Current Path : /sys/amd64/compile/hs32/modules/usr/src/sys/modules/if_gre/@/contrib/octeon-sdk/ |
FreeBSD hs32.drive.ne.jp 9.1-RELEASE FreeBSD 9.1-RELEASE #1: Wed Jan 14 12:18:08 JST 2015 root@hs32.drive.ne.jp:/sys/amd64/compile/hs32 amd64 |
Current File : //sys/amd64/compile/hs32/modules/usr/src/sys/modules/if_gre/@/contrib/octeon-sdk/cvmx-lmcx-defs.h |
/***********************license start*************** * Copyright (c) 2003-2010 Cavium Networks (support@cavium.com). All rights * reserved. * * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials provided * with the distribution. * * Neither the name of Cavium Networks nor the names of * its contributors may be used to endorse or promote products * derived from this software without specific prior written * permission. * This Software, including technical data, may be subject to U.S. export control * laws, including the U.S. Export Administration Act and its associated * regulations, and may be subject to export or import regulations in other * countries. * TO THE MAXIMUM EXTENT PERMITTED BY LAW, THE SOFTWARE IS PROVIDED "AS IS" * AND WITH ALL FAULTS AND CAVIUM NETWORKS MAKES NO PROMISES, REPRESENTATIONS OR * WARRANTIES, EITHER EXPRESS, IMPLIED, STATUTORY, OR OTHERWISE, WITH RESPECT TO * THE SOFTWARE, INCLUDING ITS CONDITION, ITS CONFORMITY TO ANY REPRESENTATION OR * DESCRIPTION, OR THE EXISTENCE OF ANY LATENT OR PATENT DEFECTS, AND CAVIUM * SPECIFICALLY DISCLAIMS ALL IMPLIED (IF ANY) WARRANTIES OF TITLE, * MERCHANTABILITY, NONINFRINGEMENT, FITNESS FOR A PARTICULAR PURPOSE, LACK OF * VIRUSES, ACCURACY OR COMPLETENESS, QUIET ENJOYMENT, QUIET POSSESSION OR * CORRESPONDENCE TO DESCRIPTION. THE ENTIRE RISK ARISING OUT OF USE OR * PERFORMANCE OF THE SOFTWARE LIES WITH YOU. ***********************license end**************************************/ /** * cvmx-lmcx-defs.h * * Configuration and status register (CSR) type definitions for * Octeon lmcx. * * This file is auto generated. Do not edit. * * <hr>$Revision$<hr> * */ #ifndef __CVMX_LMCX_TYPEDEFS_H__ #define __CVMX_LMCX_TYPEDEFS_H__ #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_BIST_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_BIST_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000F0ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_BIST_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880000F0ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_BIST_RESULT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_BIST_RESULT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000F8ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_BIST_RESULT(block_id) (CVMX_ADD_IO_SEG(0x00011800880000F8ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000220ull); } #else #define CVMX_LMCX_CHAR_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000220ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_MASK0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_MASK0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000228ull); } #else #define CVMX_LMCX_CHAR_MASK0(block_id) (CVMX_ADD_IO_SEG(0x0001180088000228ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_MASK1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_MASK1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000230ull); } #else #define CVMX_LMCX_CHAR_MASK1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000230ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_MASK2(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_MASK2(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000238ull); } #else #define CVMX_LMCX_CHAR_MASK2(block_id) (CVMX_ADD_IO_SEG(0x0001180088000238ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_MASK3(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_MASK3(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000240ull); } #else #define CVMX_LMCX_CHAR_MASK3(block_id) (CVMX_ADD_IO_SEG(0x0001180088000240ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CHAR_MASK4(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CHAR_MASK4(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000318ull); } #else #define CVMX_LMCX_CHAR_MASK4(block_id) (CVMX_ADD_IO_SEG(0x0001180088000318ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_COMP_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_COMP_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000028ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_COMP_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000028ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_COMP_CTL2(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_COMP_CTL2(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001B8ull); } #else #define CVMX_LMCX_COMP_CTL2(block_id) (CVMX_ADD_IO_SEG(0x00011800880001B8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CONFIG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CONFIG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000188ull); } #else #define CVMX_LMCX_CONFIG(block_id) (CVMX_ADD_IO_SEG(0x0001180088000188ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CONTROL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CONTROL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000190ull); } #else #define CVMX_LMCX_CONTROL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000190ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000010ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000010ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_CTL1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_CTL1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000090ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_CTL1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000090ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DCLK_CNT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DCLK_CNT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001E0ull); } #else #define CVMX_LMCX_DCLK_CNT(block_id) (CVMX_ADD_IO_SEG(0x00011800880001E0ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DCLK_CNT_HI(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DCLK_CNT_HI(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000070ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DCLK_CNT_HI(block_id) (CVMX_ADD_IO_SEG(0x0001180088000070ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DCLK_CNT_LO(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DCLK_CNT_LO(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000068ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DCLK_CNT_LO(block_id) (CVMX_ADD_IO_SEG(0x0001180088000068ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DCLK_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_DCLK_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000B8ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DCLK_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880000B8ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DDR2_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DDR2_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000018ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DDR2_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000018ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DDR_PLL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DDR_PLL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000258ull); } #else #define CVMX_LMCX_DDR_PLL_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000258ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DELAY_CFG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DELAY_CFG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000088ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DELAY_CFG(block_id) (CVMX_ADD_IO_SEG(0x0001180088000088ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DIMMX_PARAMS(unsigned long offset, unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && (((offset <= 1)) && ((block_id == 0)))))) cvmx_warn("CVMX_LMCX_DIMMX_PARAMS(%lu,%lu) is invalid on this chip\n", offset, block_id); return CVMX_ADD_IO_SEG(0x0001180088000270ull) + (((offset) & 1) + ((block_id) & 0) * 0x0ull) * 8; } #else #define CVMX_LMCX_DIMMX_PARAMS(offset, block_id) (CVMX_ADD_IO_SEG(0x0001180088000270ull) + (((offset) & 1) + ((block_id) & 0) * 0x0ull) * 8) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DIMM_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DIMM_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000310ull); } #else #define CVMX_LMCX_DIMM_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000310ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DLL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_DLL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000C0ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DLL_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880000C0ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DLL_CTL2(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DLL_CTL2(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001C8ull); } #else #define CVMX_LMCX_DLL_CTL2(block_id) (CVMX_ADD_IO_SEG(0x00011800880001C8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DLL_CTL3(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DLL_CTL3(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000218ull); } #else #define CVMX_LMCX_DLL_CTL3(block_id) (CVMX_ADD_IO_SEG(0x0001180088000218ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_DUAL_MEMCFG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_DUAL_MEMCFG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000098ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_DUAL_MEMCFG(block_id) (CVMX_ADD_IO_SEG(0x0001180088000098ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_ECC_SYND(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_ECC_SYND(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000038ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_ECC_SYND(block_id) (CVMX_ADD_IO_SEG(0x0001180088000038ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_FADR(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_FADR(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000020ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_FADR(block_id) (CVMX_ADD_IO_SEG(0x0001180088000020ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_IFB_CNT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_IFB_CNT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001D0ull); } #else #define CVMX_LMCX_IFB_CNT(block_id) (CVMX_ADD_IO_SEG(0x00011800880001D0ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_IFB_CNT_HI(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_IFB_CNT_HI(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000050ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_IFB_CNT_HI(block_id) (CVMX_ADD_IO_SEG(0x0001180088000050ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_IFB_CNT_LO(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_IFB_CNT_LO(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000048ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_IFB_CNT_LO(block_id) (CVMX_ADD_IO_SEG(0x0001180088000048ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_INT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_INT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001F0ull); } #else #define CVMX_LMCX_INT(block_id) (CVMX_ADD_IO_SEG(0x00011800880001F0ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_INT_EN(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_INT_EN(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001E8ull); } #else #define CVMX_LMCX_INT_EN(block_id) (CVMX_ADD_IO_SEG(0x00011800880001E8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_MEM_CFG0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_MEM_CFG0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000000ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_MEM_CFG0(block_id) (CVMX_ADD_IO_SEG(0x0001180088000000ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_MEM_CFG1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_MEM_CFG1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000008ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_MEM_CFG1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000008ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_MODEREG_PARAMS0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_MODEREG_PARAMS0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001A8ull); } #else #define CVMX_LMCX_MODEREG_PARAMS0(block_id) (CVMX_ADD_IO_SEG(0x00011800880001A8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_MODEREG_PARAMS1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_MODEREG_PARAMS1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000260ull); } #else #define CVMX_LMCX_MODEREG_PARAMS1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000260ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_NXM(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_NXM(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000C8ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_NXM(block_id) (CVMX_ADD_IO_SEG(0x00011800880000C8ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_OPS_CNT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_OPS_CNT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001D8ull); } #else #define CVMX_LMCX_OPS_CNT(block_id) (CVMX_ADD_IO_SEG(0x00011800880001D8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_OPS_CNT_HI(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_OPS_CNT_HI(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000060ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_OPS_CNT_HI(block_id) (CVMX_ADD_IO_SEG(0x0001180088000060ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_OPS_CNT_LO(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_OPS_CNT_LO(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000058ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_OPS_CNT_LO(block_id) (CVMX_ADD_IO_SEG(0x0001180088000058ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_PHY_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_PHY_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000210ull); } #else #define CVMX_LMCX_PHY_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000210ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_PLL_BWCTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_PLL_BWCTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000040ull); } #else #define CVMX_LMCX_PLL_BWCTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000040ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_PLL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_PLL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000A8ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_PLL_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880000A8ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_PLL_STATUS(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_PLL_STATUS(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000B0ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_PLL_STATUS(block_id) (CVMX_ADD_IO_SEG(0x00011800880000B0ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_READ_LEVEL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_READ_LEVEL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000140ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_READ_LEVEL_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000140ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_READ_LEVEL_DBG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_READ_LEVEL_DBG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000148ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_READ_LEVEL_DBG(block_id) (CVMX_ADD_IO_SEG(0x0001180088000148ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_READ_LEVEL_RANKX(unsigned long offset, unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN52XX) && (((offset <= 3)) && ((block_id == 0)))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && (((offset <= 3)) && ((block_id <= 1)))))) cvmx_warn("CVMX_LMCX_READ_LEVEL_RANKX(%lu,%lu) is invalid on this chip\n", offset, block_id); return CVMX_ADD_IO_SEG(0x0001180088000100ull) + (((offset) & 3) + ((block_id) & 1) * 0xC000000ull) * 8; } #else #define CVMX_LMCX_READ_LEVEL_RANKX(offset, block_id) (CVMX_ADD_IO_SEG(0x0001180088000100ull) + (((offset) & 3) + ((block_id) & 1) * 0xC000000ull) * 8) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RESET_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RESET_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000180ull); } #else #define CVMX_LMCX_RESET_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000180ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RLEVEL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RLEVEL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880002A0ull); } #else #define CVMX_LMCX_RLEVEL_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880002A0ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RLEVEL_DBG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RLEVEL_DBG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880002A8ull); } #else #define CVMX_LMCX_RLEVEL_DBG(block_id) (CVMX_ADD_IO_SEG(0x00011800880002A8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RLEVEL_RANKX(unsigned long offset, unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && (((offset <= 3)) && ((block_id == 0)))))) cvmx_warn("CVMX_LMCX_RLEVEL_RANKX(%lu,%lu) is invalid on this chip\n", offset, block_id); return CVMX_ADD_IO_SEG(0x0001180088000280ull) + (((offset) & 3) + ((block_id) & 0) * 0x0ull) * 8; } #else #define CVMX_LMCX_RLEVEL_RANKX(offset, block_id) (CVMX_ADD_IO_SEG(0x0001180088000280ull) + (((offset) & 3) + ((block_id) & 0) * 0x0ull) * 8) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RODT_COMP_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RODT_COMP_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880000A0ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_RODT_COMP_CTL(block_id) (CVMX_ADD_IO_SEG(0x00011800880000A0ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RODT_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RODT_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000078ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_RODT_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000078ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_RODT_MASK(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_RODT_MASK(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000268ull); } #else #define CVMX_LMCX_RODT_MASK(block_id) (CVMX_ADD_IO_SEG(0x0001180088000268ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_SLOT_CTL0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_SLOT_CTL0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001F8ull); } #else #define CVMX_LMCX_SLOT_CTL0(block_id) (CVMX_ADD_IO_SEG(0x00011800880001F8ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_SLOT_CTL1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_SLOT_CTL1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000200ull); } #else #define CVMX_LMCX_SLOT_CTL1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000200ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_SLOT_CTL2(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_SLOT_CTL2(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000208ull); } #else #define CVMX_LMCX_SLOT_CTL2(block_id) (CVMX_ADD_IO_SEG(0x0001180088000208ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_TIMING_PARAMS0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_TIMING_PARAMS0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000198ull); } #else #define CVMX_LMCX_TIMING_PARAMS0(block_id) (CVMX_ADD_IO_SEG(0x0001180088000198ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_TIMING_PARAMS1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_TIMING_PARAMS1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001A0ull); } #else #define CVMX_LMCX_TIMING_PARAMS1(block_id) (CVMX_ADD_IO_SEG(0x00011800880001A0ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_TRO_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_TRO_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000248ull); } #else #define CVMX_LMCX_TRO_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000248ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_TRO_STAT(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_TRO_STAT(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000250ull); } #else #define CVMX_LMCX_TRO_STAT(block_id) (CVMX_ADD_IO_SEG(0x0001180088000250ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WLEVEL_CTL(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_WLEVEL_CTL(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000300ull); } #else #define CVMX_LMCX_WLEVEL_CTL(block_id) (CVMX_ADD_IO_SEG(0x0001180088000300ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WLEVEL_DBG(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_WLEVEL_DBG(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000308ull); } #else #define CVMX_LMCX_WLEVEL_DBG(block_id) (CVMX_ADD_IO_SEG(0x0001180088000308ull)) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WLEVEL_RANKX(unsigned long offset, unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && (((offset <= 3)) && ((block_id == 0)))))) cvmx_warn("CVMX_LMCX_WLEVEL_RANKX(%lu,%lu) is invalid on this chip\n", offset, block_id); return CVMX_ADD_IO_SEG(0x00011800880002B0ull) + (((offset) & 3) + ((block_id) & 0) * 0x0ull) * 8; } #else #define CVMX_LMCX_WLEVEL_RANKX(offset, block_id) (CVMX_ADD_IO_SEG(0x00011800880002B0ull) + (((offset) & 3) + ((block_id) & 0) * 0x0ull) * 8) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WODT_CTL0(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN38XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN50XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))) || (OCTEON_IS_MODEL(OCTEON_CN58XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_WODT_CTL0(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000030ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_WODT_CTL0(block_id) (CVMX_ADD_IO_SEG(0x0001180088000030ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WODT_CTL1(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN30XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN31XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN52XX) && ((block_id == 0))) || (OCTEON_IS_MODEL(OCTEON_CN56XX) && ((block_id <= 1))))) cvmx_warn("CVMX_LMCX_WODT_CTL1(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x0001180088000080ull) + ((block_id) & 1) * 0x60000000ull; } #else #define CVMX_LMCX_WODT_CTL1(block_id) (CVMX_ADD_IO_SEG(0x0001180088000080ull) + ((block_id) & 1) * 0x60000000ull) #endif #if CVMX_ENABLE_CSR_ADDRESS_CHECKING static inline uint64_t CVMX_LMCX_WODT_MASK(unsigned long block_id) { if (!( (OCTEON_IS_MODEL(OCTEON_CN63XX) && ((block_id == 0))))) cvmx_warn("CVMX_LMCX_WODT_MASK(%lu) is invalid on this chip\n", block_id); return CVMX_ADD_IO_SEG(0x00011800880001B0ull); } #else #define CVMX_LMCX_WODT_MASK(block_id) (CVMX_ADD_IO_SEG(0x00011800880001B0ull)) #endif /** * cvmx_lmc#_bist_ctl * * Notes: * This controls BiST only for the memories that operate on DCLK. The normal, chip-wide BiST flow * controls BiST for the memories that operate on ECLK. */ union cvmx_lmcx_bist_ctl { uint64_t u64; struct cvmx_lmcx_bist_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_1_63 : 63; uint64_t start : 1; /**< A 0->1 transition causes BiST to run. */ #else uint64_t start : 1; uint64_t reserved_1_63 : 63; #endif } s; struct cvmx_lmcx_bist_ctl_s cn50xx; struct cvmx_lmcx_bist_ctl_s cn52xx; struct cvmx_lmcx_bist_ctl_s cn52xxp1; struct cvmx_lmcx_bist_ctl_s cn56xx; struct cvmx_lmcx_bist_ctl_s cn56xxp1; }; typedef union cvmx_lmcx_bist_ctl cvmx_lmcx_bist_ctl_t; /** * cvmx_lmc#_bist_result * * Notes: * Access to the internal BiST results * Each bit is the BiST result of an individual memory (per bit, 0=pass and 1=fail). */ union cvmx_lmcx_bist_result { uint64_t u64; struct cvmx_lmcx_bist_result_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_11_63 : 53; uint64_t csrd2e : 1; /**< BiST result of CSRD2E memory (0=pass, !0=fail) */ uint64_t csre2d : 1; /**< BiST result of CSRE2D memory (0=pass, !0=fail) */ uint64_t mwf : 1; /**< BiST result of MWF memories (0=pass, !0=fail) */ uint64_t mwd : 3; /**< BiST result of MWD memories (0=pass, !0=fail) */ uint64_t mwc : 1; /**< BiST result of MWC memories (0=pass, !0=fail) */ uint64_t mrf : 1; /**< BiST result of MRF memories (0=pass, !0=fail) */ uint64_t mrd : 3; /**< BiST result of MRD memories (0=pass, !0=fail) */ #else uint64_t mrd : 3; uint64_t mrf : 1; uint64_t mwc : 1; uint64_t mwd : 3; uint64_t mwf : 1; uint64_t csre2d : 1; uint64_t csrd2e : 1; uint64_t reserved_11_63 : 53; #endif } s; struct cvmx_lmcx_bist_result_cn50xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_9_63 : 55; uint64_t mwf : 1; /**< BiST result of MWF memories (0=pass, !0=fail) */ uint64_t mwd : 3; /**< BiST result of MWD memories (0=pass, !0=fail) */ uint64_t mwc : 1; /**< BiST result of MWC memories (0=pass, !0=fail) */ uint64_t mrf : 1; /**< BiST result of MRF memories (0=pass, !0=fail) */ uint64_t mrd : 3; /**< BiST result of MRD memories (0=pass, !0=fail) */ #else uint64_t mrd : 3; uint64_t mrf : 1; uint64_t mwc : 1; uint64_t mwd : 3; uint64_t mwf : 1; uint64_t reserved_9_63 : 55; #endif } cn50xx; struct cvmx_lmcx_bist_result_s cn52xx; struct cvmx_lmcx_bist_result_s cn52xxp1; struct cvmx_lmcx_bist_result_s cn56xx; struct cvmx_lmcx_bist_result_s cn56xxp1; }; typedef union cvmx_lmcx_bist_result cvmx_lmcx_bist_result_t; /** * cvmx_lmc#_char_ctl * * LMC_CHAR_CTL = LMC Characterization Control * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_ctl { uint64_t u64; struct cvmx_lmcx_char_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_42_63 : 22; uint64_t en : 1; /**< Enable characterization */ uint64_t sel : 1; /**< Pattern select 0 = PRBS 1 = Programmable pattern */ uint64_t prog : 8; /**< Programmable pattern */ uint64_t prbs : 32; /**< PRBS Polynomial */ #else uint64_t prbs : 32; uint64_t prog : 8; uint64_t sel : 1; uint64_t en : 1; uint64_t reserved_42_63 : 22; #endif } s; struct cvmx_lmcx_char_ctl_s cn63xx; struct cvmx_lmcx_char_ctl_s cn63xxp1; }; typedef union cvmx_lmcx_char_ctl cvmx_lmcx_char_ctl_t; /** * cvmx_lmc#_char_mask0 * * LMC_CHAR_MASK0 = LMC Characterization Mask0 * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_mask0 { uint64_t u64; struct cvmx_lmcx_char_mask0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t mask : 64; /**< Mask for DQ0[63:0] */ #else uint64_t mask : 64; #endif } s; struct cvmx_lmcx_char_mask0_s cn63xx; struct cvmx_lmcx_char_mask0_s cn63xxp1; }; typedef union cvmx_lmcx_char_mask0 cvmx_lmcx_char_mask0_t; /** * cvmx_lmc#_char_mask1 * * LMC_CHAR_MASK1 = LMC Characterization Mask1 * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_mask1 { uint64_t u64; struct cvmx_lmcx_char_mask1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_8_63 : 56; uint64_t mask : 8; /**< Mask for DQ0[71:64] */ #else uint64_t mask : 8; uint64_t reserved_8_63 : 56; #endif } s; struct cvmx_lmcx_char_mask1_s cn63xx; struct cvmx_lmcx_char_mask1_s cn63xxp1; }; typedef union cvmx_lmcx_char_mask1 cvmx_lmcx_char_mask1_t; /** * cvmx_lmc#_char_mask2 * * LMC_CHAR_MASK2 = LMC Characterization Mask2 * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_mask2 { uint64_t u64; struct cvmx_lmcx_char_mask2_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t mask : 64; /**< Mask for DQ1[63:0] */ #else uint64_t mask : 64; #endif } s; struct cvmx_lmcx_char_mask2_s cn63xx; struct cvmx_lmcx_char_mask2_s cn63xxp1; }; typedef union cvmx_lmcx_char_mask2 cvmx_lmcx_char_mask2_t; /** * cvmx_lmc#_char_mask3 * * LMC_CHAR_MASK3 = LMC Characterization Mask3 * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_mask3 { uint64_t u64; struct cvmx_lmcx_char_mask3_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_8_63 : 56; uint64_t mask : 8; /**< Mask for DQ1[71:64] */ #else uint64_t mask : 8; uint64_t reserved_8_63 : 56; #endif } s; struct cvmx_lmcx_char_mask3_s cn63xx; struct cvmx_lmcx_char_mask3_s cn63xxp1; }; typedef union cvmx_lmcx_char_mask3 cvmx_lmcx_char_mask3_t; /** * cvmx_lmc#_char_mask4 * * LMC_CHAR_MASK4 = LMC Characterization Mask4 * This register is an assortment of various control fields needed to charecterize the DDR3 interface */ union cvmx_lmcx_char_mask4 { uint64_t u64; struct cvmx_lmcx_char_mask4_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_33_63 : 31; uint64_t reset_n_mask : 1; /**< Mask for RESET_L */ uint64_t a_mask : 16; /**< Mask for A[15:0] */ uint64_t ba_mask : 3; /**< Mask for BA[2:0] */ uint64_t we_n_mask : 1; /**< Mask for WE_N */ uint64_t cas_n_mask : 1; /**< Mask for CAS_N */ uint64_t ras_n_mask : 1; /**< Mask for RAS_N */ uint64_t odt1_mask : 2; /**< Mask for ODT1 */ uint64_t odt0_mask : 2; /**< Mask for ODT0 */ uint64_t cs1_n_mask : 2; /**< Mask for CS1_N */ uint64_t cs0_n_mask : 2; /**< Mask for CS0_N */ uint64_t cke_mask : 2; /**< Mask for CKE* */ #else uint64_t cke_mask : 2; uint64_t cs0_n_mask : 2; uint64_t cs1_n_mask : 2; uint64_t odt0_mask : 2; uint64_t odt1_mask : 2; uint64_t ras_n_mask : 1; uint64_t cas_n_mask : 1; uint64_t we_n_mask : 1; uint64_t ba_mask : 3; uint64_t a_mask : 16; uint64_t reset_n_mask : 1; uint64_t reserved_33_63 : 31; #endif } s; struct cvmx_lmcx_char_mask4_s cn63xx; struct cvmx_lmcx_char_mask4_s cn63xxp1; }; typedef union cvmx_lmcx_char_mask4 cvmx_lmcx_char_mask4_t; /** * cvmx_lmc#_comp_ctl * * LMC_COMP_CTL = LMC Compensation control * */ union cvmx_lmcx_comp_ctl { uint64_t u64; struct cvmx_lmcx_comp_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t nctl_csr : 4; /**< Compensation control bits */ uint64_t nctl_clk : 4; /**< Compensation control bits */ uint64_t nctl_cmd : 4; /**< Compensation control bits */ uint64_t nctl_dat : 4; /**< Compensation control bits */ uint64_t pctl_csr : 4; /**< Compensation control bits */ uint64_t pctl_clk : 4; /**< Compensation control bits */ uint64_t reserved_0_7 : 8; #else uint64_t reserved_0_7 : 8; uint64_t pctl_clk : 4; uint64_t pctl_csr : 4; uint64_t nctl_dat : 4; uint64_t nctl_cmd : 4; uint64_t nctl_clk : 4; uint64_t nctl_csr : 4; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_comp_ctl_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t nctl_csr : 4; /**< Compensation control bits */ uint64_t nctl_clk : 4; /**< Compensation control bits */ uint64_t nctl_cmd : 4; /**< Compensation control bits */ uint64_t nctl_dat : 4; /**< Compensation control bits */ uint64_t pctl_csr : 4; /**< Compensation control bits */ uint64_t pctl_clk : 4; /**< Compensation control bits */ uint64_t pctl_cmd : 4; /**< Compensation control bits */ uint64_t pctl_dat : 4; /**< Compensation control bits */ #else uint64_t pctl_dat : 4; uint64_t pctl_cmd : 4; uint64_t pctl_clk : 4; uint64_t pctl_csr : 4; uint64_t nctl_dat : 4; uint64_t nctl_cmd : 4; uint64_t nctl_clk : 4; uint64_t nctl_csr : 4; uint64_t reserved_32_63 : 32; #endif } cn30xx; struct cvmx_lmcx_comp_ctl_cn30xx cn31xx; struct cvmx_lmcx_comp_ctl_cn30xx cn38xx; struct cvmx_lmcx_comp_ctl_cn30xx cn38xxp2; struct cvmx_lmcx_comp_ctl_cn50xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t nctl_csr : 4; /**< Compensation control bits */ uint64_t reserved_20_27 : 8; uint64_t nctl_dat : 4; /**< Compensation control bits */ uint64_t pctl_csr : 4; /**< Compensation control bits */ uint64_t reserved_5_11 : 7; uint64_t pctl_dat : 5; /**< Compensation control bits */ #else uint64_t pctl_dat : 5; uint64_t reserved_5_11 : 7; uint64_t pctl_csr : 4; uint64_t nctl_dat : 4; uint64_t reserved_20_27 : 8; uint64_t nctl_csr : 4; uint64_t reserved_32_63 : 32; #endif } cn50xx; struct cvmx_lmcx_comp_ctl_cn50xx cn52xx; struct cvmx_lmcx_comp_ctl_cn50xx cn52xxp1; struct cvmx_lmcx_comp_ctl_cn50xx cn56xx; struct cvmx_lmcx_comp_ctl_cn50xx cn56xxp1; struct cvmx_lmcx_comp_ctl_cn50xx cn58xx; struct cvmx_lmcx_comp_ctl_cn58xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t nctl_csr : 4; /**< Compensation control bits */ uint64_t reserved_20_27 : 8; uint64_t nctl_dat : 4; /**< Compensation control bits */ uint64_t pctl_csr : 4; /**< Compensation control bits */ uint64_t reserved_4_11 : 8; uint64_t pctl_dat : 4; /**< Compensation control bits */ #else uint64_t pctl_dat : 4; uint64_t reserved_4_11 : 8; uint64_t pctl_csr : 4; uint64_t nctl_dat : 4; uint64_t reserved_20_27 : 8; uint64_t nctl_csr : 4; uint64_t reserved_32_63 : 32; #endif } cn58xxp1; }; typedef union cvmx_lmcx_comp_ctl cvmx_lmcx_comp_ctl_t; /** * cvmx_lmc#_comp_ctl2 * * LMC_COMP_CTL2 = LMC Compensation control * */ union cvmx_lmcx_comp_ctl2 { uint64_t u64; struct cvmx_lmcx_comp_ctl2_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_34_63 : 30; uint64_t ddr__ptune : 4; /**< DDR PCTL from compensation circuit The encoded value provides debug information for the compensation impedance on P-pullup */ uint64_t ddr__ntune : 4; /**< DDR NCTL from compensation circuit The encoded value provides debug information for the compensation impedance on N-pulldown */ uint64_t m180 : 1; /**< Cap impedance at 180 Ohm (instead of 240 Ohm) */ uint64_t byp : 1; /**< Bypass mode When set, PTUNE,NTUNE are the compensation setting. When clear, DDR_PTUNE,DDR_NTUNE are the compensation setting. */ uint64_t ptune : 4; /**< PCTL impedance control in bypass mode */ uint64_t ntune : 4; /**< NCTL impedance control in bypass mode */ uint64_t rodt_ctl : 4; /**< NCTL RODT impedance control bits This field controls ODT values during a memory read on the Octeon side 0000 = No ODT 0001 = 20 ohm 0010 = 30 ohm 0011 = 40 ohm 0100 = 60 ohm 0101 = 120 ohm 0110-1111 = Reserved */ uint64_t cmd_ctl : 4; /**< Drive strength control for CMD/A/RESET_L/CKE* drivers 0001 = 24 ohm 0010 = 26.67 ohm 0011 = 30 ohm 0100 = 34.3 ohm 0101 = 40 ohm 0110 = 48 ohm 0111 = 60 ohm 0000,1000-1111 = Reserved */ uint64_t ck_ctl : 4; /**< Drive strength control for CK/CS*_L/ODT drivers 0001 = 24 ohm 0010 = 26.67 ohm 0011 = 30 ohm 0100 = 34.3 ohm 0101 = 40 ohm 0110 = 48 ohm 0111 = 60 ohm 0000,1000-1111 = Reserved */ uint64_t dqx_ctl : 4; /**< Drive strength control for DQ/DQS drivers 0001 = 24 ohm 0010 = 26.67 ohm 0011 = 30 ohm 0100 = 34.3 ohm 0101 = 40 ohm 0110 = 48 ohm 0111 = 60 ohm 0000,1000-1111 = Reserved */ #else uint64_t dqx_ctl : 4; uint64_t ck_ctl : 4; uint64_t cmd_ctl : 4; uint64_t rodt_ctl : 4; uint64_t ntune : 4; uint64_t ptune : 4; uint64_t byp : 1; uint64_t m180 : 1; uint64_t ddr__ntune : 4; uint64_t ddr__ptune : 4; uint64_t reserved_34_63 : 30; #endif } s; struct cvmx_lmcx_comp_ctl2_s cn63xx; struct cvmx_lmcx_comp_ctl2_s cn63xxp1; }; typedef union cvmx_lmcx_comp_ctl2 cvmx_lmcx_comp_ctl2_t; /** * cvmx_lmc#_config * * LMC_CONFIG = LMC Configuration Register * * This register controls certain parameters of Memory Configuration * * Notes: * a. Priority order for hardware writes to LMC*_CONFIG/LMC*_FADR/LMC*_ECC_SYND: DED error >= NXM error > SEC error * b. The self refresh entry sequence(s) power the DLL up/down (depending on LMC*_MODEREG_PARAMS0[DLL]) * when LMC*_CONFIG[SREF_WITH_DLL] is set * c. Prior to the self-refresh exit sequence, LMC*_MODEREG_PARAMS0 and LMC*_MODEREG_PARAMS1 should be re-programmed (if needed) to the * appropriate values * * LMC Bringup Sequence: * 1. SW must ensure there are no pending DRAM transactions and that the DDR PLL and the DLL have been initialized. * 2. Write LMC*_COMP_CTL2, LMC*_CONTROL, LMC*_WODT_MASK, LMC*_DUAL_MEMCFG, LMC*_TIMING_PARAMS0, LMC*_TIMING_PARAMS1, * LMC*_MODEREG_PARAMS0, LMC*_MODEREG_PARAMS1, LMC*_RESET_CTL (with DDR3RST=0), LMC*_CONFIG (with INIT_START=0) * with appropriate values, if necessary. * 3. Wait 200us, then write LMC*_RESET_CTL[DDR3RST] = 1. * 4. Initialize all ranks at once by writing LMC*_CONFIG[RANKMASK][n] = 1, LMC*_CONFIG[INIT_STATUS][n] = 1, and LMC*_CONFIG[INIT_START] = 1 * where n is a valid rank index for the specific board configuration. * 5. for each rank n to be write-leveled [ * if auto write-leveling is desired [ * write LMC*_CONFIG[RANKMASK][n] = 1, LMC*_WLEVEL_CTL appropriately and LMC*_CONFIG[INIT_START] = 1 * wait until LMC*_WLEVEL_RANKn[STATUS] = 3 * ] else [ * write LMC*_WLEVEL_RANKn with appropriate values * ] * ] * 6. for each rank n to be read-leveled [ * if auto read-leveling is desired [ * write LMC*_CONFIG[RANKMASK][n] = 1, LMC*_RLEVEL_CTL appropriately and LMC*_CONFIG[INIT_START] = 1 * wait until LMC*_RLEVEL_RANKn[STATUS] = 3 * ] else [ * write LMC*_RLEVEL_RANKn with appropriate values * ] * ] */ union cvmx_lmcx_config { uint64_t u64; struct cvmx_lmcx_config_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_59_63 : 5; uint64_t early_unload_d1_r1 : 1; /**< When set, unload the PHY silo one cycle early for Rank 3 reads The recommended EARLY_UNLOAD_D1_R1 value can be calculated after the final LMC*_RLEVEL_RANK3[BYTE*] values are selected (as part of read-leveling initialization). Then, determine the largest read-leveling setting for rank 3 (i.e. calculate maxset=MAX(LMC*_RLEVEL_RANK3[BYTEi]) across all i), then set EARLY_UNLOAD_D1_R1 when the low two bits of this largest setting is not 3 (i.e. EARLY_UNLOAD_D1_R1 = (maxset<1:0>!=3)). */ uint64_t early_unload_d1_r0 : 1; /**< When set, unload the PHY silo one cycle early for Rank 2 reads The recommended EARLY_UNLOAD_D1_RO value can be calculated after the final LMC*_RLEVEL_RANK2[BYTE*] values are selected (as part of read-leveling initialization). Then, determine the largest read-leveling setting for rank 2 (i.e. calculate maxset=MAX(LMC*_RLEVEL_RANK2[BYTEi]) across all i), then set EARLY_UNLOAD_D1_RO when the low two bits of this largest setting is not 3 (i.e. EARLY_UNLOAD_D1_RO = (maxset<1:0>!=3)). */ uint64_t early_unload_d0_r1 : 1; /**< When set, unload the PHY silo one cycle early for Rank 1 reads The recommended EARLY_UNLOAD_D0_R1 value can be calculated after the final LMC*_RLEVEL_RANK1[BYTE*] values are selected (as part of read-leveling initialization). Then, determine the largest read-leveling setting for rank 1 (i.e. calculate maxset=MAX(LMC*_RLEVEL_RANK1[BYTEi]) across all i), then set EARLY_UNLOAD_D0_R1 when the low two bits of this largest setting is not 3 (i.e. EARLY_UNLOAD_D0_R1 = (maxset<1:0>!=3)). */ uint64_t early_unload_d0_r0 : 1; /**< When set, unload the PHY silo one cycle early for Rank 0 reads. The recommended EARLY_UNLOAD_D0_R0 value can be calculated after the final LMC*_RLEVEL_RANK0[BYTE*] values are selected (as part of read-leveling initialization). Then, determine the largest read-leveling setting for rank 0 (i.e. calculate maxset=MAX(LMC*_RLEVEL_RANK0[BYTEi]) across all i), then set EARLY_UNLOAD_D0_R0 when the low two bits of this largest setting is not 3 (i.e. EARLY_UNLOAD_D0_R0 = (maxset<1:0>!=3)). */ uint64_t init_status : 4; /**< Indicates status of initialization INIT_STATUS[n] = 1 implies rank n has been initialized SW must set necessary INIT_STATUS bits with the same LMC*_CONFIG write that initiates power-up/init and self-refresh exit sequences (if the required INIT_STATUS bits are not already set before LMC initiates the sequence). INIT_STATUS determines the chip-selects that assert during refresh, ZQCS, and precharge power-down and self-refresh entry/exit SEQUENCE's. */ uint64_t mirrmask : 4; /**< Mask determining which ranks are address-mirrored. MIRRMASK<n> = 1 means Rank n addresses are mirrored for 0 <= n <= 3 A mirrored read/write has these differences: - DDR_BA<1> is swapped with DDR_BA<0> - DDR_A<8> is swapped with DDR_A<7> - DDR_A<6> is swapped with DDR_A<5> - DDR_A<4> is swapped with DDR_A<3> When RANK_ENA=0, MIRRMASK<1> and MIRRMASK<3> MBZ */ uint64_t rankmask : 4; /**< Mask to select rank to be leveled/initialized. To write-level/read-level/initialize rank i, set RANKMASK<i> RANK_ENA=1 RANK_ENA=0 RANKMASK<0> = DIMM0_CS0 DIMM0_CS0 RANKMASK<1> = DIMM0_CS1 MBZ RANKMASK<2> = DIMM1_CS0 DIMM1_CS0 RANKMASK<3> = DIMM1_CS1 MBZ For read/write leveling, each rank has to be leveled separately, so RANKMASK should only have one bit set. RANKMASK is not used during self-refresh entry/exit and precharge power-down entry/exit instruction sequences. When RANK_ENA=0, RANKMASK<1> and RANKMASK<3> MBZ */ uint64_t rank_ena : 1; /**< RANK ena (for use with dual-rank DIMMs) For dual-rank DIMMs, the rank_ena bit will enable the drive of the CS*_L[1:0] and ODT_<1:0> pins differently based on the (pbank_lsb-1) address bit. Write 0 for SINGLE ranked DIMM's. */ uint64_t sref_with_dll : 1; /**< Self-refresh entry/exit write MR1 and MR2 When set, self-refresh entry and exit instruction sequences write MR1 and MR2 (in all ranks). (The writes occur before self-refresh entry, and after self-refresh exit.) When clear, self-refresh entry and exit instruction sequences do not write any registers in the DDR3 parts. */ uint64_t early_dqx : 1; /**< Send DQx signals one CK cycle earlier for the case when the shortest DQx lines have a larger delay than the CK line */ uint64_t sequence : 3; /**< Selects the sequence that LMC runs after a 0->1 transition on LMC*_CONFIG[INIT_START]. SEQUENCE=0=power-up/init: - RANKMASK selects participating ranks (should be all ranks with attached DRAM) - INIT_STATUS must equal RANKMASK - DDR_CKE* signals activated (if they weren't already active) - RDIMM register control words 0-15 will be written to RANKMASK-selected RDIMM's when LMC(0)_CONTROL[RDIMM_ENA]=1 and corresponding LMC*_DIMM_CTL[DIMM*_WMASK] bits are set. (Refer to LMC*_DIMM*_PARAMS and LMC*_DIMM_CTL descriptions below for more details.) - MR0, MR1, MR2, and MR3 will be written to selected ranks SEQUENCE=1=read-leveling: - RANKMASK selects the rank to be read-leveled - MR3 written to selected rank SEQUENCE=2=self-refresh entry: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - MR1 and MR2 will be written to selected ranks if SREF_WITH_DLL=1 - DDR_CKE* signals de-activated SEQUENCE=3=self-refresh exit: - INIT_STATUS must be set to indicate participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals activated - MR0, MR1, MR2, and MR3 will be written to participating ranks if SREF_WITH_DLL=1 SEQUENCE=4=precharge power-down entry: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals de-activated SEQUENCE=5=precharge power-down exit: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals activated SEQUENCE=6=write-leveling: - RANKMASK selects the rank to be write-leveled - INIT_STATUS must indicate all ranks with attached DRAM - MR1 and MR2 written to INIT_STATUS-selected ranks SEQUENCE=7=illegal Precharge power-down entry and exit SEQUENCE's may also be automatically generated by the HW when IDLEPOWER!=0. Self-refresh entry SEQUENCE's may also be automatically generated by hardware upon a chip warm or soft reset sequence when LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT] are set. LMC writes the LMC*_MODEREG_PARAMS0 and LMC*_MODEREG_PARAMS1 CSR field values to the Mode registers in the DRAM parts (i.e. MR0, MR1, MR2, and MR3) as part of some of these sequences. Refer to the LMC*_MODEREG_PARAMS0 and LMC*_MODEREG_PARAMS1 descriptions for more details. If there are two consecutive power-up/init's without a DRESET assertion between them, LMC asserts DDR_CKE* as part of the first power-up/init, and continues to assert DDR_CKE* through the remainder of the first and the second power-up/init. If DDR_CKE* deactivation and reactivation is needed for a second power-up/init, a DRESET assertion is required between the first and the second. */ uint64_t ref_zqcs_int : 19; /**< Refresh & ZQCS interval represented in \#of 512 CK cycle increments. A Refresh sequence is triggered when bits [24:18] are equal to 0, and a ZQCS sequence is triggered when [36:18] are equal to 0. Program [24:18] to RND-DN(tREFI/clkPeriod/512) Program [36:25] to RND-DN(ZQCS_Interval/clkPeriod/(512*64)). Note that this value should always be greater than 32, to account for resistor calibration delays. 000_00000000_00000000: RESERVED Max Refresh interval = 127 * 512 = 65024 CKs Max ZQCS interval = (8*256*256-1) * 512 = 268434944 CKs ~ 335ms for a 800 MHz CK LMC*_CONFIG[INIT_STATUS] determines which ranks receive the REF / ZQCS. LMC does not send any refreshes / ZQCS's when LMC*_CONFIG[INIT_STATUS]=0. */ uint64_t reset : 1; /**< Reset oneshot pulse for refresh counter, and LMC*_OPS_CNT, LMC*_IFB_CNT, and LMC*_DCLK_CNT CSR's. SW should write this to a one, then re-write it to a zero to cause the reset. */ uint64_t ecc_adr : 1; /**< Include memory reference address in the ECC calculation 0=disabled, 1=enabled */ uint64_t forcewrite : 4; /**< Force the oldest outstanding write to complete after having waited for 2^FORCEWRITE CK cycles. 0=disabled. */ uint64_t idlepower : 3; /**< Enter precharge power-down mode after the memory controller has been idle for 2^(2+IDLEPOWER) CK cycles. 0=disabled. This field should only be programmed after initialization. LMC*_MODEREG_PARAMS0[PPD] determines whether the DRAM DLL is disabled during the precharge power-down. */ uint64_t pbank_lsb : 4; /**< DIMM address bit select Reverting to the explanation for ROW_LSB, PBank_LSB would be Row_LSB bit + \#rowbits + \#rankbits Decoding for pbank_lsb - 0000:DIMM = mem_adr[28] / rank = mem_adr[27] (if RANK_ENA) - 0001:DIMM = mem_adr[29] / rank = mem_adr[28] " - 0010:DIMM = mem_adr[30] / rank = mem_adr[29] " - 0011:DIMM = mem_adr[31] / rank = mem_adr[30] " - 0100:DIMM = mem_adr[32] / rank = mem_adr[31] " - 0101:DIMM = mem_adr[33] / rank = mem_adr[32] " - 0110:DIMM = mem_adr[34] / rank = mem_adr[33] " - 0111:DIMM = 0 / rank = mem_adr[34] " - 1000-1111: RESERVED For example, for a DIMM made of Samsung's k4b1g0846c-f7 1Gb (16M x 8 bit x 8 bank) DDR3 parts, the column address width = 10, so with 10b of col, 3b of bus, 3b of bank, row_lsb = 16. So, row = mem_adr[29:16] With rank_ena = 0, pbank_lsb = 2 With rank_ena = 1, pbank_lsb = 3 */ uint64_t row_lsb : 3; /**< Row Address bit select Encoding used to determine which memory address bit position represents the low order DDR ROW address. The processor's memory address[34:7] needs to be translated to DRAM addresses (bnk,row,col,rank and DIMM) and that is a function of the following: 1. Datapath Width (64) 2. \# Banks (8) 3. \# Column Bits of the memory part - spec'd indirectly by this register. 4. \# Row Bits of the memory part - spec'd indirectly 5. \# Ranks in a DIMM - spec'd by RANK_ENA 6. \# DIMM's in the system by the register below (PBANK_LSB). Decoding for row_lsb - 000: row_lsb = mem_adr[14] - 001: row_lsb = mem_adr[15] - 010: row_lsb = mem_adr[16] - 011: row_lsb = mem_adr[17] - 100: row_lsb = mem_adr[18] - 101: row_lsb = mem_adr[19] - 110: row_lsb = mem_adr[20] - 111: RESERVED For example, for a DIMM made of Samsung's k4b1g0846c-f7 1Gb (16M x 8 bit x 8 bank) DDR3 parts, the column address width = 10, so with 10b of col, 3b of bus, 3b of bank, row_lsb = 16. So, row = mem_adr[29:16] */ uint64_t ecc_ena : 1; /**< ECC Enable: When set will enable the 8b ECC check/correct logic. Should be 1 when used with DIMMs with ECC. 0, otherwise. When this mode is turned on, DQ[71:64] on writes, will contain the ECC code generated for the 64 bits of data which will written in the memory and then later on reads, used to check for Single bit error (which will be auto- corrected) and Double Bit error (which will be reported). When not turned on, DQ[71:64] are driven to 0. Please refer to SEC_ERR, DED_ERR, LMC*_FADR, and LMC*_ECC_SYND registers for diagnostics information when there is an error. */ uint64_t init_start : 1; /**< A 0->1 transition starts the DDR memory sequence that is selected by LMC*_CONFIG[SEQUENCE]. This register is a oneshot and clears itself each time it is set. */ #else uint64_t init_start : 1; uint64_t ecc_ena : 1; uint64_t row_lsb : 3; uint64_t pbank_lsb : 4; uint64_t idlepower : 3; uint64_t forcewrite : 4; uint64_t ecc_adr : 1; uint64_t reset : 1; uint64_t ref_zqcs_int : 19; uint64_t sequence : 3; uint64_t early_dqx : 1; uint64_t sref_with_dll : 1; uint64_t rank_ena : 1; uint64_t rankmask : 4; uint64_t mirrmask : 4; uint64_t init_status : 4; uint64_t early_unload_d0_r0 : 1; uint64_t early_unload_d0_r1 : 1; uint64_t early_unload_d1_r0 : 1; uint64_t early_unload_d1_r1 : 1; uint64_t reserved_59_63 : 5; #endif } s; struct cvmx_lmcx_config_s cn63xx; struct cvmx_lmcx_config_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_55_63 : 9; uint64_t init_status : 4; /**< Indicates status of initialization INIT_STATUS[n] = 1 implies rank n has been initialized SW must set necessary INIT_STATUS bits with the same LMC*_CONFIG write that initiates power-up/init and self-refresh exit sequences (if the required INIT_STATUS bits are not already set before LMC initiates the sequence). INIT_STATUS determines the chip-selects that assert during refresh, ZQCS, and precharge power-down and self-refresh entry/exit SEQUENCE's. */ uint64_t mirrmask : 4; /**< Mask determining which ranks are address-mirrored. MIRRMASK<n> = 1 means Rank n addresses are mirrored for 0 <= n <= 3 A mirrored read/write has these differences: - DDR_BA<1> is swapped with DDR_BA<0> - DDR_A<8> is swapped with DDR_A<7> - DDR_A<6> is swapped with DDR_A<5> - DDR_A<4> is swapped with DDR_A<3> When RANK_ENA=0, MIRRMASK<1> and MIRRMASK<3> MBZ */ uint64_t rankmask : 4; /**< Mask to select rank to be leveled/initialized. To write-level/read-level/initialize rank i, set RANKMASK<i> RANK_ENA=1 RANK_ENA=0 RANKMASK<0> = DIMM0_CS0 DIMM0_CS0 RANKMASK<1> = DIMM0_CS1 MBZ RANKMASK<2> = DIMM1_CS0 DIMM1_CS0 RANKMASK<3> = DIMM1_CS1 MBZ For read/write leveling, each rank has to be leveled separately, so RANKMASK should only have one bit set. RANKMASK is not used during self-refresh entry/exit and precharge power-down entry/exit instruction sequences. When RANK_ENA=0, RANKMASK<1> and RANKMASK<3> MBZ */ uint64_t rank_ena : 1; /**< RANK ena (for use with dual-rank DIMMs) For dual-rank DIMMs, the rank_ena bit will enable the drive of the CS*_L[1:0] and ODT_<1:0> pins differently based on the (pbank_lsb-1) address bit. Write 0 for SINGLE ranked DIMM's. */ uint64_t sref_with_dll : 1; /**< Self-refresh entry/exit write MR1 and MR2 When set, self-refresh entry and exit instruction sequences write MR1 and MR2 (in all ranks). (The writes occur before self-refresh entry, and after self-refresh exit.) When clear, self-refresh entry and exit instruction sequences do not write any registers in the DDR3 parts. */ uint64_t early_dqx : 1; /**< Send DQx signals one CK cycle earlier for the case when the shortest DQx lines have a larger delay than the CK line */ uint64_t sequence : 3; /**< Selects the sequence that LMC runs after a 0->1 transition on LMC*_CONFIG[INIT_START]. SEQUENCE=0=power-up/init: - RANKMASK selects participating ranks (should be all ranks with attached DRAM) - INIT_STATUS must equal RANKMASK - DDR_CKE* signals activated (if they weren't already active) - RDIMM register control words 0-15 will be written to RANKMASK-selected RDIMM's when LMC(0)_CONTROL[RDIMM_ENA]=1 and corresponding LMC*_DIMM_CTL[DIMM*_WMASK] bits are set. (Refer to LMC*_DIMM*_PARAMS and LMC*_DIMM_CTL descriptions below for more details.) - MR0, MR1, MR2, and MR3 will be written to selected ranks SEQUENCE=1=read-leveling: - RANKMASK selects the rank to be read-leveled - MR3 written to selected rank SEQUENCE=2=self-refresh entry: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - MR1 and MR2 will be written to selected ranks if SREF_WITH_DLL=1 - DDR_CKE* signals de-activated SEQUENCE=3=self-refresh exit: - INIT_STATUS must be set to indicate participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals activated - MR0, MR1, MR2, and MR3 will be written to participating ranks if SREF_WITH_DLL=1 SEQUENCE=4=precharge power-down entry: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals de-activated SEQUENCE=5=precharge power-down exit: - INIT_STATUS selects participating ranks (should be all ranks with attached DRAM) - DDR_CKE* signals activated SEQUENCE=6=write-leveling: - RANKMASK selects the rank to be write-leveled - INIT_STATUS must indicate all ranks with attached DRAM - MR1 and MR2 written to INIT_STATUS-selected ranks SEQUENCE=7=illegal Precharge power-down entry and exit SEQUENCE's may also be automatically generated by the HW when IDLEPOWER!=0. Self-refresh entry SEQUENCE's may also be automatically generated by hardware upon a chip warm or soft reset sequence when LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT] are set. LMC writes the LMC*_MODEREG_PARAMS0 and LMC*_MODEREG_PARAMS1 CSR field values to the Mode registers in the DRAM parts (i.e. MR0, MR1, MR2, and MR3) as part of some of these sequences. Refer to the LMC*_MODEREG_PARAMS0 and LMC*_MODEREG_PARAMS1 descriptions for more details. If there are two consecutive power-up/init's without a DRESET assertion between them, LMC asserts DDR_CKE* as part of the first power-up/init, and continues to assert DDR_CKE* through the remainder of the first and the second power-up/init. If DDR_CKE* deactivation and reactivation is needed for a second power-up/init, a DRESET assertion is required between the first and the second. */ uint64_t ref_zqcs_int : 19; /**< Refresh & ZQCS interval represented in \#of 512 CK cycle increments. A Refresh sequence is triggered when bits [24:18] are equal to 0, and a ZQCS sequence is triggered when [36:18] are equal to 0. Program [24:18] to RND-DN(tREFI/clkPeriod/512) Program [36:25] to RND-DN(ZQCS_Interval/clkPeriod/(512*64)). Note that this value should always be greater than 32, to account for resistor calibration delays. 000_00000000_00000000: RESERVED Max Refresh interval = 127 * 512 = 65024 CKs Max ZQCS interval = (8*256*256-1) * 512 = 268434944 CKs ~ 335ms for a 800 MHz CK LMC*_CONFIG[INIT_STATUS] determines which ranks receive the REF / ZQCS. LMC does not send any refreshes / ZQCS's when LMC*_CONFIG[INIT_STATUS]=0. */ uint64_t reset : 1; /**< Reset oneshot pulse for refresh counter, and LMC*_OPS_CNT, LMC*_IFB_CNT, and LMC*_DCLK_CNT CSR's. SW should write this to a one, then re-write it to a zero to cause the reset. */ uint64_t ecc_adr : 1; /**< Include memory reference address in the ECC calculation 0=disabled, 1=enabled */ uint64_t forcewrite : 4; /**< Force the oldest outstanding write to complete after having waited for 2^FORCEWRITE CK cycles. 0=disabled. */ uint64_t idlepower : 3; /**< Enter precharge power-down mode after the memory controller has been idle for 2^(2+IDLEPOWER) CK cycles. 0=disabled. This field should only be programmed after initialization. LMC*_MODEREG_PARAMS0[PPD] determines whether the DRAM DLL is disabled during the precharge power-down. */ uint64_t pbank_lsb : 4; /**< DIMM address bit select Reverting to the explanation for ROW_LSB, PBank_LSB would be Row_LSB bit + \#rowbits + \#rankbits Decoding for pbank_lsb - 0000:DIMM = mem_adr[28] / rank = mem_adr[27] (if RANK_ENA) - 0001:DIMM = mem_adr[29] / rank = mem_adr[28] " - 0010:DIMM = mem_adr[30] / rank = mem_adr[29] " - 0011:DIMM = mem_adr[31] / rank = mem_adr[30] " - 0100:DIMM = mem_adr[32] / rank = mem_adr[31] " - 0101:DIMM = mem_adr[33] / rank = mem_adr[32] " - 0110:DIMM = mem_adr[34] / rank = mem_adr[33] " - 0111:DIMM = 0 / rank = mem_adr[34] " - 1000-1111: RESERVED For example, for a DIMM made of Samsung's k4b1g0846c-f7 1Gb (16M x 8 bit x 8 bank) DDR3 parts, the column address width = 10, so with 10b of col, 3b of bus, 3b of bank, row_lsb = 16. So, row = mem_adr[29:16] With rank_ena = 0, pbank_lsb = 2 With rank_ena = 1, pbank_lsb = 3 */ uint64_t row_lsb : 3; /**< Row Address bit select Encoding used to determine which memory address bit position represents the low order DDR ROW address. The processor's memory address[34:7] needs to be translated to DRAM addresses (bnk,row,col,rank and DIMM) and that is a function of the following: 1. Datapath Width (64) 2. \# Banks (8) 3. \# Column Bits of the memory part - spec'd indirectly by this register. 4. \# Row Bits of the memory part - spec'd indirectly 5. \# Ranks in a DIMM - spec'd by RANK_ENA 6. \# DIMM's in the system by the register below (PBANK_LSB). Decoding for row_lsb - 000: row_lsb = mem_adr[14] - 001: row_lsb = mem_adr[15] - 010: row_lsb = mem_adr[16] - 011: row_lsb = mem_adr[17] - 100: row_lsb = mem_adr[18] - 101: row_lsb = mem_adr[19] - 110: row_lsb = mem_adr[20] - 111: RESERVED For example, for a DIMM made of Samsung's k4b1g0846c-f7 1Gb (16M x 8 bit x 8 bank) DDR3 parts, the column address width = 10, so with 10b of col, 3b of bus, 3b of bank, row_lsb = 16. So, row = mem_adr[29:16] */ uint64_t ecc_ena : 1; /**< ECC Enable: When set will enable the 8b ECC check/correct logic. Should be 1 when used with DIMMs with ECC. 0, otherwise. When this mode is turned on, DQ[71:64] on writes, will contain the ECC code generated for the 64 bits of data which will written in the memory and then later on reads, used to check for Single bit error (which will be auto- corrected) and Double Bit error (which will be reported). When not turned on, DQ[71:64] are driven to 0. Please refer to SEC_ERR, DED_ERR, LMC*_FADR, and LMC*_ECC_SYND registers for diagnostics information when there is an error. */ uint64_t init_start : 1; /**< A 0->1 transition starts the DDR memory sequence that is selected by LMC*_CONFIG[SEQUENCE]. This register is a oneshot and clears itself each time it is set. */ #else uint64_t init_start : 1; uint64_t ecc_ena : 1; uint64_t row_lsb : 3; uint64_t pbank_lsb : 4; uint64_t idlepower : 3; uint64_t forcewrite : 4; uint64_t ecc_adr : 1; uint64_t reset : 1; uint64_t ref_zqcs_int : 19; uint64_t sequence : 3; uint64_t early_dqx : 1; uint64_t sref_with_dll : 1; uint64_t rank_ena : 1; uint64_t rankmask : 4; uint64_t mirrmask : 4; uint64_t init_status : 4; uint64_t reserved_55_63 : 9; #endif } cn63xxp1; }; typedef union cvmx_lmcx_config cvmx_lmcx_config_t; /** * cvmx_lmc#_control * * LMC_CONTROL = LMC Control * This register is an assortment of various control fields needed by the memory controller */ union cvmx_lmcx_control { uint64_t u64; struct cvmx_lmcx_control_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_24_63 : 40; uint64_t rodt_bprch : 1; /**< When set, the turn-off time for the ODT pin during a RD cmd is delayed an additional CK cycle. */ uint64_t wodt_bprch : 1; /**< When set, the turn-off time for the ODT pin during a WR cmd is delayed an additional CK cycle. */ uint64_t bprch : 2; /**< Back Porch Enable: When set, the turn-on time for the default DDR_DQ/DQS drivers is delayed an additional BPRCH CK cycles. 00 = 0 CKs 01 = 1 CKs 10 = 2 CKs 11 = 3 CKs */ uint64_t ext_zqcs_dis : 1; /**< Disable (external) auto-zqcs calibration When clear, LMC runs external ZQ calibration every LMC*_CONFIG[REF_ZQCS_INT] CK cycles. */ uint64_t int_zqcs_dis : 1; /**< Disable (internal) auto-zqcs calibration When clear, LMC runs internal ZQ calibration every LMC*_CONFIG[REF_ZQCS_INT] CK cycles. */ uint64_t auto_dclkdis : 1; /**< When 1, LMC will automatically shut off its internal clock to conserve power when there is no traffic. Note that this has no effect on the DDR3 PHY and pads clocks. */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[2:0]=address[9:7] ^ address[14:12] else bank[2:0]=address[9:7] */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before forcing reads to interrupt. */ uint64_t nxm_write_en : 1; /**< NXM Write mode When clear, LMC discards writes to addresses that don't exist in the DRAM (as defined by LMC*_NXM configuration). When set, LMC completes writes to addresses that don't exist in the DRAM at an aliased address. */ uint64_t elev_prio_dis : 1; /**< Disable elevate priority logic. When set, writes are sent in regardless of priority information from L2C. */ uint64_t inorder_wr : 1; /**< Send writes in order(regardless of priority) */ uint64_t inorder_rd : 1; /**< Send reads in order (regardless of priority) */ uint64_t throttle_wr : 1; /**< When set, use at most one IFB for writes */ uint64_t throttle_rd : 1; /**< When set, use at most one IFB for reads */ uint64_t fprch2 : 2; /**< Front Porch Enable: When set, the turn-off time for the default DDR_DQ/DQS drivers is FPRCH2 CKs earlier. 00 = 0 CKs 01 = 1 CKs 10 = 2 CKs 11 = RESERVED */ uint64_t pocas : 1; /**< Enable the Posted CAS feature of DDR3. This bit must be set whenever LMC*_MODEREG_PARAMS0[AL]!=0, and clear otherwise. */ uint64_t ddr2t : 1; /**< Turn on the DDR 2T mode. 2 CK cycle window for CMD and address. This mode helps relieve setup time pressure on the Address and command bus which nominally have a very large fanout. Please refer to Micron's tech note tn_47_01 titled "DDR2-533 Memory Design Guide for Two Dimm Unbuffered Systems" for physical details. */ uint64_t bwcnt : 1; /**< Bus utilization counter Clear. Clears the LMC*_OPS_CNT, LMC*_IFB_CNT, and LMC*_DCLK_CNT registers. SW should first write this field to a one, then write this field to a zero to clear the CSR's. */ uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require address and control bits to be registered in the controller. */ #else uint64_t rdimm_ena : 1; uint64_t bwcnt : 1; uint64_t ddr2t : 1; uint64_t pocas : 1; uint64_t fprch2 : 2; uint64_t throttle_rd : 1; uint64_t throttle_wr : 1; uint64_t inorder_rd : 1; uint64_t inorder_wr : 1; uint64_t elev_prio_dis : 1; uint64_t nxm_write_en : 1; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t auto_dclkdis : 1; uint64_t int_zqcs_dis : 1; uint64_t ext_zqcs_dis : 1; uint64_t bprch : 2; uint64_t wodt_bprch : 1; uint64_t rodt_bprch : 1; uint64_t reserved_24_63 : 40; #endif } s; struct cvmx_lmcx_control_s cn63xx; struct cvmx_lmcx_control_s cn63xxp1; }; typedef union cvmx_lmcx_control cvmx_lmcx_control_t; /** * cvmx_lmc#_ctl * * LMC_CTL = LMC Control * This register is an assortment of various control fields needed by the memory controller */ union cvmx_lmcx_ctl { uint64_t u64; struct cvmx_lmcx_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< Should be cleared to zero */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t pll_div2 : 1; /**< PLL Div2. */ uint64_t pll_bypass : 1; /**< PLL Bypass. */ uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< Reads as zero */ uint64_t inorder_mrf : 1; /**< Always clear to zero */ uint64_t reserved_10_11 : 2; uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks */ uint64_t qs_dic : 2; /**< DDR2 Termination Resistor Setting A non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the 4/8 ODT pins (64/128b mode) based on what the masks (LMC_WODT_CTL) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. 0 = Normal 1 = Reduced DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t reserved_10_11 : 2; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t pll_bypass : 1; uint64_t pll_div2 : 1; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ctl_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< 1=SCF has pass1 latency, 0=SCF has 1 cycle lower latency when compared to pass1 */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t pll_div2 : 1; /**< PLL Div2. */ uint64_t pll_bypass : 1; /**< PLL Bypass. */ uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< Reads as zero */ uint64_t inorder_mrf : 1; /**< Always set to zero */ uint64_t dreset : 1; /**< Dclk domain reset. The reset signal that is used by the Dclk domain is (DRESET || ECLK_RESET). */ uint64_t mode32b : 1; /**< 32b data Path Mode Set to 1 if we use only 32 DQ pins 0 for 16b DQ mode. */ uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks */ uint64_t qs_dic : 2; /**< QS Drive Strength Control (DDR1): & DDR2 Termination Resistor Setting When in DDR2, a non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the 8 ODT pins based on what the masks (LMC_WODT_CTL1 & 2) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: For DDR-I/II Mode, DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. (see DDR-I data sheet EMRS description) 0 = Normal 1 = Reduced For DDR-II Mode, DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t mode32b : 1; uint64_t dreset : 1; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t pll_bypass : 1; uint64_t pll_div2 : 1; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } cn30xx; struct cvmx_lmcx_ctl_cn30xx cn31xx; struct cvmx_lmcx_ctl_cn38xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< 1=SCF has pass1 latency, 0=SCF has 1 cycle lower latency when compared to pass1 NOTE - This bit has NO effect in PASS1 */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t reserved_16_17 : 2; uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< When set, forces LMC_MWF (writes) into strict, in-order mode. When clear, writes may be serviced out of order (optimized to keep multiple banks active). This bit is ONLY to be set at power-on and should not be set for normal use. NOTE: For PASS1, set as follows: DDR-I -> 1 DDR-II -> 0 For Pass2, this bit is RA0, write ignore (this feature is permanently disabled) */ uint64_t inorder_mrf : 1; /**< When set, forces LMC_MRF (reads) into strict, in-order mode. When clear, reads may be serviced out of order (optimized to keep multiple banks active). This bit is ONLY to be set at power-on and should not be set for normal use. NOTE: For PASS1, set as follows: DDR-I -> 1 DDR-II -> 0 For Pass2, this bit should be written ZERO for DDR I & II */ uint64_t set_zero : 1; /**< Reserved. Always Set this Bit to Zero */ uint64_t mode128b : 1; /**< 128b data Path Mode Set to 1 if we use all 128 DQ pins 0 for 64b DQ mode. */ uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks */ uint64_t qs_dic : 2; /**< QS Drive Strength Control (DDR1): & DDR2 Termination Resistor Setting When in DDR2, a non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the 4/8 ODT pins (64/128b mode) based on what the masks (LMC_WODT_CTL) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: For DDR-I/II Mode, DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. (see DDR-I data sheet EMRS description) 0 = Normal 1 = Reduced For DDR-II Mode, DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t mode128b : 1; uint64_t set_zero : 1; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t reserved_16_17 : 2; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } cn38xx; struct cvmx_lmcx_ctl_cn38xx cn38xxp2; struct cvmx_lmcx_ctl_cn50xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< Should be cleared to zero */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t reserved_17_17 : 1; uint64_t pll_bypass : 1; /**< PLL Bypass. */ uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< Reads as zero */ uint64_t inorder_mrf : 1; /**< Always clear to zero */ uint64_t dreset : 1; /**< Dclk domain reset. The reset signal that is used by the Dclk domain is (DRESET || ECLK_RESET). */ uint64_t mode32b : 1; /**< 32b data Path Mode Set to 1 if we use 32 DQ pins 0 for 16b DQ mode. */ uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks */ uint64_t qs_dic : 2; /**< DDR2 Termination Resistor Setting When in DDR2, a non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the ODT pins based on what the masks (LMC_WODT_CTL) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. 0 = Normal 1 = Reduced DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t mode32b : 1; uint64_t dreset : 1; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t pll_bypass : 1; uint64_t reserved_17_17 : 1; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } cn50xx; struct cvmx_lmcx_ctl_cn52xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< Always clear to zero */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t reserved_16_17 : 2; uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< Reads as zero */ uint64_t inorder_mrf : 1; /**< Always set to zero */ uint64_t dreset : 1; /**< MBZ THIS IS OBSOLETE. Use LMC_DLL_CTL[DRESET] instead. */ uint64_t mode32b : 1; /**< 32b data Path Mode Set to 1 if we use only 32 DQ pins 0 for 64b DQ mode. */ uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. THIS IS OBSOLETE. Use READ_LEVEL_RANK instead. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks THIS IS OBSOLETE. Use READ_LEVEL_RANK instead. */ uint64_t qs_dic : 2; /**< DDR2 Termination Resistor Setting When in DDR2, a non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the 4/8 ODT pins (64/128b mode) based on what the masks (LMC_WODT_CTL0 & 1) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. 0 = Normal 1 = Reduced DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t mode32b : 1; uint64_t dreset : 1; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t reserved_16_17 : 2; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } cn52xx; struct cvmx_lmcx_ctl_cn52xx cn52xxp1; struct cvmx_lmcx_ctl_cn52xx cn56xx; struct cvmx_lmcx_ctl_cn52xx cn56xxp1; struct cvmx_lmcx_ctl_cn58xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 4; /**< DDR nctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pulldns. */ uint64_t ddr__pctl : 4; /**< DDR pctl from compensation circuit The encoded value on this will adjust the drive strength of the DDR DQ pullup. */ uint64_t slow_scf : 1; /**< Should be cleared to zero */ uint64_t xor_bank : 1; /**< If (XOR_BANK == 1), then bank[n:0]=address[n+7:7] ^ address[n+7+5:7+5] else bank[n:0]=address[n+7:7] where n=1 for a 4 bank part and n=2 for an 8 bank part */ uint64_t max_write_batch : 4; /**< Maximum number of consecutive writes to service before allowing reads to interrupt. */ uint64_t reserved_16_17 : 2; uint64_t rdimm_ena : 1; /**< Registered DIMM Enable - When set allows the use of JEDEC Registered DIMMs which require Write data to be registered in the controller. */ uint64_t r2r_slot : 1; /**< R2R Slot Enable: When set, all read-to-read trans will slot an additional 1 cycle data bus bubble to avoid DQ/DQS bus contention. This is only a CYA bit, in case the "built-in" DIMM and RANK crossing logic which should auto-detect and perfectly slot read-to-reads to the same DIMM/RANK. */ uint64_t inorder_mwf : 1; /**< Reads as zero */ uint64_t inorder_mrf : 1; /**< Always clear to zero */ uint64_t dreset : 1; /**< Dclk domain reset. The reset signal that is used by the Dclk domain is (DRESET || ECLK_RESET). */ uint64_t mode128b : 1; /**< 128b data Path Mode Set to 1 if we use all 128 DQ pins 0 for 64b DQ mode. */ uint64_t fprch2 : 1; /**< Front Porch Enable: When set, the turn-off time for the DDR_DQ/DQS drivers is 1 dclk earlier. This bit should typically be set. */ uint64_t bprch : 1; /**< Back Porch Enable: When set, the turn-on time for the DDR_DQ/DQS drivers is delayed an additional DCLK cycle. This should be set to one whenever both SILO_HC and SILO_QC are set. */ uint64_t sil_lat : 2; /**< SILO Latency: On reads, determines how many additional dclks to wait (on top of TCL+1+TSKW) before pulling data out of the pad silos. - 00: illegal - 01: 1 dclks - 10: 2 dclks - 11: illegal This should always be set to 1. */ uint64_t tskw : 2; /**< This component is a representation of total BOARD DELAY on DQ (used in the controller to determine the R->W spacing to avoid DQS/DQ bus conflicts). Enter the largest of the per byte Board delay - 00: 0 dclk - 01: 1 dclks - 10: 2 dclks - 11: 3 dclks */ uint64_t qs_dic : 2; /**< DDR2 Termination Resistor Setting A non Zero value in this register enables the On Die Termination (ODT) in DDR parts. These two bits are loaded into the RTT portion of the EMRS register bits A6 & A2. If DDR2's termination (for the memory's DQ/DQS/DM pads) is not desired, set it to 00. If it is, chose between 01 for 75 ohm and 10 for 150 ohm termination. 00 = ODT Disabled 01 = 75 ohm Termination 10 = 150 ohm Termination 11 = 50 ohm Termination Octeon, on writes, by default, drives the 4/8 ODT pins (64/128b mode) based on what the masks (LMC_WODT_CTL) are programmed to. LMC_DDR2_CTL->ODT_ENA enables Octeon to drive ODT pins for READS. LMC_RODT_CTL needs to be programmed based on the system's needs for ODT. */ uint64_t dic : 2; /**< Drive Strength Control: DIC[0] is loaded into the Extended Mode Register (EMRS) A1 bit during initialization. 0 = Normal 1 = Reduced DIC[1] is used to load into EMRS bit 10 - DQSN Enable/Disable field. By default, we program the DDR's to drive the DQSN also. Set it to 1 if DQSN should be Hi-Z. 0 - DQSN Enable 1 - DQSN Disable */ #else uint64_t dic : 2; uint64_t qs_dic : 2; uint64_t tskw : 2; uint64_t sil_lat : 2; uint64_t bprch : 1; uint64_t fprch2 : 1; uint64_t mode128b : 1; uint64_t dreset : 1; uint64_t inorder_mrf : 1; uint64_t inorder_mwf : 1; uint64_t r2r_slot : 1; uint64_t rdimm_ena : 1; uint64_t reserved_16_17 : 2; uint64_t max_write_batch : 4; uint64_t xor_bank : 1; uint64_t slow_scf : 1; uint64_t ddr__pctl : 4; uint64_t ddr__nctl : 4; uint64_t reserved_32_63 : 32; #endif } cn58xx; struct cvmx_lmcx_ctl_cn58xx cn58xxp1; }; typedef union cvmx_lmcx_ctl cvmx_lmcx_ctl_t; /** * cvmx_lmc#_ctl1 * * LMC_CTL1 = LMC Control1 * This register is an assortment of various control fields needed by the memory controller */ union cvmx_lmcx_ctl1 { uint64_t u64; struct cvmx_lmcx_ctl1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_21_63 : 43; uint64_t ecc_adr : 1; /**< Include memory reference address in the ECC calculation 0=disabled, 1=enabled */ uint64_t forcewrite : 4; /**< Force the oldest outstanding write to complete after having waited for 2^FORCEWRITE cycles. 0=disabled. */ uint64_t idlepower : 3; /**< Enter power-down mode after the memory controller has been idle for 2^(2+IDLEPOWER) cycles. 0=disabled. */ uint64_t sequence : 3; /**< Instruction sequence that is run after a 0->1 transition on LMC_MEM_CFG0[INIT_START]. 0=DDR2 power-up/init, 1=read-leveling 2=self-refresh entry, 3=self-refresh exit, 4=power-down entry, 5=power-down exit, 6=7=illegal */ uint64_t sil_mode : 1; /**< Read Silo mode. 0=envelope, 1=self-timed. */ uint64_t dcc_enable : 1; /**< Duty Cycle Corrector Enable. 0=disable, 1=enable If the memory part does not support DCC, then this bit must be set to 0. */ uint64_t reserved_2_7 : 6; uint64_t data_layout : 2; /**< Logical data layout per DQ byte lane: In 32b mode, this setting has no effect and the data layout DQ[35:0] is the following: [E[3:0], D[31:24], D[23:16], D[15:8], D[7:0]] In 16b mode, the DQ[35:0] layouts are the following: 0 - [0[3:0], 0[7:0], [0[7:2], E[1:0]], D[15:8], D[7:0]] 1 - [0[3:0], [0[7:2], E[1:0]], D[15:8], D[7:0], 0[7:0]] 2 - [[0[1:0], E[1:0]], D[15:8], D[7:0], 0[7:0], 0[7:0]] where E means ecc, D means data, and 0 means unused (ignored on reads and written as 0 on writes) */ #else uint64_t data_layout : 2; uint64_t reserved_2_7 : 6; uint64_t dcc_enable : 1; uint64_t sil_mode : 1; uint64_t sequence : 3; uint64_t idlepower : 3; uint64_t forcewrite : 4; uint64_t ecc_adr : 1; uint64_t reserved_21_63 : 43; #endif } s; struct cvmx_lmcx_ctl1_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_2_63 : 62; uint64_t data_layout : 2; /**< Logical data layout per DQ byte lane: In 32b mode, this setting has no effect and the data layout DQ[35:0] is the following: [E[3:0], D[31:24], D[23:16], D[15:8], D[7:0]] In 16b mode, the DQ[35:0] layouts are the following: 0 - [0[3:0], 0[7:0], [0[7:2], E[1:0]], D[15:8], D[7:0]] 1 - [0[3:0], [0[7:2], E[1:0]], D[15:8], D[7:0], 0[7:0]] 2 - [[0[1:0], E[1:0]], D[15:8], D[7:0], 0[7:0], 0[7:0]] where E means ecc, D means data, and 0 means unused (ignored on reads and written as 0 on writes) */ #else uint64_t data_layout : 2; uint64_t reserved_2_63 : 62; #endif } cn30xx; struct cvmx_lmcx_ctl1_cn50xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_10_63 : 54; uint64_t sil_mode : 1; /**< Read Silo mode. 0=envelope, 1=self-timed. */ uint64_t dcc_enable : 1; /**< Duty Cycle Corrector Enable. 0=disable, 1=enable If the memory part does not support DCC, then this bit must be set to 0. */ uint64_t reserved_2_7 : 6; uint64_t data_layout : 2; /**< Logical data layout per DQ byte lane: In 32b mode, this setting has no effect and the data layout DQ[35:0] is the following: [E[3:0], D[31:24], D[23:16], D[15:8], D[7:0]] In 16b mode, the DQ[35:0] layouts are the following: 0 - [0[3:0], 0[7:0], [0[7:2], E[1:0]], D[15:8], D[7:0]] 1 - [0[3:0], [0[7:2], E[1:0]], D[15:8], D[7:0], 0[7:0]] 2 - [[0[1:0], E[1:0]], D[15:8], D[7:0], 0[7:0], 0[7:0]] where E means ecc, D means data, and 0 means unused (ignored on reads and written as 0 on writes) */ #else uint64_t data_layout : 2; uint64_t reserved_2_7 : 6; uint64_t dcc_enable : 1; uint64_t sil_mode : 1; uint64_t reserved_10_63 : 54; #endif } cn50xx; struct cvmx_lmcx_ctl1_cn52xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_21_63 : 43; uint64_t ecc_adr : 1; /**< Include memory reference address in the ECC calculation 0=disabled, 1=enabled */ uint64_t forcewrite : 4; /**< Force the oldest outstanding write to complete after having waited for 2^FORCEWRITE cycles. 0=disabled. */ uint64_t idlepower : 3; /**< Enter power-down mode after the memory controller has been idle for 2^(2+IDLEPOWER) cycles. 0=disabled. */ uint64_t sequence : 3; /**< Instruction sequence that is run after a 0->1 transition on LMC_MEM_CFG0[INIT_START]. 0=DDR2 power-up/init, 1=read-leveling 2=self-refresh entry, 3=self-refresh exit, 4=power-down entry, 5=power-down exit, 6=7=illegal */ uint64_t sil_mode : 1; /**< Read Silo mode. 0=envelope, 1=self-timed. */ uint64_t dcc_enable : 1; /**< Duty Cycle Corrector Enable. 0=disable, 1=enable If the memory part does not support DCC, then this bit must be set to 0. */ uint64_t reserved_0_7 : 8; #else uint64_t reserved_0_7 : 8; uint64_t dcc_enable : 1; uint64_t sil_mode : 1; uint64_t sequence : 3; uint64_t idlepower : 3; uint64_t forcewrite : 4; uint64_t ecc_adr : 1; uint64_t reserved_21_63 : 43; #endif } cn52xx; struct cvmx_lmcx_ctl1_cn52xx cn52xxp1; struct cvmx_lmcx_ctl1_cn52xx cn56xx; struct cvmx_lmcx_ctl1_cn52xx cn56xxp1; struct cvmx_lmcx_ctl1_cn58xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_10_63 : 54; uint64_t sil_mode : 1; /**< Read Silo mode. 0=envelope, 1=self-timed. */ uint64_t dcc_enable : 1; /**< Duty Cycle Corrector Enable. 0=disable, 1=enable If the memory part does not support DCC, then this bit must be set to 0. */ uint64_t reserved_0_7 : 8; #else uint64_t reserved_0_7 : 8; uint64_t dcc_enable : 1; uint64_t sil_mode : 1; uint64_t reserved_10_63 : 54; #endif } cn58xx; struct cvmx_lmcx_ctl1_cn58xx cn58xxp1; }; typedef union cvmx_lmcx_ctl1 cvmx_lmcx_ctl1_t; /** * cvmx_lmc#_dclk_cnt * * LMC_DCLK_CNT = Performance Counters * */ union cvmx_lmcx_dclk_cnt { uint64_t u64; struct cvmx_lmcx_dclk_cnt_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t dclkcnt : 64; /**< Performance Counter 64-bit counter that increments every CK cycle */ #else uint64_t dclkcnt : 64; #endif } s; struct cvmx_lmcx_dclk_cnt_s cn63xx; struct cvmx_lmcx_dclk_cnt_s cn63xxp1; }; typedef union cvmx_lmcx_dclk_cnt cvmx_lmcx_dclk_cnt_t; /** * cvmx_lmc#_dclk_cnt_hi * * LMC_DCLK_CNT_HI = Performance Counters * */ union cvmx_lmcx_dclk_cnt_hi { uint64_t u64; struct cvmx_lmcx_dclk_cnt_hi_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t dclkcnt_hi : 32; /**< Performance Counter that counts dclks Upper 32-bits of a 64-bit counter. */ #else uint64_t dclkcnt_hi : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_dclk_cnt_hi_s cn30xx; struct cvmx_lmcx_dclk_cnt_hi_s cn31xx; struct cvmx_lmcx_dclk_cnt_hi_s cn38xx; struct cvmx_lmcx_dclk_cnt_hi_s cn38xxp2; struct cvmx_lmcx_dclk_cnt_hi_s cn50xx; struct cvmx_lmcx_dclk_cnt_hi_s cn52xx; struct cvmx_lmcx_dclk_cnt_hi_s cn52xxp1; struct cvmx_lmcx_dclk_cnt_hi_s cn56xx; struct cvmx_lmcx_dclk_cnt_hi_s cn56xxp1; struct cvmx_lmcx_dclk_cnt_hi_s cn58xx; struct cvmx_lmcx_dclk_cnt_hi_s cn58xxp1; }; typedef union cvmx_lmcx_dclk_cnt_hi cvmx_lmcx_dclk_cnt_hi_t; /** * cvmx_lmc#_dclk_cnt_lo * * LMC_DCLK_CNT_LO = Performance Counters * */ union cvmx_lmcx_dclk_cnt_lo { uint64_t u64; struct cvmx_lmcx_dclk_cnt_lo_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t dclkcnt_lo : 32; /**< Performance Counter that counts dclks Lower 32-bits of a 64-bit counter. */ #else uint64_t dclkcnt_lo : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_dclk_cnt_lo_s cn30xx; struct cvmx_lmcx_dclk_cnt_lo_s cn31xx; struct cvmx_lmcx_dclk_cnt_lo_s cn38xx; struct cvmx_lmcx_dclk_cnt_lo_s cn38xxp2; struct cvmx_lmcx_dclk_cnt_lo_s cn50xx; struct cvmx_lmcx_dclk_cnt_lo_s cn52xx; struct cvmx_lmcx_dclk_cnt_lo_s cn52xxp1; struct cvmx_lmcx_dclk_cnt_lo_s cn56xx; struct cvmx_lmcx_dclk_cnt_lo_s cn56xxp1; struct cvmx_lmcx_dclk_cnt_lo_s cn58xx; struct cvmx_lmcx_dclk_cnt_lo_s cn58xxp1; }; typedef union cvmx_lmcx_dclk_cnt_lo cvmx_lmcx_dclk_cnt_lo_t; /** * cvmx_lmc#_dclk_ctl * * LMC_DCLK_CTL = LMC DCLK generation control * * * Notes: * This CSR is only relevant for LMC1. LMC0_DCLK_CTL is not used. * */ union cvmx_lmcx_dclk_ctl { uint64_t u64; struct cvmx_lmcx_dclk_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_8_63 : 56; uint64_t off90_ena : 1; /**< 0=use global DCLK (i.e. the PLL) directly for LMC1 1=use the 90 degree DCLK DLL to offset LMC1 DCLK */ uint64_t dclk90_byp : 1; /**< 0=90 degree DCLK DLL uses sampled delay from LMC0 1=90 degree DCLK DLL uses DCLK90_VLU See DCLK90_VLU. */ uint64_t dclk90_ld : 1; /**< The 90 degree DCLK DLL samples the delay setting from LMC0's DLL when this field transitions 0->1 */ uint64_t dclk90_vlu : 5; /**< Manual open-loop delay setting. The LMC1 90 degree DCLK DLL uses DCLK90_VLU rather than the delay setting sampled from LMC0 when DCLK90_BYP=1. */ #else uint64_t dclk90_vlu : 5; uint64_t dclk90_ld : 1; uint64_t dclk90_byp : 1; uint64_t off90_ena : 1; uint64_t reserved_8_63 : 56; #endif } s; struct cvmx_lmcx_dclk_ctl_s cn56xx; struct cvmx_lmcx_dclk_ctl_s cn56xxp1; }; typedef union cvmx_lmcx_dclk_ctl cvmx_lmcx_dclk_ctl_t; /** * cvmx_lmc#_ddr2_ctl * * LMC_DDR2_CTL = LMC DDR2 & DLL Control Register * */ union cvmx_lmcx_ddr2_ctl { uint64_t u64; struct cvmx_lmcx_ddr2_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t bank8 : 1; /**< For 8 bank DDR2 parts 1 - DDR2 parts have 8 internal banks (BA is 3 bits wide). 0 - DDR2 parts have 4 internal banks (BA is 2 bits wide). */ uint64_t burst8 : 1; /**< 8-burst mode. 1 - DDR data transfer happens in burst of 8 0 - DDR data transfer happens in burst of 4 BURST8 should be set when DDR2T is set to minimize the command bandwidth loss. */ uint64_t addlat : 3; /**< Additional Latency for posted CAS When Posted CAS is on, this configures the additional latency. This should be set to 1 .. LMC_MEM_CFG1[TRCD]-2 (Note the implication that posted CAS should not be used when tRCD is two.) */ uint64_t pocas : 1; /**< Enable the Posted CAS feature of DDR2. */ uint64_t bwcnt : 1; /**< Bus utilization counter Clear. Clears the LMC_OPS_CNT_*, LMC_IFB_CNT_*, and LMC_DCLK_CNT_* registers. SW should first write this field to a one, then write this field to a zero to clear the CSR's. */ uint64_t twr : 3; /**< DDR Write Recovery time (tWR). Last Wr Brst to Pre delay This is not a direct encoding of the value. Its programmed as below per DDR2 spec. The decimal number on the right is RNDUP(tWR(ns) / tCYC(ns)) TYP=15ns - 000: RESERVED - 001: 2 - 010: 3 - 011: 4 - 100: 5 - 101: 6 - 110: 7 - 111: 8 */ uint64_t silo_hc : 1; /**< Delays the read sample window by a Half Cycle. */ uint64_t ddr_eof : 4; /**< Early Fill Counter Init. L2 needs to know a few cycle before a fill completes so it can get its Control pipe started (for better overall performance). This counter contains an init value which is a function of Eclk/Dclk ratio to account for the asynchronous boundary between L2 cache and the DRAM controller. This init value will determine when to safely let the L2 know that a fill termination is coming up. Set DDR_EOF according to the following rule: eclkFreq/dclkFreq = dclkPeriod/eclkPeriod = RATIO RATIO < 6/6 -> illegal 6/6 <= RATIO < 6/5 -> DDR_EOF=3 6/5 <= RATIO < 6/4 -> DDR_EOF=3 6/4 <= RATIO < 6/3 -> DDR_EOF=2 6/3 <= RATIO < 6/2 -> DDR_EOF=1 6/2 <= RATIO < 6/1 -> DDR_EOF=0 6/1 <= RATIO -> DDR_EOF=0 */ uint64_t tfaw : 5; /**< tFAW - Cycles = RNDUP[tFAW(ns)/tcyc(ns)] - 1 Four Access Window time. Relevant only in DDR2 AND in 8-bank parts. tFAW = 5'b0 in DDR2-4bank tFAW = RNDUP[tFAW(ns)/tcyc(ns)] - 1 in DDR2-8bank */ uint64_t crip_mode : 1; /**< Cripple Mode - When set, the LMC allows only 1 inflight transaction (.vs. 8 in normal mode). This bit is ONLY to be set at power-on and should not be set for normal use. */ uint64_t ddr2t : 1; /**< Turn on the DDR 2T mode. 2 cycle window for CMD and address. This mode helps relieve setup time pressure on the Address and command bus which nominally have a very large fanout. Please refer to Micron's tech note tn_47_01 titled "DDR2-533 Memory Design Guide for Two Dimm Unbuffered Systems" for physical details. BURST8 should be set when DDR2T is set to minimize add/cmd loss. */ uint64_t odt_ena : 1; /**< Enable Obsolete ODT on Reads Obsolete Read ODT wiggles DDR_ODT_* pins on reads. Should normally be cleared to zero. When this is on, the following fields must also be programmed: LMC_CTL->QS_DIC - programs the termination value LMC_RODT_CTL - programs the ODT I/O mask for Reads */ uint64_t qdll_ena : 1; /**< DDR Quad DLL Enable: A 0->1 transition on this bit after DCLK init sequence will reset the DDR 90 DLL. Should happen at startup before any activity in DDR. DRESET should be asserted before and for 10 usec following the 0->1 transition on QDLL_ENA. */ uint64_t dll90_vlu : 5; /**< Contains the open loop setting value for the DDR90 delay line. */ uint64_t dll90_byp : 1; /**< DDR DLL90 Bypass: When set, the DDR90 DLL is to be bypassed and the setting is defined by DLL90_VLU */ uint64_t rdqs : 1; /**< DDR2 RDQS mode. When set, configures memory subsystem to use unidirectional DQS pins. RDQS/DM - Rcv & DQS - Xmit */ uint64_t ddr2 : 1; /**< Should be set */ #else uint64_t ddr2 : 1; uint64_t rdqs : 1; uint64_t dll90_byp : 1; uint64_t dll90_vlu : 5; uint64_t qdll_ena : 1; uint64_t odt_ena : 1; uint64_t ddr2t : 1; uint64_t crip_mode : 1; uint64_t tfaw : 5; uint64_t ddr_eof : 4; uint64_t silo_hc : 1; uint64_t twr : 3; uint64_t bwcnt : 1; uint64_t pocas : 1; uint64_t addlat : 3; uint64_t burst8 : 1; uint64_t bank8 : 1; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ddr2_ctl_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t bank8 : 1; /**< For 8 bank DDR2 parts 1 - DDR2 parts have 8 internal banks (BA is 3 bits wide). 0 - DDR2 parts have 4 internal banks (BA is 2 bits wide). */ uint64_t burst8 : 1; /**< 8-burst mode. 1 - DDR data transfer happens in burst of 8 0 - DDR data transfer happens in burst of 4 BURST8 should be set when DDR2T is set to minimize add/cmd bandwidth loss. */ uint64_t addlat : 3; /**< Additional Latency for posted CAS When Posted CAS is on, this configures the additional latency. This should be set to 1 .. LMC_MEM_CFG1[TRCD]-2 (Note the implication that posted CAS should not be used when tRCD is two.) */ uint64_t pocas : 1; /**< Enable the Posted CAS feature of DDR2. */ uint64_t bwcnt : 1; /**< Bus utilization counter Clear. Clears the LMC_OPS_CNT_*, LMC_IFB_CNT_*, and LMC_DCLK_CNT_* registers. SW should first write this field to a one, then write this field to a zero to clear the CSR's. */ uint64_t twr : 3; /**< DDR Write Recovery time (tWR). Last Wr Brst to Pre delay This is not a direct encoding of the value. Its programmed as below per DDR2 spec. The decimal number on the right is RNDUP(tWR(ns) / tCYC(ns)) TYP=15ns - 000: RESERVED - 001: 2 - 010: 3 - 011: 4 - 100: 5 - 101: 6 - 110-111: RESERVED */ uint64_t silo_hc : 1; /**< Delays the read sample window by a Half Cycle. */ uint64_t ddr_eof : 4; /**< Early Fill Counter Init. L2 needs to know a few cycle before a fill completes so it can get its Control pipe started (for better overall performance). This counter contains an init value which is a function of Eclk/Dclk ratio to account for the asynchronous boundary between L2 cache and the DRAM controller. This init value will determine when to safely let the L2 know that a fill termination is coming up. DDR_EOF = RNDUP (DCLK period/Eclk Period). If the ratio is above 3, set DDR_EOF to 3. DCLK/ECLK period DDR_EOF Less than 1 1 Less than 2 2 More than 2 3 */ uint64_t tfaw : 5; /**< tFAW - Cycles = RNDUP[tFAW(ns)/tcyc(ns)] - 1 Four Access Window time. Relevant only in 8-bank parts. TFAW = 5'b0 for DDR2-4bank TFAW = RNDUP[tFAW(ns)/tcyc(ns)] - 1 in DDR2-8bank */ uint64_t crip_mode : 1; /**< Cripple Mode - When set, the LMC allows only 1 inflight transaction (.vs. 8 in normal mode). This bit is ONLY to be set at power-on and should not be set for normal use. */ uint64_t ddr2t : 1; /**< Turn on the DDR 2T mode. 2 cycle window for CMD and address. This mode helps relieve setup time pressure on the Address and command bus which nominally have a very large fanout. Please refer to Micron's tech note tn_47_01 titled "DDR2-533 Memory Design Guide for Two Dimm Unbuffered Systems" for physical details. BURST8 should be used when DDR2T is set to minimize add/cmd bandwidth loss. */ uint64_t odt_ena : 1; /**< Enable ODT for DDR2 on Reads When this is on, the following fields must also be programmed: LMC_CTL->QS_DIC - programs the termination value LMC_RODT_CTL - programs the ODT I/O mask for writes Program as 0 for DDR1 mode and ODT needs to be off on Octeon Reads */ uint64_t qdll_ena : 1; /**< DDR Quad DLL Enable: A 0->1 transition on this bit after erst deassertion will reset the DDR 90 DLL. Should happen at startup before any activity in DDR. */ uint64_t dll90_vlu : 5; /**< Contains the open loop setting value for the DDR90 delay line. */ uint64_t dll90_byp : 1; /**< DDR DLL90 Bypass: When set, the DDR90 DLL is to be bypassed and the setting is defined by DLL90_VLU */ uint64_t reserved_1_1 : 1; uint64_t ddr2 : 1; /**< DDR2 Enable: When set, configures memory subsystem for DDR-II SDRAMs. */ #else uint64_t ddr2 : 1; uint64_t reserved_1_1 : 1; uint64_t dll90_byp : 1; uint64_t dll90_vlu : 5; uint64_t qdll_ena : 1; uint64_t odt_ena : 1; uint64_t ddr2t : 1; uint64_t crip_mode : 1; uint64_t tfaw : 5; uint64_t ddr_eof : 4; uint64_t silo_hc : 1; uint64_t twr : 3; uint64_t bwcnt : 1; uint64_t pocas : 1; uint64_t addlat : 3; uint64_t burst8 : 1; uint64_t bank8 : 1; uint64_t reserved_32_63 : 32; #endif } cn30xx; struct cvmx_lmcx_ddr2_ctl_cn30xx cn31xx; struct cvmx_lmcx_ddr2_ctl_s cn38xx; struct cvmx_lmcx_ddr2_ctl_s cn38xxp2; struct cvmx_lmcx_ddr2_ctl_s cn50xx; struct cvmx_lmcx_ddr2_ctl_s cn52xx; struct cvmx_lmcx_ddr2_ctl_s cn52xxp1; struct cvmx_lmcx_ddr2_ctl_s cn56xx; struct cvmx_lmcx_ddr2_ctl_s cn56xxp1; struct cvmx_lmcx_ddr2_ctl_s cn58xx; struct cvmx_lmcx_ddr2_ctl_s cn58xxp1; }; typedef union cvmx_lmcx_ddr2_ctl cvmx_lmcx_ddr2_ctl_t; /** * cvmx_lmc#_ddr_pll_ctl * * LMC_DDR_PLL_CTL = LMC DDR PLL control * * * Notes: * DDR PLL Bringup sequence: * 1. Write CLKF, DDR_PS_EN, DFM_PS_EN, DIFFAMP, CPS, CPB. * If test mode is going to be activated, then also write jtg__ddr_pll_tm_en1, jtg__ddr_pll_tm_en2, jtg__ddr_pll_tm_en3, * jtg__ddr_pll_tm_en4, jtg__dfa_pll_tm_en1, jtg__dfa_pll_tm_en2, jtg__dfa_pll_tm_en3, jtg__dfa_pll_tm_en4, JTAG_TEST_MODE * 2. Wait 128 ref clock cycles (7680 rclk cycles) * 3. Write 1 to RESET_N * 4. Wait 1152 ref clocks (1152*16 rclk cycles) * 5. Write 0 to DDR_DIV_RESET and DFM_DIV_RESET * 6. Wait 10 ref clock cycles (160 rclk cycles) before bringing up the DDR interface * If test mode is going to be activated, wait an additional 8191 ref clocks (8191*16 rclk cycles) to allow PLL * clock alignment */ union cvmx_lmcx_ddr_pll_ctl { uint64_t u64; struct cvmx_lmcx_ddr_pll_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_27_63 : 37; uint64_t jtg_test_mode : 1; /**< JTAG Test Mode Clock alignment between DCLK & REFCLK as well as FCLK & REFCLK can only be performed after the ddr_pll_divider_reset is deasserted. SW need to wait atleast 10 reference clock cycles after deasserting pll_divider_reset before asserting LMC(0)_DDR_PLL_CTL[JTG_TEST_MODE]. During alignment (which can take upto 160 microseconds) DCLK and FCLK can exhibit some high frequency pulses. Therefore, all bring up activities in that clock domain need to be delayed (when the chip operates in jtg_test_mode) by about 160 microseconds to ensure that lock is achieved. */ uint64_t dfm_div_reset : 1; /**< DFM postscalar divider reset */ uint64_t dfm_ps_en : 3; /**< DFM postscalar divide ratio Determines the DFM CK speed. 0x0 : Divide LMC+DFM PLL output by 1 0x1 : Divide LMC+DFM PLL output by 2 0x2 : Divide LMC+DFM PLL output by 3 0x3 : Divide LMC+DFM PLL output by 4 0x4 : Divide LMC+DFM PLL output by 6 0x5 : Divide LMC+DFM PLL output by 8 0x6 : Divide LMC+DFM PLL output by 12 0x7 : Divide LMC+DFM PLL output by 12 DFM_PS_EN is not used when DFM_DIV_RESET = 1 */ uint64_t ddr_div_reset : 1; /**< DDR postscalar divider reset */ uint64_t ddr_ps_en : 3; /**< DDR postscalar divide ratio Determines the LMC CK speed. 0x0 : Divide LMC+DFM PLL output by 1 0x1 : Divide LMC+DFM PLL output by 2 0x2 : Divide LMC+DFM PLL output by 3 0x3 : Divide LMC+DFM PLL output by 4 0x4 : Divide LMC+DFM PLL output by 6 0x5 : Divide LMC+DFM PLL output by 8 0x6 : Divide LMC+DFM PLL output by 12 0x7 : Divide LMC+DFM PLL output by 12 DDR_PS_EN is not used when DDR_DIV_RESET = 1 */ uint64_t diffamp : 4; /**< PLL diffamp input transconductance */ uint64_t cps : 3; /**< PLL charge-pump current */ uint64_t cpb : 3; /**< PLL charge-pump current */ uint64_t reset_n : 1; /**< PLL reset */ uint64_t clkf : 7; /**< Multiply reference by CLKF 32 <= CLKF <= 64 LMC+DFM PLL frequency = 50 * CLKF min = 1.6 GHz, max = 3.2 GHz */ #else uint64_t clkf : 7; uint64_t reset_n : 1; uint64_t cpb : 3; uint64_t cps : 3; uint64_t diffamp : 4; uint64_t ddr_ps_en : 3; uint64_t ddr_div_reset : 1; uint64_t dfm_ps_en : 3; uint64_t dfm_div_reset : 1; uint64_t jtg_test_mode : 1; uint64_t reserved_27_63 : 37; #endif } s; struct cvmx_lmcx_ddr_pll_ctl_s cn63xx; struct cvmx_lmcx_ddr_pll_ctl_s cn63xxp1; }; typedef union cvmx_lmcx_ddr_pll_ctl cvmx_lmcx_ddr_pll_ctl_t; /** * cvmx_lmc#_delay_cfg * * LMC_DELAY_CFG = Open-loop delay line settings * * * Notes: * The DQ bits add OUTGOING delay only to dq, dqs_[p,n], cb, cbs_[p,n], dqm. Delay is approximately * 50-80ps per setting depending on process/voltage. There is no need to add incoming delay since by * default all strobe bits are delayed internally by 90 degrees (as was always the case in previous * passes and past chips. * * The CMD add delay to all command bits DDR_RAS, DDR_CAS, DDR_A<15:0>, DDR_BA<2:0>, DDR_n_CS<1:0>_L, * DDR_WE, DDR_CKE and DDR_ODT_<7:0>. Again, delay is 50-80ps per tap. * * The CLK bits add delay to all clock signals DDR_CK_<5:0>_P and DDR_CK_<5:0>_N. Again, delay is * 50-80ps per tap. * * The usage scenario is the following: There is too much delay on command signals and setup on command * is not met. The user can then delay the clock until setup is met. * * At the same time though, dq/dqs should be delayed because there is also a DDR spec tying dqs with * clock. If clock is too much delayed with respect to dqs, writes will start to fail. * * This scheme should eliminate the board need of adding routing delay to clock signals to make high * frequencies work. */ union cvmx_lmcx_delay_cfg { uint64_t u64; struct cvmx_lmcx_delay_cfg_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_15_63 : 49; uint64_t dq : 5; /**< Setting for DQ delay line */ uint64_t cmd : 5; /**< Setting for CMD delay line */ uint64_t clk : 5; /**< Setting for CLK delay line */ #else uint64_t clk : 5; uint64_t cmd : 5; uint64_t dq : 5; uint64_t reserved_15_63 : 49; #endif } s; struct cvmx_lmcx_delay_cfg_s cn30xx; struct cvmx_lmcx_delay_cfg_cn38xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_14_63 : 50; uint64_t dq : 4; /**< Setting for DQ delay line */ uint64_t reserved_9_9 : 1; uint64_t cmd : 4; /**< Setting for CMD delay line */ uint64_t reserved_4_4 : 1; uint64_t clk : 4; /**< Setting for CLK delay line */ #else uint64_t clk : 4; uint64_t reserved_4_4 : 1; uint64_t cmd : 4; uint64_t reserved_9_9 : 1; uint64_t dq : 4; uint64_t reserved_14_63 : 50; #endif } cn38xx; struct cvmx_lmcx_delay_cfg_cn38xx cn50xx; struct cvmx_lmcx_delay_cfg_cn38xx cn52xx; struct cvmx_lmcx_delay_cfg_cn38xx cn52xxp1; struct cvmx_lmcx_delay_cfg_cn38xx cn56xx; struct cvmx_lmcx_delay_cfg_cn38xx cn56xxp1; struct cvmx_lmcx_delay_cfg_cn38xx cn58xx; struct cvmx_lmcx_delay_cfg_cn38xx cn58xxp1; }; typedef union cvmx_lmcx_delay_cfg cvmx_lmcx_delay_cfg_t; /** * cvmx_lmc#_dimm#_params * * LMC_DIMMX_PARAMS = LMC DIMMX Params * This register contains values to be programmed into each control word in the corresponding (registered) DIMM. The control words allow * optimization of the device properties for different raw card designs. * * Notes: * LMC only uses this CSR when LMC*_CONTROL[RDIMM_ENA]=1. During a power-up/init sequence, LMC writes * these fields into the control words in the JEDEC standard SSTE32882 registering clock driver on an * RDIMM when corresponding LMC*_DIMM_CTL[DIMM*_WMASK] bits are set. */ union cvmx_lmcx_dimmx_params { uint64_t u64; struct cvmx_lmcx_dimmx_params_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t rc15 : 4; /**< RC15, Reserved */ uint64_t rc14 : 4; /**< RC14, Reserved */ uint64_t rc13 : 4; /**< RC13, Reserved */ uint64_t rc12 : 4; /**< RC12, Reserved */ uint64_t rc11 : 4; /**< RC11, Encoding for RDIMM Operating VDD */ uint64_t rc10 : 4; /**< RC10, Encoding for RDIMM Operating Speed */ uint64_t rc9 : 4; /**< RC9 , Power Savings Settings Control Word */ uint64_t rc8 : 4; /**< RC8 , Additional IBT Settings Control Word */ uint64_t rc7 : 4; /**< RC7 , Reserved */ uint64_t rc6 : 4; /**< RC6 , Reserved */ uint64_t rc5 : 4; /**< RC5 , CK Driver Characterstics Control Word */ uint64_t rc4 : 4; /**< RC4 , Control Signals Driver Characteristics Control Word */ uint64_t rc3 : 4; /**< RC3 , CA Signals Driver Characterstics Control Word */ uint64_t rc2 : 4; /**< RC2 , Timing Control Word */ uint64_t rc1 : 4; /**< RC1 , Clock Driver Enable Control Word */ uint64_t rc0 : 4; /**< RC0 , Global Features Control Word */ #else uint64_t rc0 : 4; uint64_t rc1 : 4; uint64_t rc2 : 4; uint64_t rc3 : 4; uint64_t rc4 : 4; uint64_t rc5 : 4; uint64_t rc6 : 4; uint64_t rc7 : 4; uint64_t rc8 : 4; uint64_t rc9 : 4; uint64_t rc10 : 4; uint64_t rc11 : 4; uint64_t rc12 : 4; uint64_t rc13 : 4; uint64_t rc14 : 4; uint64_t rc15 : 4; #endif } s; struct cvmx_lmcx_dimmx_params_s cn63xx; struct cvmx_lmcx_dimmx_params_s cn63xxp1; }; typedef union cvmx_lmcx_dimmx_params cvmx_lmcx_dimmx_params_t; /** * cvmx_lmc#_dimm_ctl * * LMC_DIMM_CTL = LMC DIMM Control * * * Notes: * This CSR is only used when LMC*_CONTROL[RDIMM_ENA]=1. During a power-up/init sequence, this CSR * controls LMC's writes to the control words in the JEDEC standard SSTE32882 registering clock driver * on an RDIMM. */ union cvmx_lmcx_dimm_ctl { uint64_t u64; struct cvmx_lmcx_dimm_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_46_63 : 18; uint64_t parity : 1; /**< Parity */ uint64_t tcws : 13; /**< LMC waits for this time period before and after a RDIMM Control Word Access during a power-up/init SEQUENCE. TCWS is in multiples of 8 CK cycles. Set TCWS (CSR field) = RNDUP[tcws(ns)/(8*tCYC(ns))], where tCWS is the desired time (ns), and tCYC(ns) is the DDR clock frequency (not data rate). TYP=0x4e0 (equivalent to 15us) when changing clock timing (RC2.DBA1, RC6.DA4, RC10.DA3, RC10.DA4, RC11.DA3, and RC11.DA4) TYP=0x8, otherwise 0x0 = Reserved */ uint64_t dimm1_wmask : 16; /**< DIMM1 Write Mask if (DIMM1_WMASK[n] = 1) Write DIMM1.RCn */ uint64_t dimm0_wmask : 16; /**< DIMM0 Write Mask if (DIMM0_WMASK[n] = 1) Write DIMM0.RCn */ #else uint64_t dimm0_wmask : 16; uint64_t dimm1_wmask : 16; uint64_t tcws : 13; uint64_t parity : 1; uint64_t reserved_46_63 : 18; #endif } s; struct cvmx_lmcx_dimm_ctl_s cn63xx; struct cvmx_lmcx_dimm_ctl_s cn63xxp1; }; typedef union cvmx_lmcx_dimm_ctl cvmx_lmcx_dimm_ctl_t; /** * cvmx_lmc#_dll_ctl * * LMC_DLL_CTL = LMC DLL control and DCLK reset * */ union cvmx_lmcx_dll_ctl { uint64_t u64; struct cvmx_lmcx_dll_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_8_63 : 56; uint64_t dreset : 1; /**< Dclk domain reset. The reset signal that is used by the Dclk domain is (DRESET || ECLK_RESET). */ uint64_t dll90_byp : 1; /**< DDR DLL90 Bypass: When set, the DDR90 DLL is to be bypassed and the setting is defined by DLL90_VLU */ uint64_t dll90_ena : 1; /**< DDR Quad DLL Enable: A 0->1 transition on this bit after DCLK init sequence resets the DDR 90 DLL. Should happen at startup before any activity in DDR. QDLL_ENA must not transition 1->0 outside of a DRESET sequence (i.e. it must remain 1 until the next DRESET). DRESET should be asserted before and for 10 usec following the 0->1 transition on QDLL_ENA. */ uint64_t dll90_vlu : 5; /**< Contains the open loop setting value for the DDR90 delay line. */ #else uint64_t dll90_vlu : 5; uint64_t dll90_ena : 1; uint64_t dll90_byp : 1; uint64_t dreset : 1; uint64_t reserved_8_63 : 56; #endif } s; struct cvmx_lmcx_dll_ctl_s cn52xx; struct cvmx_lmcx_dll_ctl_s cn52xxp1; struct cvmx_lmcx_dll_ctl_s cn56xx; struct cvmx_lmcx_dll_ctl_s cn56xxp1; }; typedef union cvmx_lmcx_dll_ctl cvmx_lmcx_dll_ctl_t; /** * cvmx_lmc#_dll_ctl2 * * LMC_DLL_CTL2 = LMC (Octeon) DLL control and DCLK reset * * * Notes: * DLL Bringup sequence: * 1. If not done already, set LMC*_DLL_CTL2 = 0, except when LMC*_DLL_CTL2[DRESET] = 1. * 2. Write 1 to LMC*_DLL_CTL2[DLL_BRINGUP] * 3. Wait for 10 CK cycles, then write 1 to LMC*_DLL_CTL2[QUAD_DLL_ENA]. It may not be feasible to count 10 CK cycles, but the * idea is to configure the delay line into DLL mode by asserting DLL_BRING_UP earlier than [QUAD_DLL_ENA], even if it is one * cycle early. LMC*_DLL_CTL2[QUAD_DLL_ENA] must not change after this point without restarting the LMC and/or DRESET initialization * sequence. * 4. Read L2D_BST0 and wait for the result. (L2D_BST0 is subject to change depending on how it called in o63. It is still ok to go * without step 4, since step 5 has enough time) * 5. Wait 10 us. * 6. Write 0 to LMC*_DLL_CTL2[DLL_BRINGUP]. LMC*_DLL_CTL2[DLL_BRINGUP] must not change after this point without restarting the LMC * and/or DRESET initialization sequence. * 7. Read L2D_BST0 and wait for the result. (same as step 4, but the idea here is the wait some time before going to step 8, even it * is one cycle is fine) * 8. Write 0 to LMC*_DLL_CTL2[DRESET]. LMC*_DLL_CTL2[DRESET] must not change after this point without restarting the LMC and/or * DRESET initialization sequence. */ union cvmx_lmcx_dll_ctl2 { uint64_t u64; struct cvmx_lmcx_dll_ctl2_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_15_63 : 49; uint64_t dll_bringup : 1; /**< DLL Bringup */ uint64_t dreset : 1; /**< Dclk domain reset. The reset signal that is used by the Dclk domain is (DRESET || ECLK_RESET). */ uint64_t quad_dll_ena : 1; /**< DLL Enable */ uint64_t byp_sel : 4; /**< Bypass select 0000 : no byte 0001 : byte 0 - ... 1001 : byte 8 1010 : all bytes 1011-1111 : Reserved */ uint64_t byp_setting : 8; /**< Bypass setting DDR3-1600: 00100010 DDR3-1333: 00110010 DDR3-1066: 01001011 DDR3-800 : 01110101 DDR3-667 : 10010110 DDR3-600 : 10101100 */ #else uint64_t byp_setting : 8; uint64_t byp_sel : 4; uint64_t quad_dll_ena : 1; uint64_t dreset : 1; uint64_t dll_bringup : 1; uint64_t reserved_15_63 : 49; #endif } s; struct cvmx_lmcx_dll_ctl2_s cn63xx; struct cvmx_lmcx_dll_ctl2_s cn63xxp1; }; typedef union cvmx_lmcx_dll_ctl2 cvmx_lmcx_dll_ctl2_t; /** * cvmx_lmc#_dll_ctl3 * * LMC_DLL_CTL3 = LMC DLL control and DCLK reset * */ union cvmx_lmcx_dll_ctl3 { uint64_t u64; struct cvmx_lmcx_dll_ctl3_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_29_63 : 35; uint64_t dll_fast : 1; /**< DLL lock 0 = DLL locked */ uint64_t dll90_setting : 8; /**< Encoded DLL settings. Works in conjuction with DLL90_BYTE_SEL */ uint64_t fine_tune_mode : 1; /**< DLL Fine Tune Mode 0 = disabled 1 = enable. When enabled, calibrate internal PHY DLL every LMC*_CONFIG[REF_ZQCS_INT] CK cycles. */ uint64_t dll_mode : 1; /**< DLL Mode */ uint64_t dll90_byte_sel : 4; /**< Observe DLL settings for selected byte 0001 : byte 0 - ... 1001 : byte 8 0000,1010-1111 : Reserved */ uint64_t offset_ena : 1; /**< Offset enable 0 = disable 1 = enable */ uint64_t load_offset : 1; /**< Load offset 0 : disable 1 : load (generates a 1 cycle pulse to the PHY) This register is oneshot and clears itself each time it is set */ uint64_t mode_sel : 2; /**< Mode select 00 : reset 01 : write 10 : read 11 : write & read */ uint64_t byte_sel : 4; /**< Byte select 0000 : no byte 0001 : byte 0 - ... 1001 : byte 8 1010 : all bytes 1011-1111 : Reserved */ uint64_t offset : 6; /**< Write/read offset setting [4:0] : offset [5] : 0 = increment, 1 = decrement Not a 2's complement value */ #else uint64_t offset : 6; uint64_t byte_sel : 4; uint64_t mode_sel : 2; uint64_t load_offset : 1; uint64_t offset_ena : 1; uint64_t dll90_byte_sel : 4; uint64_t dll_mode : 1; uint64_t fine_tune_mode : 1; uint64_t dll90_setting : 8; uint64_t dll_fast : 1; uint64_t reserved_29_63 : 35; #endif } s; struct cvmx_lmcx_dll_ctl3_s cn63xx; struct cvmx_lmcx_dll_ctl3_s cn63xxp1; }; typedef union cvmx_lmcx_dll_ctl3 cvmx_lmcx_dll_ctl3_t; /** * cvmx_lmc#_dual_memcfg * * LMC_DUAL_MEMCFG = LMC Dual Memory Configuration Register * * This register controls certain parameters of Dual Memory Configuration * * Notes: * This register enables the design to have two, separate memory configurations, selected dynamically * by the reference address. Note however, that both configurations share * LMC*_CONTROL[XOR_BANK], LMC*_CONFIG[PBANK_LSB], LMC*_CONFIG[RANK_ENA], and all timing parameters. * In this description, "config0" refers to the normal memory configuration that is defined by the * LMC*_CONFIG[ROW_LSB] parameters and "config1" refers to the dual (or second) * memory configuration that is defined by this register. * * Enable mask to chip select mapping is shown below: * CS_MASK[3] -> DIMM1_CS_<1> * CS_MASK[2] -> DIMM1_CS_<0> * * CS_MASK[1] -> DIMM0_CS_<1> * CS_MASK[0] -> DIMM0_CS_<0> * * DIMM n uses the pair of chip selects DIMMn_CS_<1:0>. * * Programming restrictions for CS_MASK: * when LMC*_CONFIG[RANK_ENA] == 0, CS_MASK[2n + 1] = CS_MASK[2n] */ union cvmx_lmcx_dual_memcfg { uint64_t u64; struct cvmx_lmcx_dual_memcfg_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_20_63 : 44; uint64_t bank8 : 1; /**< See LMC_DDR2_CTL[BANK8] */ uint64_t row_lsb : 3; /**< See LMC*_CONFIG[ROW_LSB] */ uint64_t reserved_8_15 : 8; uint64_t cs_mask : 8; /**< Chip select mask. This mask corresponds to the 8 chip selects for a memory configuration. Each reference address will assert one of the chip selects. If that chip select has its corresponding CS_MASK bit set, then the "config1" parameters are used, otherwise the "config0" parameters are used. See additional notes below. [7:4] */ #else uint64_t cs_mask : 8; uint64_t reserved_8_15 : 8; uint64_t row_lsb : 3; uint64_t bank8 : 1; uint64_t reserved_20_63 : 44; #endif } s; struct cvmx_lmcx_dual_memcfg_s cn50xx; struct cvmx_lmcx_dual_memcfg_s cn52xx; struct cvmx_lmcx_dual_memcfg_s cn52xxp1; struct cvmx_lmcx_dual_memcfg_s cn56xx; struct cvmx_lmcx_dual_memcfg_s cn56xxp1; struct cvmx_lmcx_dual_memcfg_s cn58xx; struct cvmx_lmcx_dual_memcfg_s cn58xxp1; struct cvmx_lmcx_dual_memcfg_cn63xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_19_63 : 45; uint64_t row_lsb : 3; /**< See LMC*_CONFIG[ROW_LSB] */ uint64_t reserved_8_15 : 8; uint64_t cs_mask : 8; /**< Chip select mask. This mask corresponds to the 8 chip selects for a memory configuration. Each reference address will assert one of the chip selects. If that chip select has its corresponding CS_MASK bit set, then the "config1" parameters are used, otherwise the "config0" parameters are used. See additional notes below. [7:4] */ #else uint64_t cs_mask : 8; uint64_t reserved_8_15 : 8; uint64_t row_lsb : 3; uint64_t reserved_19_63 : 45; #endif } cn63xx; struct cvmx_lmcx_dual_memcfg_cn63xx cn63xxp1; }; typedef union cvmx_lmcx_dual_memcfg cvmx_lmcx_dual_memcfg_t; /** * cvmx_lmc#_ecc_synd * * LMC_ECC_SYND = MRD ECC Syndromes * */ union cvmx_lmcx_ecc_synd { uint64_t u64; struct cvmx_lmcx_ecc_synd_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t mrdsyn3 : 8; /**< MRD ECC Syndrome Quad3 MRDSYN3 corresponds to DQ[63:0]_c1_p1 where _cC_pP denotes cycle C and phase P */ uint64_t mrdsyn2 : 8; /**< MRD ECC Syndrome Quad2 MRDSYN2 corresponds to DQ[63:0]_c1_p0 where _cC_pP denotes cycle C and phase P */ uint64_t mrdsyn1 : 8; /**< MRD ECC Syndrome Quad1 MRDSYN1 corresponds to DQ[63:0]_c0_p1 where _cC_pP denotes cycle C and phase P */ uint64_t mrdsyn0 : 8; /**< MRD ECC Syndrome Quad0 MRDSYN0 corresponds to DQ[63:0]_c0_p0 where _cC_pP denotes cycle C and phase P */ #else uint64_t mrdsyn0 : 8; uint64_t mrdsyn1 : 8; uint64_t mrdsyn2 : 8; uint64_t mrdsyn3 : 8; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ecc_synd_s cn30xx; struct cvmx_lmcx_ecc_synd_s cn31xx; struct cvmx_lmcx_ecc_synd_s cn38xx; struct cvmx_lmcx_ecc_synd_s cn38xxp2; struct cvmx_lmcx_ecc_synd_s cn50xx; struct cvmx_lmcx_ecc_synd_s cn52xx; struct cvmx_lmcx_ecc_synd_s cn52xxp1; struct cvmx_lmcx_ecc_synd_s cn56xx; struct cvmx_lmcx_ecc_synd_s cn56xxp1; struct cvmx_lmcx_ecc_synd_s cn58xx; struct cvmx_lmcx_ecc_synd_s cn58xxp1; struct cvmx_lmcx_ecc_synd_s cn63xx; struct cvmx_lmcx_ecc_synd_s cn63xxp1; }; typedef union cvmx_lmcx_ecc_synd cvmx_lmcx_ecc_synd_t; /** * cvmx_lmc#_fadr * * LMC_FADR = LMC Failing Address Register (SEC/DED/NXM) * * This register only captures the first transaction with ecc/nxm errors. A DED/NXM error can * over-write this register with its failing addresses if the first error was a SEC. If you write * LMC*_CONFIG->SEC_ERR/DED_ERR/NXM_ERR then it will clear the error bits and capture the * next failing address. * * If FDIMM is 2 that means the error is in the higher bits DIMM. */ union cvmx_lmcx_fadr { uint64_t u64; struct cvmx_lmcx_fadr_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_0_63 : 64; #else uint64_t reserved_0_63 : 64; #endif } s; struct cvmx_lmcx_fadr_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t fdimm : 2; /**< Failing DIMM# */ uint64_t fbunk : 1; /**< Failing Rank */ uint64_t fbank : 3; /**< Failing Bank[2:0] */ uint64_t frow : 14; /**< Failing Row Address[13:0] */ uint64_t fcol : 12; /**< Failing Column Start Address[11:0] Represents the Failing read's starting column address (and not the exact column address in which the SEC/DED was detected) */ #else uint64_t fcol : 12; uint64_t frow : 14; uint64_t fbank : 3; uint64_t fbunk : 1; uint64_t fdimm : 2; uint64_t reserved_32_63 : 32; #endif } cn30xx; struct cvmx_lmcx_fadr_cn30xx cn31xx; struct cvmx_lmcx_fadr_cn30xx cn38xx; struct cvmx_lmcx_fadr_cn30xx cn38xxp2; struct cvmx_lmcx_fadr_cn30xx cn50xx; struct cvmx_lmcx_fadr_cn30xx cn52xx; struct cvmx_lmcx_fadr_cn30xx cn52xxp1; struct cvmx_lmcx_fadr_cn30xx cn56xx; struct cvmx_lmcx_fadr_cn30xx cn56xxp1; struct cvmx_lmcx_fadr_cn30xx cn58xx; struct cvmx_lmcx_fadr_cn30xx cn58xxp1; struct cvmx_lmcx_fadr_cn63xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_36_63 : 28; uint64_t fdimm : 2; /**< Failing DIMM# */ uint64_t fbunk : 1; /**< Failing Rank */ uint64_t fbank : 3; /**< Failing Bank[2:0] */ uint64_t frow : 16; /**< Failing Row Address[15:0] */ uint64_t fcol : 14; /**< Failing Column Address[13:0] Technically, represents the address of the 128b data that had an ecc error, i.e., fcol[0] is always 0. Can be used in conjuction with LMC*_CONFIG[DED_ERR] to isolate the 64b chunk of data in error */ #else uint64_t fcol : 14; uint64_t frow : 16; uint64_t fbank : 3; uint64_t fbunk : 1; uint64_t fdimm : 2; uint64_t reserved_36_63 : 28; #endif } cn63xx; struct cvmx_lmcx_fadr_cn63xx cn63xxp1; }; typedef union cvmx_lmcx_fadr cvmx_lmcx_fadr_t; /** * cvmx_lmc#_ifb_cnt * * LMC_IFB_CNT = Performance Counters * */ union cvmx_lmcx_ifb_cnt { uint64_t u64; struct cvmx_lmcx_ifb_cnt_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t ifbcnt : 64; /**< Performance Counter 64-bit counter that increments every CK cycle there is something in the in-flight buffer. */ #else uint64_t ifbcnt : 64; #endif } s; struct cvmx_lmcx_ifb_cnt_s cn63xx; struct cvmx_lmcx_ifb_cnt_s cn63xxp1; }; typedef union cvmx_lmcx_ifb_cnt cvmx_lmcx_ifb_cnt_t; /** * cvmx_lmc#_ifb_cnt_hi * * LMC_IFB_CNT_HI = Performance Counters * */ union cvmx_lmcx_ifb_cnt_hi { uint64_t u64; struct cvmx_lmcx_ifb_cnt_hi_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ifbcnt_hi : 32; /**< Performance Counter to measure Bus Utilization Upper 32-bits of 64-bit counter that increments every cycle there is something in the in-flight buffer. */ #else uint64_t ifbcnt_hi : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ifb_cnt_hi_s cn30xx; struct cvmx_lmcx_ifb_cnt_hi_s cn31xx; struct cvmx_lmcx_ifb_cnt_hi_s cn38xx; struct cvmx_lmcx_ifb_cnt_hi_s cn38xxp2; struct cvmx_lmcx_ifb_cnt_hi_s cn50xx; struct cvmx_lmcx_ifb_cnt_hi_s cn52xx; struct cvmx_lmcx_ifb_cnt_hi_s cn52xxp1; struct cvmx_lmcx_ifb_cnt_hi_s cn56xx; struct cvmx_lmcx_ifb_cnt_hi_s cn56xxp1; struct cvmx_lmcx_ifb_cnt_hi_s cn58xx; struct cvmx_lmcx_ifb_cnt_hi_s cn58xxp1; }; typedef union cvmx_lmcx_ifb_cnt_hi cvmx_lmcx_ifb_cnt_hi_t; /** * cvmx_lmc#_ifb_cnt_lo * * LMC_IFB_CNT_LO = Performance Counters * */ union cvmx_lmcx_ifb_cnt_lo { uint64_t u64; struct cvmx_lmcx_ifb_cnt_lo_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ifbcnt_lo : 32; /**< Performance Counter Low 32-bits of 64-bit counter that increments every cycle there is something in the in-flight buffer. */ #else uint64_t ifbcnt_lo : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ifb_cnt_lo_s cn30xx; struct cvmx_lmcx_ifb_cnt_lo_s cn31xx; struct cvmx_lmcx_ifb_cnt_lo_s cn38xx; struct cvmx_lmcx_ifb_cnt_lo_s cn38xxp2; struct cvmx_lmcx_ifb_cnt_lo_s cn50xx; struct cvmx_lmcx_ifb_cnt_lo_s cn52xx; struct cvmx_lmcx_ifb_cnt_lo_s cn52xxp1; struct cvmx_lmcx_ifb_cnt_lo_s cn56xx; struct cvmx_lmcx_ifb_cnt_lo_s cn56xxp1; struct cvmx_lmcx_ifb_cnt_lo_s cn58xx; struct cvmx_lmcx_ifb_cnt_lo_s cn58xxp1; }; typedef union cvmx_lmcx_ifb_cnt_lo cvmx_lmcx_ifb_cnt_lo_t; /** * cvmx_lmc#_int * * LMC_INT = LMC Interrupt Register * */ union cvmx_lmcx_int { uint64_t u64; struct cvmx_lmcx_int_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_9_63 : 55; uint64_t ded_err : 4; /**< Double Error detected (DED) of Rd Data [0] corresponds to DQ[63:0]_c0_p0 [1] corresponds to DQ[63:0]_c0_p1 [2] corresponds to DQ[63:0]_c1_p0 [3] corresponds to DQ[63:0]_c1_p1 where _cC_pP denotes cycle C and phase P Write of 1 will clear the corresponding error bit */ uint64_t sec_err : 4; /**< Single Error (corrected) of Rd Data [0] corresponds to DQ[63:0]_c0_p0 [1] corresponds to DQ[63:0]_c0_p1 [2] corresponds to DQ[63:0]_c1_p0 [3] corresponds to DQ[63:0]_c1_p1 where _cC_pP denotes cycle C and phase P Write of 1 will clear the corresponding error bit */ uint64_t nxm_wr_err : 1; /**< Write to non-existent memory Write of 1 will clear the corresponding error bit */ #else uint64_t nxm_wr_err : 1; uint64_t sec_err : 4; uint64_t ded_err : 4; uint64_t reserved_9_63 : 55; #endif } s; struct cvmx_lmcx_int_s cn63xx; struct cvmx_lmcx_int_s cn63xxp1; }; typedef union cvmx_lmcx_int cvmx_lmcx_int_t; /** * cvmx_lmc#_int_en * * LMC_INT_EN = LMC Interrupt Enable Register * */ union cvmx_lmcx_int_en { uint64_t u64; struct cvmx_lmcx_int_en_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_3_63 : 61; uint64_t intr_ded_ena : 1; /**< ECC Double Error Detect(DED) Interrupt Enable bit When set, the memory controller raises a processor interrupt on detecting an uncorrectable Dbl Bit ECC error. */ uint64_t intr_sec_ena : 1; /**< ECC Single Error Correct(SEC) Interrupt Enable bit When set, the memory controller raises a processor interrupt on detecting a correctable Single Bit ECC error. */ uint64_t intr_nxm_wr_ena : 1; /**< Non Write Error Interrupt Enable bit When set, the memory controller raises a processor interrupt on detecting an non-existent memory write */ #else uint64_t intr_nxm_wr_ena : 1; uint64_t intr_sec_ena : 1; uint64_t intr_ded_ena : 1; uint64_t reserved_3_63 : 61; #endif } s; struct cvmx_lmcx_int_en_s cn63xx; struct cvmx_lmcx_int_en_s cn63xxp1; }; typedef union cvmx_lmcx_int_en cvmx_lmcx_int_en_t; /** * cvmx_lmc#_mem_cfg0 * * Specify the RSL base addresses for the block * * LMC_MEM_CFG0 = LMC Memory Configuration Register0 * * This register controls certain parameters of Memory Configuration */ union cvmx_lmcx_mem_cfg0 { uint64_t u64; struct cvmx_lmcx_mem_cfg0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t reset : 1; /**< Reset oneshot pulse for refresh counter, and LMC_OPS_CNT_*, LMC_IFB_CNT_*, and LMC_DCLK_CNT_* CSR's. SW should write this to a one, then re-write it to a zero to cause the reset. */ uint64_t silo_qc : 1; /**< Adds a Quarter Cycle granularity to generate dqs pulse generation for silo. Combination of Silo_HC and Silo_QC gives the ability to position the read enable with quarter cycle resolution. This is applied on all the bytes uniformly. */ uint64_t bunk_ena : 1; /**< Bunk Enable aka RANK ena (for use with dual-rank DIMMs) For dual-rank DIMMs, the bunk_ena bit will enable the drive of the CS_N[1:0] pins based on the (pbank_lsb-1) address bit. Write 0 for SINGLE ranked DIMM's. */ uint64_t ded_err : 4; /**< Double Error detected (DED) of Rd Data In 128b mode, ecc is calulated on 1 cycle worth of data [25] corresponds to DQ[63:0], Phase0 [26] corresponds to DQ[127:64], Phase0 [27] corresponds to DQ[63:0], Phase1 [28] corresponds to DQ[127:64], Phase1 In 64b mode, ecc is calculated on 2 cycle worth of data [25] corresponds to DQ[63:0], Phase0, cycle0 [26] corresponds to DQ[63:0], Phase0, cycle1 [27] corresponds to DQ[63:0], Phase1, cycle0 [28] corresponds to DQ[63:0], Phase1, cycle1 Write of 1 will clear the corresponding error bit */ uint64_t sec_err : 4; /**< Single Error (corrected) of Rd Data In 128b mode, ecc is calulated on 1 cycle worth of data [21] corresponds to DQ[63:0], Phase0 [22] corresponds to DQ[127:64], Phase0 [23] corresponds to DQ[63:0], Phase1 [24] corresponds to DQ[127:64], Phase1 In 64b mode, ecc is calculated on 2 cycle worth of data [21] corresponds to DQ[63:0], Phase0, cycle0 [22] corresponds to DQ[63:0], Phase0, cycle1 [23] corresponds to DQ[63:0], Phase1, cycle0 [24] corresponds to DQ[63:0], Phase1, cycle1 Write of 1 will clear the corresponding error bit */ uint64_t intr_ded_ena : 1; /**< ECC Double Error Detect(DED) Interrupt Enable bit When set, the memory controller raises a processor interrupt on detecting an uncorrectable Dbl Bit ECC error. */ uint64_t intr_sec_ena : 1; /**< ECC Single Error Correct(SEC) Interrupt Enable bit When set, the memory controller raises a processor interrupt on detecting a correctable Single Bit ECC error. */ uint64_t tcl : 4; /**< This register is not used */ uint64_t ref_int : 6; /**< Refresh interval represented in \#of 512 dclk increments. Program this to RND-DN(tREFI/clkPeriod/512) - 000000: RESERVED - 000001: 1 * 512 = 512 dclks - ... - 111111: 63 * 512 = 32256 dclks */ uint64_t pbank_lsb : 4; /**< Physical Bank address select Reverting to the explanation for ROW_LSB, PBank_LSB would be Row_LSB bit + \#rowbits + \#rankbits In the 512MB DIMM Example, assuming no rank bits: pbank_lsb=mem_addr[15+13] for 64 b mode =mem_addr[16+13] for 128b mode Hence the parameter 0000:pbank[1:0] = mem_adr[28:27] / rank = mem_adr[26] (if bunk_ena) 0001:pbank[1:0] = mem_adr[29:28] / rank = mem_adr[27] " 0010:pbank[1:0] = mem_adr[30:29] / rank = mem_adr[28] " 0011:pbank[1:0] = mem_adr[31:30] / rank = mem_adr[29] " 0100:pbank[1:0] = mem_adr[32:31] / rank = mem_adr[30] " 0101:pbank[1:0] = mem_adr[33:32] / rank = mem_adr[31] " 0110:pbank[1:0] =[1'b0,mem_adr[33]] / rank = mem_adr[32] " 0111:pbank[1:0] =[2'b0] / rank = mem_adr[33] " 1000-1111: RESERVED */ uint64_t row_lsb : 3; /**< Encoding used to determine which memory address bit position represents the low order DDR ROW address. The processor's memory address[33:7] needs to be translated to DRAM addresses (bnk,row,col,rank and dimm) and that is a function of the following: 1. \# Banks (4 or 8) - spec'd by BANK8 2. Datapath Width(64 or 128) - MODE128b 3. \# Ranks in a DIMM - spec'd by BUNK_ENA 4. \# DIMM's in the system 5. \# Column Bits of the memory part - spec'd indirectly by this register. 6. \# Row Bits of the memory part - spec'd indirectly by the register below (PBANK_LSB). Illustration: For Micron's MT18HTF6472A,512MB DDR2 Unbuffered DIMM which uses 256Mb parts (8M x 8 x 4), \# Banks = 4 -> 2 bits of BA \# Columns = 1K -> 10 bits of Col \# Rows = 8K -> 13 bits of Row Assuming that the total Data width is 128, this is how we arrive at row_lsb: Col Address starts from mem_addr[4] for 128b (16Bytes) dq width or from mem_addr[3] for 64b (8Bytes) dq width \# col + \# bank = 12. Hence row_lsb is mem_adr[15] for 64bmode or mem_adr[16] for 128b mode. Hence row_lsb parameter should be set to 001 (64b) or 010 (128b). - 000: row_lsb = mem_adr[14] - 001: row_lsb = mem_adr[15] - 010: row_lsb = mem_adr[16] - 011: row_lsb = mem_adr[17] - 100: row_lsb = mem_adr[18] - 101-111:row_lsb = RESERVED */ uint64_t ecc_ena : 1; /**< ECC Enable: When set will enable the 8b ECC check/correct logic. Should be 1 when used with DIMMs with ECC. 0, otherwise. When this mode is turned on, DQ[71:64] and DQ[143:137] on writes, will contain the ECC code generated for the lower 64 and upper 64 bits of data which will written in the memory and then later on reads, used to check for Single bit error (which will be auto- corrected) and Double Bit error (which will be reported). When not turned on, DQ[71:64] and DQ[143:137] are driven to 0. Please refer to SEC_ERR, DED_ERR, LMC_FADR, and LMC_ECC_SYND registers for diagnostics information when there is an error. */ uint64_t init_start : 1; /**< A 0->1 transition starts the DDR memory initialization sequence. */ #else uint64_t init_start : 1; uint64_t ecc_ena : 1; uint64_t row_lsb : 3; uint64_t pbank_lsb : 4; uint64_t ref_int : 6; uint64_t tcl : 4; uint64_t intr_sec_ena : 1; uint64_t intr_ded_ena : 1; uint64_t sec_err : 4; uint64_t ded_err : 4; uint64_t bunk_ena : 1; uint64_t silo_qc : 1; uint64_t reset : 1; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_mem_cfg0_s cn30xx; struct cvmx_lmcx_mem_cfg0_s cn31xx; struct cvmx_lmcx_mem_cfg0_s cn38xx; struct cvmx_lmcx_mem_cfg0_s cn38xxp2; struct cvmx_lmcx_mem_cfg0_s cn50xx; struct cvmx_lmcx_mem_cfg0_s cn52xx; struct cvmx_lmcx_mem_cfg0_s cn52xxp1; struct cvmx_lmcx_mem_cfg0_s cn56xx; struct cvmx_lmcx_mem_cfg0_s cn56xxp1; struct cvmx_lmcx_mem_cfg0_s cn58xx; struct cvmx_lmcx_mem_cfg0_s cn58xxp1; }; typedef union cvmx_lmcx_mem_cfg0 cvmx_lmcx_mem_cfg0_t; /** * cvmx_lmc#_mem_cfg1 * * LMC_MEM_CFG1 = LMC Memory Configuration Register1 * * This register controls the External Memory Configuration Timing Parameters. Please refer to the * appropriate DDR part spec from your memory vendor for the various values in this CSR. * The details of each of these timing parameters can be found in the JEDEC spec or the vendor * spec of the memory parts. */ union cvmx_lmcx_mem_cfg1 { uint64_t u64; struct cvmx_lmcx_mem_cfg1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t comp_bypass : 1; /**< Compensation bypass. */ uint64_t trrd : 3; /**< tRRD cycles: ACT-ACT timing parameter for different banks. (Represented in tCYC cycles == 1dclks) TYP=15ns (66MHz=1,167MHz=3,200MHz=3) For DDR2, TYP=7.5ns - 000: RESERVED - 001: 1 tCYC - 010: 2 tCYC - 011: 3 tCYC - 100: 4 tCYC - 101: 5 tCYC - 110: 6 tCYC - 111: 7 tCYC */ uint64_t caslat : 3; /**< CAS Latency Encoding which is loaded into each DDR SDRAM device (MRS[6:4]) upon power-up (INIT_START=1). (Represented in tCYC cycles == 1 dclks) 000 RESERVED 001 RESERVED 010 2.0 tCYC 011 3.0 tCYC 100 4.0 tCYC 101 5.0 tCYC 110 6.0 tCYC 111 RESERVED eg). The parameters TSKW, SILO_HC, and SILO_QC can account for 1/4 cycle granularity in board/etch delays. */ uint64_t tmrd : 3; /**< tMRD Cycles (Represented in dclk tCYC) For DDR2, its TYP 2*tCYC) - 000: RESERVED - 001: 1 - 010: 2 - 011: 3 - 100: 4 - 101-111: RESERVED */ uint64_t trfc : 5; /**< Indicates tRFC constraints. Set TRFC (CSR field) = RNDUP[tRFC(ns)/4*tcyc(ns)], where tRFC is from the DDR2 spec, and tcyc(ns) is the DDR clock frequency (not data rate). For example, with 2Gb, DDR2-667 parts, typ tRFC=195ns, so TRFC (CSR field) = 0x11. TRFC (binary): Corresponding tRFC Cycles ---------------------------------------- - 00000-00001: RESERVED - 00010: 0-8 - 00011: 9-12 - 00100: 13-16 - ... - 11110: 117-120 - 11111: 121-124 */ uint64_t trp : 4; /**< tRP Cycles = RNDUP[tRP(ns)/tcyc(ns)] (Represented in tCYC cycles == 1dclk) TYP=15ns (66MHz=1,167MHz=3,400MHz=6 for TYP) - 0000: RESERVED - 0001: 1 - ... - 1001: 9 - 1010-1111: RESERVED When using parts with 8 banks (LMC_DDR2_CTL->BANK8 is 1), load tRP cycles + 1 into this register. */ uint64_t twtr : 4; /**< tWTR Cycles = RNDUP[tWTR(ns)/tcyc(ns)] Last Wr Data to Rd Command time. (Represented in tCYC cycles == 1dclks) TYP=15ns (66MHz=1,167MHz=3,400MHz=6, for TYP) - 0000: RESERVED - 0001: 1 - ... - 0111: 7 - 1000-1111: RESERVED */ uint64_t trcd : 4; /**< tRCD Cycles = RNDUP[tRCD(ns)/tcyc(ns)] (Represented in tCYC cycles == 1dclk) TYP=15ns (66MHz=1,167MHz=3,400MHz=6 for TYP) - 0000: RESERVED - 0001: 2 (2 is the smallest value allowed) - 0002: 2 - ... - 1001: 9 - 1010-1111: RESERVED In 2T mode, make this register TRCD-1, not going below 2. */ uint64_t tras : 5; /**< tRAS Cycles = RNDUP[tRAS(ns)/tcyc(ns)] (Represented in tCYC cycles == 1 dclk) - 00000-0001: RESERVED - 00010: 2 - ... - 11111: 31 */ #else uint64_t tras : 5; uint64_t trcd : 4; uint64_t twtr : 4; uint64_t trp : 4; uint64_t trfc : 5; uint64_t tmrd : 3; uint64_t caslat : 3; uint64_t trrd : 3; uint64_t comp_bypass : 1; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_mem_cfg1_s cn30xx; struct cvmx_lmcx_mem_cfg1_s cn31xx; struct cvmx_lmcx_mem_cfg1_cn38xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_31_63 : 33; uint64_t trrd : 3; /**< tRRD cycles: ACT-ACT timing parameter for different banks. (Represented in tCYC cycles == 1dclks) TYP=15ns (66MHz=1,167MHz=3,200MHz=3) For DDR2, TYP=7.5ns - 000: RESERVED - 001: 1 tCYC - 010: 2 tCYC - 011: 3 tCYC - 100: 4 tCYC - 101: 5 tCYC - 110-111: RESERVED */ uint64_t caslat : 3; /**< CAS Latency Encoding which is loaded into each DDR SDRAM device (MRS[6:4]) upon power-up (INIT_START=1). (Represented in tCYC cycles == 1 dclks) 000 RESERVED 001 RESERVED 010 2.0 tCYC 011 3.0 tCYC 100 4.0 tCYC 101 5.0 tCYC 110 6.0 tCYC (DDR2) 2.5 tCYC (DDR1) 111 RESERVED eg). The parameters TSKW, SILO_HC, and SILO_QC can account for 1/4 cycle granularity in board/etch delays. */ uint64_t tmrd : 3; /**< tMRD Cycles (Represented in dclk tCYC) For DDR2, its TYP 2*tCYC) - 000: RESERVED - 001: 1 - 010: 2 - 011: 3 - 100: 4 - 101-111: RESERVED */ uint64_t trfc : 5; /**< Indicates tRFC constraints. Set TRFC (CSR field) = RNDUP[tRFC(ns)/4*tcyc(ns)], where tRFC is from the DDR2 spec, and tcyc(ns) is the DDR clock frequency (not data rate). For example, with 2Gb, DDR2-667 parts, typ tRFC=195ns, so TRFC (CSR field) = 0x11. TRFC (binary): Corresponding tRFC Cycles ---------------------------------------- - 00000-00001: RESERVED - 00010: 0-8 - 00011: 9-12 - 00100: 13-16 - ... - 11110: 117-120 - 11111: 121-124 */ uint64_t trp : 4; /**< tRP Cycles = RNDUP[tRP(ns)/tcyc(ns)] (Represented in tCYC cycles == 1dclk) TYP=15ns (66MHz=1,167MHz=3,400MHz=6 for TYP) - 0000: RESERVED - 0001: 1 - ... - 0111: 7 - 1000-1111: RESERVED When using parts with 8 banks (LMC_DDR2_CTL->BANK8 is 1), load tRP cycles + 1 into this register. */ uint64_t twtr : 4; /**< tWTR Cycles = RNDUP[tWTR(ns)/tcyc(ns)] Last Wr Data to Rd Command time. (Represented in tCYC cycles == 1dclks) TYP=15ns (66MHz=1,167MHz=3,400MHz=6, for TYP) - 0000: RESERVED - 0001: 1 - ... - 0111: 7 - 1000-1111: RESERVED */ uint64_t trcd : 4; /**< tRCD Cycles = RNDUP[tRCD(ns)/tcyc(ns)] (Represented in tCYC cycles == 1dclk) TYP=15ns (66MHz=1,167MHz=3,400MHz=6 for TYP) - 0000: RESERVED - 0001: 2 (2 is the smallest value allowed) - 0002: 2 - ... - 0111: 7 - 1110-1111: RESERVED In 2T mode, make this register TRCD-1, not going below 2. */ uint64_t tras : 5; /**< tRAS Cycles = RNDUP[tRAS(ns)/tcyc(ns)] (Represented in tCYC cycles == 1 dclk) For DDR-I mode: TYP=45ns (66MHz=3,167MHz=8,400MHz=18 - 00000-0001: RESERVED - 00010: 2 - ... - 10100: 20 - 10101-11111: RESERVED */ #else uint64_t tras : 5; uint64_t trcd : 4; uint64_t twtr : 4; uint64_t trp : 4; uint64_t trfc : 5; uint64_t tmrd : 3; uint64_t caslat : 3; uint64_t trrd : 3; uint64_t reserved_31_63 : 33; #endif } cn38xx; struct cvmx_lmcx_mem_cfg1_cn38xx cn38xxp2; struct cvmx_lmcx_mem_cfg1_s cn50xx; struct cvmx_lmcx_mem_cfg1_cn38xx cn52xx; struct cvmx_lmcx_mem_cfg1_cn38xx cn52xxp1; struct cvmx_lmcx_mem_cfg1_cn38xx cn56xx; struct cvmx_lmcx_mem_cfg1_cn38xx cn56xxp1; struct cvmx_lmcx_mem_cfg1_cn38xx cn58xx; struct cvmx_lmcx_mem_cfg1_cn38xx cn58xxp1; }; typedef union cvmx_lmcx_mem_cfg1 cvmx_lmcx_mem_cfg1_t; /** * cvmx_lmc#_modereg_params0 * * Notes: * These parameters are written into the DDR3 MR0, MR1, MR2 and MR3 registers. * */ union cvmx_lmcx_modereg_params0 { uint64_t u64; struct cvmx_lmcx_modereg_params0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_25_63 : 39; uint64_t ppd : 1; /**< DLL Control for precharge powerdown 0 = Slow exit (DLL off) 1 = Fast exit (DLL on) LMC writes this value to MR0[PPD] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. This value must equal the MR0[PPD] value in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t wrp : 3; /**< Write recovery for auto precharge Should be programmed to be equal to or greater than RNDUP[tWR(ns)/tCYC(ns)] 000 = Reserved 001 = 5 010 = 6 011 = 7 100 = 8 101 = 10 110 = 12 111 = Reserved LMC writes this value to MR0[WR] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. This value must equal the MR0[WR] value in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t dllr : 1; /**< DLL Reset LMC writes this value to MR0[DLL] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR0[DLL] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t tm : 1; /**< Test Mode LMC writes this value to MR0[TM] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR0[TM] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t rbt : 1; /**< Read Burst Type 1 = interleaved (fixed) LMC writes this value to MR0[RBT] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR0[RBT] value must be 1 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t cl : 4; /**< CAS Latency 0010 = 5 0100 = 6 0110 = 7 1000 = 8 1010 = 9 1100 = 10 1110 = 11 0000, ???1 = Reserved LMC writes this value to MR0[CAS Latency / CL] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. This value must equal the MR0[CAS Latency / CL] value in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t bl : 2; /**< Burst Length 0 = 8 (fixed) LMC writes this value to MR0[BL] in the selected DDR3 parts during power-up/init and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR0[BL] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t qoff : 1; /**< Qoff Enable 0 = enable 1 = disable LMC writes this value to MR1[Qoff] in the DDR3 parts in the selected ranks during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK,INIT_STATUS] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. The MR1[Qoff] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t tdqs : 1; /**< TDQS Enable 0 = disable LMC writes this value to MR1[TDQS] in the DDR3 parts in the selected ranks during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK,INIT_STATUS] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t wlev : 1; /**< Write Leveling Enable 0 = disable LMC writes MR1[Level]=0 in the DDR3 parts in the selected ranks during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. (Write-leveling can only be initiated via the write-leveling instruction sequence.) See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK,INIT_STATUS] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t al : 2; /**< Additive Latency 00 = 0 01 = CL-1 10 = CL-2 11 = Reserved LMC writes this value to MR1[AL] in the selected DDR3 parts during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. This value must equal the MR1[AL] value in all the DDR3 parts attached to all ranks during normal operation. See also LMC*_CONTROL[POCAS]. */ uint64_t dll : 1; /**< DLL Enable 0 = enable 1 = disable. LMC writes this value to MR1[DLL] in the selected DDR3 parts during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. This value must equal the MR1[DLL] value in all the DDR3 parts attached to all ranks during normal operation. In dll-off mode, CL/CWL must be programmed equal to 6/6, respectively, as per the DDR3 specifications. */ uint64_t mpr : 1; /**< MPR LMC writes this value to MR3[MPR] in the selected DDR3 parts during power-up/init, read-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. (LMC also writes MR3[MPR]=1 at the beginning of the read-leveling instruction sequence. Read-leveling should only be initiated via the read-leveling instruction sequence.) See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR3[MPR] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t mprloc : 2; /**< MPR Location LMC writes this value to MR3[MPRLoc] in the selected DDR3 parts during power-up/init, read-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh exit instruction sequences. (LMC also writes MR3[MPRLoc]=0 at the beginning of the read-leveling instruction sequence.) See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK]. The MR3[MPRLoc] value must be 0 in all the DDR3 parts attached to all ranks during normal operation. */ uint64_t cwl : 3; /**< CAS Write Latency - 000: 5 - 001: 6 - 010: 7 - 011: 8 1xx: Reserved LMC writes this value to MR2[CWL] in the selected DDR3 parts during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. This value must equal the MR2[CWL] value in all the DDR3 parts attached to all ranks during normal operation. */ #else uint64_t cwl : 3; uint64_t mprloc : 2; uint64_t mpr : 1; uint64_t dll : 1; uint64_t al : 2; uint64_t wlev : 1; uint64_t tdqs : 1; uint64_t qoff : 1; uint64_t bl : 2; uint64_t cl : 4; uint64_t rbt : 1; uint64_t tm : 1; uint64_t dllr : 1; uint64_t wrp : 3; uint64_t ppd : 1; uint64_t reserved_25_63 : 39; #endif } s; struct cvmx_lmcx_modereg_params0_s cn63xx; struct cvmx_lmcx_modereg_params0_s cn63xxp1; }; typedef union cvmx_lmcx_modereg_params0 cvmx_lmcx_modereg_params0_t; /** * cvmx_lmc#_modereg_params1 * * Notes: * These parameters are written into the DDR3 MR0, MR1, MR2 and MR3 registers. * */ union cvmx_lmcx_modereg_params1 { uint64_t u64; struct cvmx_lmcx_modereg_params1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_48_63 : 16; uint64_t rtt_nom_11 : 3; /**< RTT_NOM Rank 3 LMC writes this value to MR1[Rtt_Nom] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. Per JEDEC DDR3 specifications, if RTT_Nom is used during writes, only values MR1[Rtt_Nom] = 1 (RQZ/4), 2 (RQZ/2), or 3 (RQZ/6) are allowed. Otherwise, values MR1[Rtt_Nom] = 4 (RQZ/12) and 5 (RQZ/8) are also allowed. */ uint64_t dic_11 : 2; /**< Output Driver Impedance Control Rank 3 LMC writes this value to MR1[D.I.C.] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_wr_11 : 2; /**< RTT_WR Rank 3 LMC writes this value to MR2[Rtt_WR] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t srt_11 : 1; /**< Self-refresh temperature range Rank 3 LMC writes this value to MR2[SRT] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t asr_11 : 1; /**< Auto self-refresh Rank 3 LMC writes this value to MR2[ASR] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t pasr_11 : 3; /**< Partial array self-refresh Rank 3 LMC writes this value to MR2[PASR] in the rank 3 (i.e. DIMM1_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_nom_10 : 3; /**< RTT_NOM Rank 2 LMC writes this value to MR1[Rtt_Nom] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. Per JEDEC DDR3 specifications, if RTT_Nom is used during writes, only values MR1[Rtt_Nom] = 1 (RQZ/4), 2 (RQZ/2), or 3 (RQZ/6) are allowed. Otherwise, values MR1[Rtt_Nom] = 4 (RQZ/12) and 5 (RQZ/8) are also allowed. */ uint64_t dic_10 : 2; /**< Output Driver Impedance Control Rank 2 LMC writes this value to MR1[D.I.C.] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_wr_10 : 2; /**< RTT_WR Rank 2 LMC writes this value to MR2[Rtt_WR] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t srt_10 : 1; /**< Self-refresh temperature range Rank 2 LMC writes this value to MR2[SRT] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t asr_10 : 1; /**< Auto self-refresh Rank 2 LMC writes this value to MR2[ASR] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t pasr_10 : 3; /**< Partial array self-refresh Rank 2 LMC writes this value to MR2[PASR] in the rank 2 (i.e. DIMM1_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_nom_01 : 3; /**< RTT_NOM Rank 1 LMC writes this value to MR1[Rtt_Nom] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. Per JEDEC DDR3 specifications, if RTT_Nom is used during writes, only values MR1[Rtt_Nom] = 1 (RQZ/4), 2 (RQZ/2), or 3 (RQZ/6) are allowed. Otherwise, values MR1[Rtt_Nom] = 4 (RQZ/12) and 5 (RQZ/8) are also allowed. */ uint64_t dic_01 : 2; /**< Output Driver Impedance Control Rank 1 LMC writes this value to MR1[D.I.C.] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_wr_01 : 2; /**< RTT_WR Rank 1 LMC writes this value to MR2[Rtt_WR] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t srt_01 : 1; /**< Self-refresh temperature range Rank 1 LMC writes this value to MR2[SRT] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t asr_01 : 1; /**< Auto self-refresh Rank 1 LMC writes this value to MR2[ASR] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t pasr_01 : 3; /**< Partial array self-refresh Rank 1 LMC writes this value to MR2[PASR] in the rank 1 (i.e. DIMM0_CS1) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_nom_00 : 3; /**< RTT_NOM Rank 0 LMC writes this value to MR1[Rtt_Nom] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. Per JEDEC DDR3 specifications, if RTT_Nom is used during writes, only values MR1[Rtt_Nom] = 1 (RQZ/4), 2 (RQZ/2), or 3 (RQZ/6) are allowed. Otherwise, values MR1[Rtt_Nom] = 4 (RQZ/12) and 5 (RQZ/8) are also allowed. */ uint64_t dic_00 : 2; /**< Output Driver Impedance Control Rank 0 LMC writes this value to MR1[D.I.C.] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t rtt_wr_00 : 2; /**< RTT_WR Rank 0 LMC writes this value to MR2[Rtt_WR] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t srt_00 : 1; /**< Self-refresh temperature range Rank 0 LMC writes this value to MR2[SRT] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t asr_00 : 1; /**< Auto self-refresh Rank 0 LMC writes this value to MR2[ASR] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ uint64_t pasr_00 : 3; /**< Partial array self-refresh Rank 0 LMC writes this value to MR2[PASR] in the rank 0 (i.e. DIMM0_CS0) DDR3 parts when selected during power-up/init, write-leveling, and, if LMC*_CONFIG[SREF_WITH_DLL] is set, self-refresh entry and exit instruction sequences. See LMC*_CONFIG[SEQUENCE,INIT_START,RANKMASK] and LMC*_RESET_CTL[DDR3PWARM,DDR3PSOFT]. */ #else uint64_t pasr_00 : 3; uint64_t asr_00 : 1; uint64_t srt_00 : 1; uint64_t rtt_wr_00 : 2; uint64_t dic_00 : 2; uint64_t rtt_nom_00 : 3; uint64_t pasr_01 : 3; uint64_t asr_01 : 1; uint64_t srt_01 : 1; uint64_t rtt_wr_01 : 2; uint64_t dic_01 : 2; uint64_t rtt_nom_01 : 3; uint64_t pasr_10 : 3; uint64_t asr_10 : 1; uint64_t srt_10 : 1; uint64_t rtt_wr_10 : 2; uint64_t dic_10 : 2; uint64_t rtt_nom_10 : 3; uint64_t pasr_11 : 3; uint64_t asr_11 : 1; uint64_t srt_11 : 1; uint64_t rtt_wr_11 : 2; uint64_t dic_11 : 2; uint64_t rtt_nom_11 : 3; uint64_t reserved_48_63 : 16; #endif } s; struct cvmx_lmcx_modereg_params1_s cn63xx; struct cvmx_lmcx_modereg_params1_s cn63xxp1; }; typedef union cvmx_lmcx_modereg_params1 cvmx_lmcx_modereg_params1_t; /** * cvmx_lmc#_nxm * * LMC_NXM = LMC non-existent memory * * * Notes: * Decoding for mem_msb/rank * - 0000: mem_msb = mem_adr[25] * - 0001: mem_msb = mem_adr[26] * - 0010: mem_msb = mem_adr[27] * - 0011: mem_msb = mem_adr[28] * - 0100: mem_msb = mem_adr[29] * - 0101: mem_msb = mem_adr[30] * - 0110: mem_msb = mem_adr[31] * - 0111: mem_msb = mem_adr[32] * - 1000: mem_msb = mem_adr[33] * - 1001: mem_msb = mem_adr[34] * 1010-1111 = Reserved * For example, for a DIMM made of Samsung's k4b1g0846c-f7 1Gb (16M x 8 bit x 8 bank) * DDR3 parts, the column address width = 10, so with * 10b of col, 3b of bus, 3b of bank, row_lsb = 16. So, row = mem_adr[29:16] and * mem_msb = 4 * * Note also that addresses greater the max defined space (pbank_msb) are also treated * as NXM accesses */ union cvmx_lmcx_nxm { uint64_t u64; struct cvmx_lmcx_nxm_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_40_63 : 24; uint64_t mem_msb_d3_r1 : 4; /**< Max Row MSB for DIMM3, RANK1/DIMM3 in Single Ranked */ uint64_t mem_msb_d3_r0 : 4; /**< Max Row MSB for DIMM3, RANK0 */ uint64_t mem_msb_d2_r1 : 4; /**< Max Row MSB for DIMM2, RANK1/DIMM2 in Single Ranked */ uint64_t mem_msb_d2_r0 : 4; /**< Max Row MSB for DIMM2, RANK0 */ uint64_t mem_msb_d1_r1 : 4; /**< Max Row MSB for DIMM1, RANK1/DIMM1 in Single Ranked */ uint64_t mem_msb_d1_r0 : 4; /**< Max Row MSB for DIMM1, RANK0 */ uint64_t mem_msb_d0_r1 : 4; /**< Max Row MSB for DIMM0, RANK1/DIMM0 in Single Ranked */ uint64_t mem_msb_d0_r0 : 4; /**< Max Row MSB for DIMM0, RANK0 */ uint64_t cs_mask : 8; /**< Chip select mask. This mask corresponds to the 8 chip selects for a memory configuration. If LMC*_CONFIG[RANK_ENA]==0 then this mask must be set in pairs because each reference address will assert a pair of chip selects. If the chip select(s) have a corresponding CS_MASK bit set, then the reference is to non-existent memory (NXM). LMC will alias a NXM read reference to use the lowest, legal chip select(s) and return 0's. LMC normally discards NXM writes, but will also alias them when LMC*_CONTROL[NXM_WRITE_EN]=1. CS_MASK<7:4> MBZ in 63xx */ #else uint64_t cs_mask : 8; uint64_t mem_msb_d0_r0 : 4; uint64_t mem_msb_d0_r1 : 4; uint64_t mem_msb_d1_r0 : 4; uint64_t mem_msb_d1_r1 : 4; uint64_t mem_msb_d2_r0 : 4; uint64_t mem_msb_d2_r1 : 4; uint64_t mem_msb_d3_r0 : 4; uint64_t mem_msb_d3_r1 : 4; uint64_t reserved_40_63 : 24; #endif } s; struct cvmx_lmcx_nxm_cn52xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_8_63 : 56; uint64_t cs_mask : 8; /**< Chip select mask. This mask corresponds to the 8 chip selects for a memory configuration. If LMC_MEM_CFG0[BUNK_ENA]==0 then this mask must be set in pairs because each reference address will assert a pair of chip selects. If the chip select(s) have a corresponding CS_MASK bit set, then the reference is to non-existent memory. LMC will alias the reference to use the lowest, legal chip select(s) in that case. */ #else uint64_t cs_mask : 8; uint64_t reserved_8_63 : 56; #endif } cn52xx; struct cvmx_lmcx_nxm_cn52xx cn56xx; struct cvmx_lmcx_nxm_cn52xx cn58xx; struct cvmx_lmcx_nxm_s cn63xx; struct cvmx_lmcx_nxm_s cn63xxp1; }; typedef union cvmx_lmcx_nxm cvmx_lmcx_nxm_t; /** * cvmx_lmc#_ops_cnt * * LMC_OPS_CNT = Performance Counters * */ union cvmx_lmcx_ops_cnt { uint64_t u64; struct cvmx_lmcx_ops_cnt_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t opscnt : 64; /**< Performance Counter 64-bit counter that increments when the DDR3 data bus is being used DRAM bus utilization = LMC*_OPS_CNT/LMC*_DCLK_CNT */ #else uint64_t opscnt : 64; #endif } s; struct cvmx_lmcx_ops_cnt_s cn63xx; struct cvmx_lmcx_ops_cnt_s cn63xxp1; }; typedef union cvmx_lmcx_ops_cnt cvmx_lmcx_ops_cnt_t; /** * cvmx_lmc#_ops_cnt_hi * * LMC_OPS_CNT_HI = Performance Counters * */ union cvmx_lmcx_ops_cnt_hi { uint64_t u64; struct cvmx_lmcx_ops_cnt_hi_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t opscnt_hi : 32; /**< Performance Counter to measure Bus Utilization Upper 32-bits of 64-bit counter DRAM bus utilization = LMC_OPS_CNT_* /LMC_DCLK_CNT_* */ #else uint64_t opscnt_hi : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ops_cnt_hi_s cn30xx; struct cvmx_lmcx_ops_cnt_hi_s cn31xx; struct cvmx_lmcx_ops_cnt_hi_s cn38xx; struct cvmx_lmcx_ops_cnt_hi_s cn38xxp2; struct cvmx_lmcx_ops_cnt_hi_s cn50xx; struct cvmx_lmcx_ops_cnt_hi_s cn52xx; struct cvmx_lmcx_ops_cnt_hi_s cn52xxp1; struct cvmx_lmcx_ops_cnt_hi_s cn56xx; struct cvmx_lmcx_ops_cnt_hi_s cn56xxp1; struct cvmx_lmcx_ops_cnt_hi_s cn58xx; struct cvmx_lmcx_ops_cnt_hi_s cn58xxp1; }; typedef union cvmx_lmcx_ops_cnt_hi cvmx_lmcx_ops_cnt_hi_t; /** * cvmx_lmc#_ops_cnt_lo * * LMC_OPS_CNT_LO = Performance Counters * */ union cvmx_lmcx_ops_cnt_lo { uint64_t u64; struct cvmx_lmcx_ops_cnt_lo_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t opscnt_lo : 32; /**< Performance Counter Low 32-bits of 64-bit counter DRAM bus utilization = LMC_OPS_CNT_* /LMC_DCLK_CNT_* */ #else uint64_t opscnt_lo : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_ops_cnt_lo_s cn30xx; struct cvmx_lmcx_ops_cnt_lo_s cn31xx; struct cvmx_lmcx_ops_cnt_lo_s cn38xx; struct cvmx_lmcx_ops_cnt_lo_s cn38xxp2; struct cvmx_lmcx_ops_cnt_lo_s cn50xx; struct cvmx_lmcx_ops_cnt_lo_s cn52xx; struct cvmx_lmcx_ops_cnt_lo_s cn52xxp1; struct cvmx_lmcx_ops_cnt_lo_s cn56xx; struct cvmx_lmcx_ops_cnt_lo_s cn56xxp1; struct cvmx_lmcx_ops_cnt_lo_s cn58xx; struct cvmx_lmcx_ops_cnt_lo_s cn58xxp1; }; typedef union cvmx_lmcx_ops_cnt_lo cvmx_lmcx_ops_cnt_lo_t; /** * cvmx_lmc#_phy_ctl * * LMC_PHY_CTL = LMC PHY Control * */ union cvmx_lmcx_phy_ctl { uint64_t u64; struct cvmx_lmcx_phy_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_15_63 : 49; uint64_t rx_always_on : 1; /**< Disable dynamic DDR3 IO Rx power gating */ uint64_t lv_mode : 1; /**< Low Voltage Mode (1.35V) */ uint64_t ck_tune1 : 1; /**< Clock Tune */ uint64_t ck_dlyout1 : 4; /**< Clock delay out setting */ uint64_t ck_tune0 : 1; /**< Clock Tune */ uint64_t ck_dlyout0 : 4; /**< Clock delay out setting */ uint64_t loopback : 1; /**< Loopback enable */ uint64_t loopback_pos : 1; /**< Loopback pos mode */ uint64_t ts_stagger : 1; /**< TS Staggermode This mode configures output drivers with 2-stage drive strength to avoid undershoot issues on the bus when strong drivers are suddenly turned on. When this mode is asserted, Octeon will configure output drivers to be weak drivers (60 ohm output impedance) at the first CK cycle, and change drivers to the designated drive strengths specified in $LMC(0)_COMP_CTL2 [CMD_CTL/CK_CTL/DQX_CTL] starting at the following cycle */ #else uint64_t ts_stagger : 1; uint64_t loopback_pos : 1; uint64_t loopback : 1; uint64_t ck_dlyout0 : 4; uint64_t ck_tune0 : 1; uint64_t ck_dlyout1 : 4; uint64_t ck_tune1 : 1; uint64_t lv_mode : 1; uint64_t rx_always_on : 1; uint64_t reserved_15_63 : 49; #endif } s; struct cvmx_lmcx_phy_ctl_s cn63xx; struct cvmx_lmcx_phy_ctl_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_14_63 : 50; uint64_t lv_mode : 1; /**< Low Voltage Mode (1.35V) */ uint64_t ck_tune1 : 1; /**< Clock Tune */ uint64_t ck_dlyout1 : 4; /**< Clock delay out setting */ uint64_t ck_tune0 : 1; /**< Clock Tune */ uint64_t ck_dlyout0 : 4; /**< Clock delay out setting */ uint64_t loopback : 1; /**< Loopback enable */ uint64_t loopback_pos : 1; /**< Loopback pos mode */ uint64_t ts_stagger : 1; /**< TS Staggermode This mode configures output drivers with 2-stage drive strength to avoid undershoot issues on the bus when strong drivers are suddenly turned on. When this mode is asserted, Octeon will configure output drivers to be weak drivers (60 ohm output impedance) at the first CK cycle, and change drivers to the designated drive strengths specified in $LMC(0)_COMP_CTL2 [CMD_CTL/CK_CTL/DQX_CTL] starting at the following cycle */ #else uint64_t ts_stagger : 1; uint64_t loopback_pos : 1; uint64_t loopback : 1; uint64_t ck_dlyout0 : 4; uint64_t ck_tune0 : 1; uint64_t ck_dlyout1 : 4; uint64_t ck_tune1 : 1; uint64_t lv_mode : 1; uint64_t reserved_14_63 : 50; #endif } cn63xxp1; }; typedef union cvmx_lmcx_phy_ctl cvmx_lmcx_phy_ctl_t; /** * cvmx_lmc#_pll_bwctl * * LMC_PLL_BWCTL = DDR PLL Bandwidth Control Register * */ union cvmx_lmcx_pll_bwctl { uint64_t u64; struct cvmx_lmcx_pll_bwctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_5_63 : 59; uint64_t bwupd : 1; /**< Load this Bandwidth Register value into the PLL */ uint64_t bwctl : 4; /**< Bandwidth Control Register for DDR PLL */ #else uint64_t bwctl : 4; uint64_t bwupd : 1; uint64_t reserved_5_63 : 59; #endif } s; struct cvmx_lmcx_pll_bwctl_s cn30xx; struct cvmx_lmcx_pll_bwctl_s cn31xx; struct cvmx_lmcx_pll_bwctl_s cn38xx; struct cvmx_lmcx_pll_bwctl_s cn38xxp2; }; typedef union cvmx_lmcx_pll_bwctl cvmx_lmcx_pll_bwctl_t; /** * cvmx_lmc#_pll_ctl * * LMC_PLL_CTL = LMC pll control * * * Notes: * This CSR is only relevant for LMC0. LMC1_PLL_CTL is not used. * * Exactly one of EN2, EN4, EN6, EN8, EN12, EN16 must be set. * * The resultant DDR_CK frequency is the DDR2_REF_CLK * frequency multiplied by: * * (CLKF + 1) / ((CLKR + 1) * EN(2,4,6,8,12,16)) * * The PLL frequency, which is: * * (DDR2_REF_CLK freq) * ((CLKF + 1) / (CLKR + 1)) * * must reside between 1.2 and 2.5 GHz. A faster PLL frequency is desirable if there is a choice. */ union cvmx_lmcx_pll_ctl { uint64_t u64; struct cvmx_lmcx_pll_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_30_63 : 34; uint64_t bypass : 1; /**< PLL Bypass */ uint64_t fasten_n : 1; /**< Should be set, especially when CLKF > ~80 */ uint64_t div_reset : 1; /**< Analog pll divider reset De-assert at least 500*(CLKR+1) reference clock cycles following RESET_N de-assertion. */ uint64_t reset_n : 1; /**< Analog pll reset De-assert at least 5 usec after CLKF, CLKR, and EN* are set up. */ uint64_t clkf : 12; /**< Multiply reference by CLKF + 1 CLKF must be <= 128 */ uint64_t clkr : 6; /**< Divide reference by CLKR + 1 */ uint64_t reserved_6_7 : 2; uint64_t en16 : 1; /**< Divide output by 16 */ uint64_t en12 : 1; /**< Divide output by 12 */ uint64_t en8 : 1; /**< Divide output by 8 */ uint64_t en6 : 1; /**< Divide output by 6 */ uint64_t en4 : 1; /**< Divide output by 4 */ uint64_t en2 : 1; /**< Divide output by 2 */ #else uint64_t en2 : 1; uint64_t en4 : 1; uint64_t en6 : 1; uint64_t en8 : 1; uint64_t en12 : 1; uint64_t en16 : 1; uint64_t reserved_6_7 : 2; uint64_t clkr : 6; uint64_t clkf : 12; uint64_t reset_n : 1; uint64_t div_reset : 1; uint64_t fasten_n : 1; uint64_t bypass : 1; uint64_t reserved_30_63 : 34; #endif } s; struct cvmx_lmcx_pll_ctl_cn50xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_29_63 : 35; uint64_t fasten_n : 1; /**< Should be set, especially when CLKF > ~80 */ uint64_t div_reset : 1; /**< Analog pll divider reset De-assert at least 500*(CLKR+1) reference clock cycles following RESET_N de-assertion. */ uint64_t reset_n : 1; /**< Analog pll reset De-assert at least 5 usec after CLKF, CLKR, and EN* are set up. */ uint64_t clkf : 12; /**< Multiply reference by CLKF + 1 CLKF must be <= 256 */ uint64_t clkr : 6; /**< Divide reference by CLKR + 1 */ uint64_t reserved_6_7 : 2; uint64_t en16 : 1; /**< Divide output by 16 */ uint64_t en12 : 1; /**< Divide output by 12 */ uint64_t en8 : 1; /**< Divide output by 8 */ uint64_t en6 : 1; /**< Divide output by 6 */ uint64_t en4 : 1; /**< Divide output by 4 */ uint64_t en2 : 1; /**< Divide output by 2 */ #else uint64_t en2 : 1; uint64_t en4 : 1; uint64_t en6 : 1; uint64_t en8 : 1; uint64_t en12 : 1; uint64_t en16 : 1; uint64_t reserved_6_7 : 2; uint64_t clkr : 6; uint64_t clkf : 12; uint64_t reset_n : 1; uint64_t div_reset : 1; uint64_t fasten_n : 1; uint64_t reserved_29_63 : 35; #endif } cn50xx; struct cvmx_lmcx_pll_ctl_s cn52xx; struct cvmx_lmcx_pll_ctl_s cn52xxp1; struct cvmx_lmcx_pll_ctl_cn50xx cn56xx; struct cvmx_lmcx_pll_ctl_cn56xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_28_63 : 36; uint64_t div_reset : 1; /**< Analog pll divider reset De-assert at least 500*(CLKR+1) reference clock cycles following RESET_N de-assertion. */ uint64_t reset_n : 1; /**< Analog pll reset De-assert at least 5 usec after CLKF, CLKR, and EN* are set up. */ uint64_t clkf : 12; /**< Multiply reference by CLKF + 1 CLKF must be <= 128 */ uint64_t clkr : 6; /**< Divide reference by CLKR + 1 */ uint64_t reserved_6_7 : 2; uint64_t en16 : 1; /**< Divide output by 16 */ uint64_t en12 : 1; /**< Divide output by 12 */ uint64_t en8 : 1; /**< Divide output by 8 */ uint64_t en6 : 1; /**< Divide output by 6 */ uint64_t en4 : 1; /**< Divide output by 4 */ uint64_t en2 : 1; /**< Divide output by 2 */ #else uint64_t en2 : 1; uint64_t en4 : 1; uint64_t en6 : 1; uint64_t en8 : 1; uint64_t en12 : 1; uint64_t en16 : 1; uint64_t reserved_6_7 : 2; uint64_t clkr : 6; uint64_t clkf : 12; uint64_t reset_n : 1; uint64_t div_reset : 1; uint64_t reserved_28_63 : 36; #endif } cn56xxp1; struct cvmx_lmcx_pll_ctl_cn56xxp1 cn58xx; struct cvmx_lmcx_pll_ctl_cn56xxp1 cn58xxp1; }; typedef union cvmx_lmcx_pll_ctl cvmx_lmcx_pll_ctl_t; /** * cvmx_lmc#_pll_status * * LMC_PLL_STATUS = LMC pll status * */ union cvmx_lmcx_pll_status { uint64_t u64; struct cvmx_lmcx_pll_status_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ddr__nctl : 5; /**< DDR nctl from compensation circuit */ uint64_t ddr__pctl : 5; /**< DDR pctl from compensation circuit */ uint64_t reserved_2_21 : 20; uint64_t rfslip : 1; /**< Reference clock slip */ uint64_t fbslip : 1; /**< Feedback clock slip */ #else uint64_t fbslip : 1; uint64_t rfslip : 1; uint64_t reserved_2_21 : 20; uint64_t ddr__pctl : 5; uint64_t ddr__nctl : 5; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_pll_status_s cn50xx; struct cvmx_lmcx_pll_status_s cn52xx; struct cvmx_lmcx_pll_status_s cn52xxp1; struct cvmx_lmcx_pll_status_s cn56xx; struct cvmx_lmcx_pll_status_s cn56xxp1; struct cvmx_lmcx_pll_status_s cn58xx; struct cvmx_lmcx_pll_status_cn58xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_2_63 : 62; uint64_t rfslip : 1; /**< Reference clock slip */ uint64_t fbslip : 1; /**< Feedback clock slip */ #else uint64_t fbslip : 1; uint64_t rfslip : 1; uint64_t reserved_2_63 : 62; #endif } cn58xxp1; }; typedef union cvmx_lmcx_pll_status cvmx_lmcx_pll_status_t; /** * cvmx_lmc#_read_level_ctl * * Notes: * The HW writes and reads the cache block selected by ROW, COL, BNK and the rank as part of a read-leveling sequence for a rank. * A cache block write is 16 72-bit words. PATTERN selects the write value. For the first 8 * words, the write value is the bit PATTERN<i> duplicated into a 72-bit vector. The write value of * the last 8 words is the inverse of the write value of the first 8 words. * See LMC*_READ_LEVEL_RANK*. */ union cvmx_lmcx_read_level_ctl { uint64_t u64; struct cvmx_lmcx_read_level_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_44_63 : 20; uint64_t rankmask : 4; /**< Selects ranks to be leveled to read-level rank i, set RANKMASK<i> */ uint64_t pattern : 8; /**< All DQ driven to PATTERN[burst], 0 <= burst <= 7 All DQ driven to ~PATTERN[burst-8], 8 <= burst <= 15 */ uint64_t row : 16; /**< Row address used to write/read data pattern */ uint64_t col : 12; /**< Column address used to write/read data pattern */ uint64_t reserved_3_3 : 1; uint64_t bnk : 3; /**< Bank address used to write/read data pattern */ #else uint64_t bnk : 3; uint64_t reserved_3_3 : 1; uint64_t col : 12; uint64_t row : 16; uint64_t pattern : 8; uint64_t rankmask : 4; uint64_t reserved_44_63 : 20; #endif } s; struct cvmx_lmcx_read_level_ctl_s cn52xx; struct cvmx_lmcx_read_level_ctl_s cn52xxp1; struct cvmx_lmcx_read_level_ctl_s cn56xx; struct cvmx_lmcx_read_level_ctl_s cn56xxp1; }; typedef union cvmx_lmcx_read_level_ctl cvmx_lmcx_read_level_ctl_t; /** * cvmx_lmc#_read_level_dbg * * Notes: * A given read of LMC*_READ_LEVEL_DBG returns the read-leveling pass/fail results for all possible * delay settings (i.e. the BITMASK) for only one byte in the last rank that the HW read-leveled. * LMC*_READ_LEVEL_DBG[BYTE] selects the particular byte. * To get these pass/fail results for another different rank, you must run the hardware read-leveling * again. For example, it is possible to get the BITMASK results for every byte of every rank * if you run read-leveling separately for each rank, probing LMC*_READ_LEVEL_DBG between each * read-leveling. */ union cvmx_lmcx_read_level_dbg { uint64_t u64; struct cvmx_lmcx_read_level_dbg_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t bitmask : 16; /**< Bitmask generated during deskew settings sweep BITMASK[n]=0 means deskew setting n failed BITMASK[n]=1 means deskew setting n passed for 0 <= n <= 15 */ uint64_t reserved_4_15 : 12; uint64_t byte : 4; /**< 0 <= BYTE <= 8 */ #else uint64_t byte : 4; uint64_t reserved_4_15 : 12; uint64_t bitmask : 16; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_read_level_dbg_s cn52xx; struct cvmx_lmcx_read_level_dbg_s cn52xxp1; struct cvmx_lmcx_read_level_dbg_s cn56xx; struct cvmx_lmcx_read_level_dbg_s cn56xxp1; }; typedef union cvmx_lmcx_read_level_dbg cvmx_lmcx_read_level_dbg_t; /** * cvmx_lmc#_read_level_rank# * * Notes: * This is four CSRs per LMC, one per each rank. * Each CSR is written by HW during a read-leveling sequence for the rank. (HW sets STATUS==3 after HW read-leveling completes for the rank.) * Each CSR may also be written by SW, but not while a read-leveling sequence is in progress. (HW sets STATUS==1 after a CSR write.) * Deskew setting is measured in units of 1/4 DCLK, so the above BYTE* values can range over 4 DCLKs. * SW initiates a HW read-leveling sequence by programming LMC*_READ_LEVEL_CTL and writing INIT_START=1 with SEQUENCE=1. * See LMC*_READ_LEVEL_CTL. */ union cvmx_lmcx_read_level_rankx { uint64_t u64; struct cvmx_lmcx_read_level_rankx_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_38_63 : 26; uint64_t status : 2; /**< Indicates status of the read-levelling and where the BYTE* programmings in <35:0> came from: 0 = BYTE* values are their reset value 1 = BYTE* values were set via a CSR write to this register 2 = read-leveling sequence currently in progress (BYTE* values are unpredictable) 3 = BYTE* values came from a complete read-leveling sequence */ uint64_t byte8 : 4; /**< Deskew setting */ uint64_t byte7 : 4; /**< Deskew setting */ uint64_t byte6 : 4; /**< Deskew setting */ uint64_t byte5 : 4; /**< Deskew setting */ uint64_t byte4 : 4; /**< Deskew setting */ uint64_t byte3 : 4; /**< Deskew setting */ uint64_t byte2 : 4; /**< Deskew setting */ uint64_t byte1 : 4; /**< Deskew setting */ uint64_t byte0 : 4; /**< Deskew setting */ #else uint64_t byte0 : 4; uint64_t byte1 : 4; uint64_t byte2 : 4; uint64_t byte3 : 4; uint64_t byte4 : 4; uint64_t byte5 : 4; uint64_t byte6 : 4; uint64_t byte7 : 4; uint64_t byte8 : 4; uint64_t status : 2; uint64_t reserved_38_63 : 26; #endif } s; struct cvmx_lmcx_read_level_rankx_s cn52xx; struct cvmx_lmcx_read_level_rankx_s cn52xxp1; struct cvmx_lmcx_read_level_rankx_s cn56xx; struct cvmx_lmcx_read_level_rankx_s cn56xxp1; }; typedef union cvmx_lmcx_read_level_rankx cvmx_lmcx_read_level_rankx_t; /** * cvmx_lmc#_reset_ctl * * Specify the RSL base addresses for the block * * * Notes: * DDR3RST - DDR3 DRAM parts have a new RESET# * pin that wasn't present in DDR2 parts. The * DDR3RST CSR field controls the assertion of * the new 63xx pin that attaches to RESET#. * When DDR3RST is set, 63xx asserts RESET#. * When DDR3RST is clear, 63xx de-asserts * RESET#. * * DDR3RST is set on a cold reset. Warm and * soft chip resets do not affect the DDR3RST * value. Outside of cold reset, only software * CSR writes change the DDR3RST value. * * DDR3PWARM - Enables preserve mode during a warm * reset. When set, the DDR3 controller hardware * automatically puts the attached DDR3 DRAM parts * into self refresh (see LMC*CONFIG[SEQUENCE] below) at the beginning of a warm * reset sequence, provided that the DDR3 controller * is up. When clear, the DDR3 controller hardware * does not put the attached DDR3 DRAM parts into * self-refresh during a warm reset sequence. * * DDR3PWARM is cleared on a cold reset. Warm and * soft chip resets do not affect the DDR3PWARM * value. Outside of cold reset, only software * CSR writes change the DDR3PWARM value. * * Note that if a warm reset follows a soft reset, * DDR3PWARM has no effect, as the DDR3 controller * is no longer up after any cold/warm/soft * reset sequence. * * DDR3PSOFT - Enables preserve mode during a soft * reset. When set, the DDR3 controller hardware * automatically puts the attached DDR3 DRAM parts * into self refresh (see LMC*CONFIG[SEQUENCE] below) at the beginning of a soft * reset sequence, provided that the DDR3 controller * is up. When clear, the DDR3 controller hardware * does not put the attached DDR3 DRAM parts into * self-refresh during a soft reset sequence. * * DDR3PSOFT is cleared on a cold reset. Warm and * soft chip resets do not affect the DDR3PSOFT * value. Outside of cold reset, only software * CSR writes change the DDR3PSOFT value. * * DDR3PSV - May be useful for system software to * determine when the DDR3 contents have been * preserved. * * Cleared by hardware during a cold reset. Never * cleared by hardware during a warm/soft reset. * Set by hardware during a warm/soft reset if * the hardware automatically put the DDR3 DRAM * into self-refresh during the reset sequence. * * Can also be written by software (to any value). */ union cvmx_lmcx_reset_ctl { uint64_t u64; struct cvmx_lmcx_reset_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_4_63 : 60; uint64_t ddr3psv : 1; /**< Memory Reset 1 = DDR contents preserved */ uint64_t ddr3psoft : 1; /**< Memory Reset 1 = Enable Preserve mode during soft reset */ uint64_t ddr3pwarm : 1; /**< Memory Reset 1 = Enable Preserve mode during warm reset */ uint64_t ddr3rst : 1; /**< Memory Reset 0 = Reset asserted 1 = Reset de-asserted */ #else uint64_t ddr3rst : 1; uint64_t ddr3pwarm : 1; uint64_t ddr3psoft : 1; uint64_t ddr3psv : 1; uint64_t reserved_4_63 : 60; #endif } s; struct cvmx_lmcx_reset_ctl_s cn63xx; struct cvmx_lmcx_reset_ctl_s cn63xxp1; }; typedef union cvmx_lmcx_reset_ctl cvmx_lmcx_reset_ctl_t; /** * cvmx_lmc#_rlevel_ctl */ union cvmx_lmcx_rlevel_ctl { uint64_t u64; struct cvmx_lmcx_rlevel_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_22_63 : 42; uint64_t delay_unload_3 : 1; /**< When set, unload the PHY silo one cycle later during read-leveling if LMC*_RLEVEL_RANKi[BYTE*<1:0>] = 3 DELAY_UNLOAD_3 should normally be set, particularly at higher speeds. */ uint64_t delay_unload_2 : 1; /**< When set, unload the PHY silo one cycle later during read-leveling if LMC*_RLEVEL_RANKi[BYTE*<1:0>] = 2 DELAY_UNLOAD_2 should normally not be set. */ uint64_t delay_unload_1 : 1; /**< When set, unload the PHY silo one cycle later during read-leveling if LMC*_RLEVEL_RANKi[BYTE*<1:0>] = 1 DELAY_UNLOAD_1 should normally not be set. */ uint64_t delay_unload_0 : 1; /**< When set, unload the PHY silo one cycle later during read-leveling if LMC*_RLEVEL_RANKi[BYTE*<1:0>] = 0 DELAY_UNLOAD_0 should normally not be set. */ uint64_t bitmask : 8; /**< Mask to select bit lanes on which read-leveling feedback is returned when OR_DIS is set to 1 */ uint64_t or_dis : 1; /**< Disable or'ing of bits in a byte lane when computing the read-leveling bitmask OR_DIS should normally not be set. */ uint64_t offset_en : 1; /**< When set, LMC attempts to select the read-leveling setting that is LMC*RLEVEL_CTL[OFFSET] settings earlier than the last passing read-leveling setting in the largest contiguous sequence of passing settings. When clear, or if the setting selected by LMC*RLEVEL_CTL[OFFSET] did not pass, LMC selects the middle setting in the largest contiguous sequence of passing settings, rounding earlier when necessary. */ uint64_t offset : 4; /**< The offset used when LMC*RLEVEL_CTL[OFFSET] is set */ uint64_t byte : 4; /**< 0 <= BYTE <= 8 Byte index for which bitmask results are saved in LMC*_RLEVEL_DBG */ #else uint64_t byte : 4; uint64_t offset : 4; uint64_t offset_en : 1; uint64_t or_dis : 1; uint64_t bitmask : 8; uint64_t delay_unload_0 : 1; uint64_t delay_unload_1 : 1; uint64_t delay_unload_2 : 1; uint64_t delay_unload_3 : 1; uint64_t reserved_22_63 : 42; #endif } s; struct cvmx_lmcx_rlevel_ctl_s cn63xx; struct cvmx_lmcx_rlevel_ctl_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_9_63 : 55; uint64_t offset_en : 1; /**< When set, LMC attempts to select the read-leveling setting that is LMC*RLEVEL_CTL[OFFSET] settings earlier than the last passing read-leveling setting in the largest contiguous sequence of passing settings. When clear, or if the setting selected by LMC*RLEVEL_CTL[OFFSET] did not pass, LMC selects the middle setting in the largest contiguous sequence of passing settings, rounding earlier when necessary. */ uint64_t offset : 4; /**< The offset used when LMC*RLEVEL_CTL[OFFSET] is set */ uint64_t byte : 4; /**< 0 <= BYTE <= 8 Byte index for which bitmask results are saved in LMC*_RLEVEL_DBG */ #else uint64_t byte : 4; uint64_t offset : 4; uint64_t offset_en : 1; uint64_t reserved_9_63 : 55; #endif } cn63xxp1; }; typedef union cvmx_lmcx_rlevel_ctl cvmx_lmcx_rlevel_ctl_t; /** * cvmx_lmc#_rlevel_dbg * * Notes: * A given read of LMC*_RLEVEL_DBG returns the read-leveling pass/fail results for all possible * delay settings (i.e. the BITMASK) for only one byte in the last rank that the HW read-leveled. * LMC*_RLEVEL_CTL[BYTE] selects the particular byte. * * To get these pass/fail results for another different rank, you must run the hardware read-leveling * again. For example, it is possible to get the BITMASK results for every byte of every rank * if you run read-leveling separately for each rank, probing LMC*_RLEVEL_DBG between each * read-leveling. */ union cvmx_lmcx_rlevel_dbg { uint64_t u64; struct cvmx_lmcx_rlevel_dbg_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t bitmask : 64; /**< Bitmask generated during deskew settings sweep BITMASK[n]=0 means deskew setting n failed BITMASK[n]=1 means deskew setting n passed for 0 <= n <= 63 */ #else uint64_t bitmask : 64; #endif } s; struct cvmx_lmcx_rlevel_dbg_s cn63xx; struct cvmx_lmcx_rlevel_dbg_s cn63xxp1; }; typedef union cvmx_lmcx_rlevel_dbg cvmx_lmcx_rlevel_dbg_t; /** * cvmx_lmc#_rlevel_rank# * * Notes: * This is four CSRs per LMC, one per each rank. * * Deskew setting is measured in units of 1/4 CK, so the above BYTE* values can range over 16 CKs. * * Each CSR is written by HW during a read-leveling sequence for the rank. (HW sets STATUS==3 after HW read-leveling completes for the rank.) * If HW is unable to find a match per LMC*_RLEVEL_CTL[OFFSET_ENA] and LMC*_RLEVEL_CTL[OFFSET], then HW will set LMC*_RLEVEL_RANKi[BYTE*<5:0>] * to 0. * * Each CSR may also be written by SW, but not while a read-leveling sequence is in progress. (HW sets STATUS==1 after a CSR write.) * * SW initiates a HW read-leveling sequence by programming LMC*_RLEVEL_CTL and writing INIT_START=1 with SEQUENCE=1. * See LMC*_RLEVEL_CTL. * * LMC*_RLEVEL_RANKi values for ranks i without attached DRAM should be set such that * they do not increase the range of possible BYTE values for any byte * lane. The easiest way to do this is to set * LMC*_RLEVEL_RANKi = LMC*_RLEVEL_RANKj, * where j is some rank with attached DRAM whose LMC*_RLEVEL_RANKj is already fully initialized. */ union cvmx_lmcx_rlevel_rankx { uint64_t u64; struct cvmx_lmcx_rlevel_rankx_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_56_63 : 8; uint64_t status : 2; /**< Indicates status of the read-levelling and where the BYTE* programmings in <35:0> came from: 0 = BYTE* values are their reset value 1 = BYTE* values were set via a CSR write to this register 2 = read-leveling sequence currently in progress (BYTE* values are unpredictable) 3 = BYTE* values came from a complete read-leveling sequence */ uint64_t byte8 : 6; /**< Deskew setting When ECC DRAM is not present (i.e. when DRAM is not attached to chip signals DDR_CBS_0_* and DDR_CB[7:0]), SW should write BYTE8 to a value that does not increase the range of possible BYTE* values. The easiest way to do this is to set LMC*_RLEVEL_RANK*[BYTE8] = LMC*_RLEVEL_RANK*[BYTE0] when there is no ECC DRAM, using the final BYTE0 value. */ uint64_t byte7 : 6; /**< Deskew setting */ uint64_t byte6 : 6; /**< Deskew setting */ uint64_t byte5 : 6; /**< Deskew setting */ uint64_t byte4 : 6; /**< Deskew setting */ uint64_t byte3 : 6; /**< Deskew setting */ uint64_t byte2 : 6; /**< Deskew setting */ uint64_t byte1 : 6; /**< Deskew setting */ uint64_t byte0 : 6; /**< Deskew setting */ #else uint64_t byte0 : 6; uint64_t byte1 : 6; uint64_t byte2 : 6; uint64_t byte3 : 6; uint64_t byte4 : 6; uint64_t byte5 : 6; uint64_t byte6 : 6; uint64_t byte7 : 6; uint64_t byte8 : 6; uint64_t status : 2; uint64_t reserved_56_63 : 8; #endif } s; struct cvmx_lmcx_rlevel_rankx_s cn63xx; struct cvmx_lmcx_rlevel_rankx_s cn63xxp1; }; typedef union cvmx_lmcx_rlevel_rankx cvmx_lmcx_rlevel_rankx_t; /** * cvmx_lmc#_rodt_comp_ctl * * LMC_RODT_COMP_CTL = LMC Compensation control * */ union cvmx_lmcx_rodt_comp_ctl { uint64_t u64; struct cvmx_lmcx_rodt_comp_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_17_63 : 47; uint64_t enable : 1; /**< 0=not enabled, 1=enable */ uint64_t reserved_12_15 : 4; uint64_t nctl : 4; /**< Compensation control bits */ uint64_t reserved_5_7 : 3; uint64_t pctl : 5; /**< Compensation control bits */ #else uint64_t pctl : 5; uint64_t reserved_5_7 : 3; uint64_t nctl : 4; uint64_t reserved_12_15 : 4; uint64_t enable : 1; uint64_t reserved_17_63 : 47; #endif } s; struct cvmx_lmcx_rodt_comp_ctl_s cn50xx; struct cvmx_lmcx_rodt_comp_ctl_s cn52xx; struct cvmx_lmcx_rodt_comp_ctl_s cn52xxp1; struct cvmx_lmcx_rodt_comp_ctl_s cn56xx; struct cvmx_lmcx_rodt_comp_ctl_s cn56xxp1; struct cvmx_lmcx_rodt_comp_ctl_s cn58xx; struct cvmx_lmcx_rodt_comp_ctl_s cn58xxp1; }; typedef union cvmx_lmcx_rodt_comp_ctl cvmx_lmcx_rodt_comp_ctl_t; /** * cvmx_lmc#_rodt_ctl * * LMC_RODT_CTL = Obsolete LMC Read OnDieTermination control * See the description in LMC_WODT_CTL1. On Reads, Octeon only supports turning on ODT's in * the lower 2 DIMM's with the masks as below. * * Notes: * When a given RANK in position N is selected, the RODT _HI and _LO masks for that position are used. * Mask[3:0] is used for RODT control of the RANKs in positions 3, 2, 1, and 0, respectively. * In 64b mode, DIMMs are assumed to be ordered in the following order: * position 3: [unused , DIMM1_RANK1_LO] * position 2: [unused , DIMM1_RANK0_LO] * position 1: [unused , DIMM0_RANK1_LO] * position 0: [unused , DIMM0_RANK0_LO] * In 128b mode, DIMMs are assumed to be ordered in the following order: * position 3: [DIMM3_RANK1_HI, DIMM1_RANK1_LO] * position 2: [DIMM3_RANK0_HI, DIMM1_RANK0_LO] * position 1: [DIMM2_RANK1_HI, DIMM0_RANK1_LO] * position 0: [DIMM2_RANK0_HI, DIMM0_RANK0_LO] */ union cvmx_lmcx_rodt_ctl { uint64_t u64; struct cvmx_lmcx_rodt_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t rodt_hi3 : 4; /**< Read ODT mask for position 3, data[127:64] */ uint64_t rodt_hi2 : 4; /**< Read ODT mask for position 2, data[127:64] */ uint64_t rodt_hi1 : 4; /**< Read ODT mask for position 1, data[127:64] */ uint64_t rodt_hi0 : 4; /**< Read ODT mask for position 0, data[127:64] */ uint64_t rodt_lo3 : 4; /**< Read ODT mask for position 3, data[ 63: 0] */ uint64_t rodt_lo2 : 4; /**< Read ODT mask for position 2, data[ 63: 0] */ uint64_t rodt_lo1 : 4; /**< Read ODT mask for position 1, data[ 63: 0] */ uint64_t rodt_lo0 : 4; /**< Read ODT mask for position 0, data[ 63: 0] */ #else uint64_t rodt_lo0 : 4; uint64_t rodt_lo1 : 4; uint64_t rodt_lo2 : 4; uint64_t rodt_lo3 : 4; uint64_t rodt_hi0 : 4; uint64_t rodt_hi1 : 4; uint64_t rodt_hi2 : 4; uint64_t rodt_hi3 : 4; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_rodt_ctl_s cn30xx; struct cvmx_lmcx_rodt_ctl_s cn31xx; struct cvmx_lmcx_rodt_ctl_s cn38xx; struct cvmx_lmcx_rodt_ctl_s cn38xxp2; struct cvmx_lmcx_rodt_ctl_s cn50xx; struct cvmx_lmcx_rodt_ctl_s cn52xx; struct cvmx_lmcx_rodt_ctl_s cn52xxp1; struct cvmx_lmcx_rodt_ctl_s cn56xx; struct cvmx_lmcx_rodt_ctl_s cn56xxp1; struct cvmx_lmcx_rodt_ctl_s cn58xx; struct cvmx_lmcx_rodt_ctl_s cn58xxp1; }; typedef union cvmx_lmcx_rodt_ctl cvmx_lmcx_rodt_ctl_t; /** * cvmx_lmc#_rodt_mask * * LMC_RODT_MASK = LMC Read OnDieTermination mask * System designers may desire to terminate DQ/DQS lines for higher frequency DDR operations * especially on a multi-rank system. DDR3 DQ/DQS I/O's have built in * Termination resistor that can be turned on or off by the controller, after meeting tAOND and tAOF * timing requirements. Each Rank has its own ODT pin that fans out to all the memory parts * in that DIMM. System designers may prefer different combinations of ODT ON's for reads * into different ranks. Octeon supports full programmability by way of the mask register below. * Each Rank position has its own 8-bit programmable field. * When the controller does a read to that rank, it sets the 4 ODT pins to the MASK pins below. * For eg., When doing a read from Rank0, a system designer may desire to terminate the lines * with the resistor on DIMM0/Rank1. The mask RODT_D0_R0 would then be [00000010]. * Octeon drives the appropriate mask values on the ODT pins by default. If this feature is not * required, write 0 in this register. Note that, as per the DDR3 specifications, the ODT pin * for the rank that is being read should always be 0. * * Notes: * When a given RANK is selected, the RODT mask for that RANK is used. The resulting RODT mask is * driven to the DIMMs in the following manner: * RANK_ENA=1 RANK_ENA=0 * Mask[3] -> DIMM1_ODT_1 MBZ * Mask[2] -> DIMM1_ODT_0 DIMM1_ODT_0 * Mask[1] -> DIMM0_ODT_1 MBZ * Mask[0] -> DIMM0_ODT_0 DIMM0_ODT_0 * * LMC always reads entire cache blocks and always reads them via two consecutive * read CAS operations to the same rank+bank+row spaced exactly 4 CK's apart. * When a RODT mask bit is set, LMC asserts the OCTEON ODT output * pin(s) starting (CL - CWL) CK's after the first read CAS operation. Then, OCTEON * normally continues to assert the ODT output pin(s) for 9+LMC*_CONTROL[RODT_BPRCH] more CK's * - for a total of 10+LMC*_CONTROL[RODT_BPRCH] CK's for the entire cache block read - * through the second read CAS operation of the cache block, * satisfying the 6 CK DDR3 ODTH8 requirements. * But it is possible for OCTEON to issue two cache block reads separated by as few as * RtR = 8 or 9 (10 if LMC*_CONTROL[RODT_BPRCH]=1) CK's. In that case, OCTEON asserts the ODT output pin(s) * for the RODT mask of the first cache block read for RtR CK's, then asserts * the ODT output pin(s) for the RODT mask of the second cache block read for 10+LMC*_CONTROL[RODT_BPRCH] CK's * (or less if a third cache block read follows within 8 or 9 (or 10) CK's of this second cache block read). * Note that it may be necessary to force LMC to space back-to-back cache block reads * to different ranks apart by at least 10+LMC*_CONTROL[RODT_BPRCH] CK's to prevent DDR3 ODTH8 violations. */ union cvmx_lmcx_rodt_mask { uint64_t u64; struct cvmx_lmcx_rodt_mask_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t rodt_d3_r1 : 8; /**< Read ODT mask DIMM3, RANK1/DIMM3 in SingleRanked *UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d3_r0 : 8; /**< Read ODT mask DIMM3, RANK0 *UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d2_r1 : 8; /**< Read ODT mask DIMM2, RANK1/DIMM2 in SingleRanked *UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d2_r0 : 8; /**< Read ODT mask DIMM2, RANK0 *UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d1_r1 : 8; /**< Read ODT mask DIMM1, RANK1/DIMM1 in SingleRanked if (RANK_ENA) then RODT_D1_R1[3] must be 0 else RODT_D1_R1[3:0] is not used and MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d1_r0 : 8; /**< Read ODT mask DIMM1, RANK0 if (RANK_ENA) then RODT_D1_RO[2] must be 0 else RODT_D1_RO[3:2,1] must be 0 *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d0_r1 : 8; /**< Read ODT mask DIMM0, RANK1/DIMM0 in SingleRanked if (RANK_ENA) then RODT_D0_R1[1] must be 0 else RODT_D0_R1[3:0] is not used and MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t rodt_d0_r0 : 8; /**< Read ODT mask DIMM0, RANK0 if (RANK_ENA) then RODT_D0_RO[0] must be 0 else RODT_D0_RO[1:0,3] must be 0 *Upper 4 bits UNUSED IN 63xx, and MBZ* */ #else uint64_t rodt_d0_r0 : 8; uint64_t rodt_d0_r1 : 8; uint64_t rodt_d1_r0 : 8; uint64_t rodt_d1_r1 : 8; uint64_t rodt_d2_r0 : 8; uint64_t rodt_d2_r1 : 8; uint64_t rodt_d3_r0 : 8; uint64_t rodt_d3_r1 : 8; #endif } s; struct cvmx_lmcx_rodt_mask_s cn63xx; struct cvmx_lmcx_rodt_mask_s cn63xxp1; }; typedef union cvmx_lmcx_rodt_mask cvmx_lmcx_rodt_mask_t; /** * cvmx_lmc#_slot_ctl0 * * LMC_SLOT_CTL0 = LMC Slot Control0 * This register is an assortment of various control fields needed by the memory controller * * Notes: * If SW has not previously written to this register (since the last DRESET), * HW updates the fields in this register to the minimum allowed value * when any of LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn, LMC*_CONTROL and * LMC*_MODEREG_PARAMS0 CSR's change. Ideally, only read this register * after LMC has been initialized and LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn * have valid data. * * The interpretation of the fields in this CSR depends on LMC*_CONFIG[DDR2T]: * - If LMC*_CONFIG[DDR2T]=1, (FieldValue + 4) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. * - If LMC*_CONFIG[DDR2T]=0, (FieldValue + 3) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. FieldValue = 0 is always illegal in this * case. * * The hardware-calculated minimums are: * * min R2R_INIT = 1 - LMC*_CONFIG[DDR2T] * min R2W_INIT = 5 - LMC*_CONFIG[DDR2T] + (RL + MaxRdSkew) - (WL + MinWrSkew) + LMC*_CONTROL[BPRCH] * min W2R_INIT = 2 - LMC*_CONFIG[DDR2T] + LMC*_TIMING_PARAMS1[TWTR] + WL * min W2W_INIT = 1 - LMC*_CONFIG[DDR2T] * * where * * RL = CL + AL (LMC*_MODEREG_PARAMS0[CL] selects CL, LMC*_MODEREG_PARAMS0[AL] selects AL) * WL = CWL + AL (LMC*_MODEREG_PARAMS0[CWL] selects CWL) * MaxRdSkew = max(LMC*_RLEVEL_RANKi[BYTEj]/4) + 1 (max is across all ranks i (0..3) and bytes j (0..8)) * MinWrSkew = min(LMC*_WLEVEL_RANKi[BYTEj]/8) - LMC*_CONFIG[EARLY_DQX] (min is across all ranks i (0..3) and bytes j (0..8)) * * R2W_INIT has 1 CK cycle built in for OCTEON-internal ODT settling/channel turnaround time. */ union cvmx_lmcx_slot_ctl0 { uint64_t u64; struct cvmx_lmcx_slot_ctl0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_24_63 : 40; uint64_t w2w_init : 6; /**< Write-to-write spacing control for back to back accesses to the same rank and DIMM */ uint64_t w2r_init : 6; /**< Write-to-read spacing control for back to back accesses to the same rank and DIMM */ uint64_t r2w_init : 6; /**< Read-to-write spacing control for back to back accesses to the same rank and DIMM */ uint64_t r2r_init : 6; /**< Read-to-read spacing control for back to back accesses to the same rank and DIMM */ #else uint64_t r2r_init : 6; uint64_t r2w_init : 6; uint64_t w2r_init : 6; uint64_t w2w_init : 6; uint64_t reserved_24_63 : 40; #endif } s; struct cvmx_lmcx_slot_ctl0_s cn63xx; struct cvmx_lmcx_slot_ctl0_s cn63xxp1; }; typedef union cvmx_lmcx_slot_ctl0 cvmx_lmcx_slot_ctl0_t; /** * cvmx_lmc#_slot_ctl1 * * LMC_SLOT_CTL1 = LMC Slot Control1 * This register is an assortment of various control fields needed by the memory controller * * Notes: * If SW has not previously written to this register (since the last DRESET), * HW updates the fields in this register to the minimum allowed value * when any of LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn, LMC*_CONTROL and * LMC*_MODEREG_PARAMS0 CSR's change. Ideally, only read this register * after LMC has been initialized and LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn * have valid data. * * The interpretation of the fields in this CSR depends on LMC*_CONFIG[DDR2T]: * - If LMC*_CONFIG[DDR2T]=1, (FieldValue + 4) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. * - If LMC*_CONFIG[DDR2T]=0, (FieldValue + 3) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. FieldValue = 0 is always illegal in this * case. * * The hardware-calculated minimums are: * * min R2R_XRANK_INIT = 2 - LMC*_CONFIG[DDR2T] + MaxRdSkew - MinRdSkew * min R2W_XRANK_INIT = 5 - LMC*_CONFIG[DDR2T] + (RL + MaxRdSkew) - (WL + MinWrSkew) + LMC*_CONTROL[BPRCH] * min W2R_XRANK_INIT = 3 - LMC*_CONFIG[DDR2T] + MaxWrSkew + LMC*_CONTROL[FPRCH2] * min W2W_XRANK_INIT = 4 - LMC*_CONFIG[DDR2T] + MaxWrSkew - MinWrSkew * * where * * RL = CL + AL (LMC*_MODEREG_PARAMS0[CL] selects CL, LMC*_MODEREG_PARAMS0[AL] selects AL) * WL = CWL + AL (LMC*_MODEREG_PARAMS0[CWL] selects CWL) * MinRdSkew = min(LMC*_RLEVEL_RANKi[BYTEj]/4) (min is across all ranks i (0..3) and bytes j (0..8)) * MaxRdSkew = max(LMC*_RLEVEL_RANKi[BYTEj]/4) + 1 (max is across all ranks i (0..3) and bytes j (0..8)) * MinWrSkew = min(LMC*_WLEVEL_RANKi[BYTEj]/8) - LMC*_CONFIG[EARLY_DQX] (min is across all ranks i (0..3) and bytes j (0..8)) * MaxWrSkew = max(LMC*_WLEVEL_RANKi[BYTEj]/8) - LMC*_CONFIG[EARLY_DQX] + 1 (max is across all ranks i (0..3) and bytes j (0..8)) * * R2W_XRANK_INIT has 1 extra CK cycle built in for OCTEON-internal ODT settling/channel turnaround time. * * W2R_XRANK_INIT has 1 extra CK cycle built in for channel turnaround time. */ union cvmx_lmcx_slot_ctl1 { uint64_t u64; struct cvmx_lmcx_slot_ctl1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_24_63 : 40; uint64_t w2w_xrank_init : 6; /**< Write-to-write spacing control for back to back accesses across ranks of the same DIMM */ uint64_t w2r_xrank_init : 6; /**< Write-to-read spacing control for back to back accesses across ranks of the same DIMM */ uint64_t r2w_xrank_init : 6; /**< Read-to-write spacing control for back to back accesses across ranks of the same DIMM */ uint64_t r2r_xrank_init : 6; /**< Read-to-read spacing control for back to back accesses across ranks of the same DIMM */ #else uint64_t r2r_xrank_init : 6; uint64_t r2w_xrank_init : 6; uint64_t w2r_xrank_init : 6; uint64_t w2w_xrank_init : 6; uint64_t reserved_24_63 : 40; #endif } s; struct cvmx_lmcx_slot_ctl1_s cn63xx; struct cvmx_lmcx_slot_ctl1_s cn63xxp1; }; typedef union cvmx_lmcx_slot_ctl1 cvmx_lmcx_slot_ctl1_t; /** * cvmx_lmc#_slot_ctl2 * * LMC_SLOT_CTL2 = LMC Slot Control2 * This register is an assortment of various control fields needed by the memory controller * * Notes: * If SW has not previously written to this register (since the last DRESET), * HW updates the fields in this register to the minimum allowed value * when any of LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn, LMC*_CONTROL and * LMC*_MODEREG_PARAMS0 CSR's change. Ideally, only read this register * after LMC has been initialized and LMC*_RLEVEL_RANKn, LMC*_WLEVEL_RANKn * have valid data. * * The interpretation of the fields in this CSR depends on LMC*_CONFIG[DDR2T]: * - If LMC*_CONFIG[DDR2T]=1, (FieldValue + 4) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. * - If LMC*_CONFIG[DDR2T]=0, (FieldValue + 3) is the minimum CK cycles * between when the DRAM part registers CAS commands of the 1st and 2nd types * from different cache blocks. FieldValue = 0 is always illegal in this * case. * * The hardware-calculated minimums are: * * min R2R_XDIMM_INIT = 3 - LMC*_CONFIG[DDR2T] + MaxRdSkew - MinRdSkew * min R2W_XDIMM_INIT = 6 - LMC*_CONFIG[DDR2T] + (RL + MaxRdSkew) - (WL + MinWrSkew) + LMC*_CONTROL[BPRCH] * min W2R_XDIMM_INIT = 3 - LMC*_CONFIG[DDR2T] + MaxWrSkew + LMC*_CONTROL[FPRCH2] * min W2W_XDIMM_INIT = 5 - LMC*_CONFIG[DDR2T] + MaxWrSkew - MinWrSkew * * where * * RL = CL + AL (LMC*_MODEREG_PARAMS0[CL] selects CL, LMC*_MODEREG_PARAMS0[AL] selects AL) * WL = CWL + AL (LMC*_MODEREG_PARAMS0[CWL] selects CWL) * MinRdSkew = min(LMC*_RLEVEL_RANKi[BYTEj]/4) (min is across all ranks i (0..3) and bytes j (0..8)) * MaxRdSkew = max(LMC*_RLEVEL_RANKi[BYTEj]/4) + 1 (max is across all ranks i (0..3) and bytes j (0..8)) * MinWrSkew = min(LMC*_WLEVEL_RANKi[BYTEj]/8) - LMC*_CONFIG[EARLY_DQX] (min is across all ranks i (0..3) and bytes j (0..8)) * MaxWrSkew = max(LMC*_WLEVEL_RANKi[BYTEj]/8) - LMC*_CONFIG[EARLY_DQX] + 1 (max is across all ranks i (0..3) and bytes j (0..8)) * * R2W_XDIMM_INIT has 2 extra CK cycles built in for OCTEON-internal ODT settling/channel turnaround time. * * R2R_XDIMM_INIT, W2R_XRANK_INIT, W2W_XDIMM_INIT have 1 extra CK cycle built in for channel turnaround time. */ union cvmx_lmcx_slot_ctl2 { uint64_t u64; struct cvmx_lmcx_slot_ctl2_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_24_63 : 40; uint64_t w2w_xdimm_init : 6; /**< Write-to-write spacing control for back to back accesses across DIMMs */ uint64_t w2r_xdimm_init : 6; /**< Write-to-read spacing control for back to back accesses across DIMMs */ uint64_t r2w_xdimm_init : 6; /**< Read-to-write spacing control for back to back accesses across DIMMs */ uint64_t r2r_xdimm_init : 6; /**< Read-to-read spacing control for back to back accesses across DIMMs */ #else uint64_t r2r_xdimm_init : 6; uint64_t r2w_xdimm_init : 6; uint64_t w2r_xdimm_init : 6; uint64_t w2w_xdimm_init : 6; uint64_t reserved_24_63 : 40; #endif } s; struct cvmx_lmcx_slot_ctl2_s cn63xx; struct cvmx_lmcx_slot_ctl2_s cn63xxp1; }; typedef union cvmx_lmcx_slot_ctl2 cvmx_lmcx_slot_ctl2_t; /** * cvmx_lmc#_timing_params0 */ union cvmx_lmcx_timing_params0 { uint64_t u64; struct cvmx_lmcx_timing_params0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_47_63 : 17; uint64_t trp_ext : 1; /**< Indicates tRP constraints. Set [TRP_EXT[0:0], TRP[3:0]] (CSR field) = RNDUP[tRP(ns)/tCYC(ns)] + (RNDUP[tRTP(ns)/tCYC(ns)]-4)-1, where tRP, tRTP are from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP tRP=10-15ns TYP tRTP=max(4nCK, 7.5ns) */ uint64_t tcksre : 4; /**< Indicates tCKSRE constraints. Set TCKSRE (CSR field) = RNDUP[tCKSRE(ns)/tCYC(ns)]-1, where tCKSRE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, 10ns) */ uint64_t trp : 4; /**< Indicates tRP constraints. Set TRP (CSR field) = RNDUP[tRP(ns)/tCYC(ns)] + (RNDUP[tRTP(ns)/tCYC(ns)])-4)-1, where tRP, tRTP are from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP tRP=10-15ns TYP tRTP=max(4nCK, 7.5ns) */ uint64_t tzqinit : 4; /**< Indicates tZQINIT constraints. Set TZQINIT (CSR field) = RNDUP[tZQINIT(ns)/(256*tCYC(ns))], where tZQINIT is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) */ uint64_t tdllk : 4; /**< Indicates tDLLK constraints. Set TDLLK (CSR field) = RNDUP[tDLLK(ns)/(256*tCYC(ns))], where tDLLK is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) This parameter is used in self-refresh exit and assumed to be greater than tRFC */ uint64_t tmod : 4; /**< Indicates tMOD constraints. Set TMOD (CSR field) = RNDUP[tMOD(ns)/tCYC(ns)]-1, where tMOD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(12nCK, 15ns) */ uint64_t tmrd : 4; /**< Indicates tMRD constraints. Set TMRD (CSR field) = RNDUP[tMRD(ns)/tCYC(ns)]-1, where tMRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4nCK */ uint64_t txpr : 4; /**< Indicates tXPR constraints. Set TXPR (CSR field) = RNDUP[tXPR(ns)/(16*tCYC(ns))], where tXPR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, tRFC+10ns) */ uint64_t tcke : 4; /**< Indicates tCKE constraints. Set TCKE (CSR field) = RNDUP[tCKE(ns)/tCYC(ns)]-1, where tCKE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(3nCK, 7.5/5.625/5.625/5ns) */ uint64_t tzqcs : 4; /**< Indicates tZQCS constraints. Set TZQCS (CSR field) = RNDUP[tZQCS(ns)/(16*tCYC(ns))], where tZQCS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4 (equivalent to 64) */ uint64_t tckeon : 10; /**< Reserved. Should be written to zero. */ #else uint64_t tckeon : 10; uint64_t tzqcs : 4; uint64_t tcke : 4; uint64_t txpr : 4; uint64_t tmrd : 4; uint64_t tmod : 4; uint64_t tdllk : 4; uint64_t tzqinit : 4; uint64_t trp : 4; uint64_t tcksre : 4; uint64_t trp_ext : 1; uint64_t reserved_47_63 : 17; #endif } s; struct cvmx_lmcx_timing_params0_cn63xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_47_63 : 17; uint64_t trp_ext : 1; /**< Indicates tRP constraints. Set [TRP_EXT[0:0], TRP[3:0]] (CSR field) = RNDUP[tRP(ns)/tCYC(ns)] + (RNDUP[tRTP(ns)/tCYC(ns)]-4)-1, where tRP, tRTP are from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP tRP=10-15ns TYP tRTP=max(4nCK, 7.5ns) */ uint64_t tcksre : 4; /**< Indicates tCKSRE constraints. Set TCKSRE (CSR field) = RNDUP[tCKSRE(ns)/tCYC(ns)]-1, where tCKSRE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, 10ns) */ uint64_t trp : 4; /**< Indicates tRP constraints. Set [TRP_EXT[0:0], TRP[3:0]] (CSR field) = RNDUP[tRP(ns)/tCYC(ns)] + (RNDUP[tRTP(ns)/tCYC(ns)])-4)-1, where tRP, tRTP are from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP tRP=10-15ns TYP tRTP=max(4nCK, 7.5ns) */ uint64_t tzqinit : 4; /**< Indicates tZQINIT constraints. Set TZQINIT (CSR field) = RNDUP[tZQINIT(ns)/(256*tCYC(ns))], where tZQINIT is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) */ uint64_t tdllk : 4; /**< Indicates tDLLK constraints. Set TDLLK (CSR field) = RNDUP[tDLLK(ns)/(256*tCYC(ns))], where tDLLK is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) This parameter is used in self-refresh exit and assumed to be greater than tRFC */ uint64_t tmod : 4; /**< Indicates tMOD constraints. Set TMOD (CSR field) = RNDUP[tMOD(ns)/tCYC(ns)]-1, where tMOD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(12nCK, 15ns) */ uint64_t tmrd : 4; /**< Indicates tMRD constraints. Set TMRD (CSR field) = RNDUP[tMRD(ns)/tCYC(ns)]-1, where tMRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4nCK */ uint64_t txpr : 4; /**< Indicates tXPR constraints. Set TXPR (CSR field) = RNDUP[tXPR(ns)/(16*tCYC(ns))], where tXPR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, tRFC+10ns) */ uint64_t tcke : 4; /**< Indicates tCKE constraints. Set TCKE (CSR field) = RNDUP[tCKE(ns)/tCYC(ns)]-1, where tCKE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(3nCK, 7.5/5.625/5.625/5ns) */ uint64_t tzqcs : 4; /**< Indicates tZQCS constraints. Set TZQCS (CSR field) = RNDUP[tZQCS(ns)/(16*tCYC(ns))], where tZQCS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4 (equivalent to 64) */ uint64_t reserved_0_9 : 10; #else uint64_t reserved_0_9 : 10; uint64_t tzqcs : 4; uint64_t tcke : 4; uint64_t txpr : 4; uint64_t tmrd : 4; uint64_t tmod : 4; uint64_t tdllk : 4; uint64_t tzqinit : 4; uint64_t trp : 4; uint64_t tcksre : 4; uint64_t trp_ext : 1; uint64_t reserved_47_63 : 17; #endif } cn63xx; struct cvmx_lmcx_timing_params0_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_46_63 : 18; uint64_t tcksre : 4; /**< Indicates tCKSRE constraints. Set TCKSRE (CSR field) = RNDUP[tCKSRE(ns)/tCYC(ns)]-1, where tCKSRE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, 10ns) */ uint64_t trp : 4; /**< Indicates tRP constraints. Set TRP (CSR field) = RNDUP[tRP(ns)/tCYC(ns)] + (RNDUP[tRTP(ns)/tCYC(ns)])-4)-1, where tRP, tRTP are from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP tRP=10-15ns TYP tRTP=max(4nCK, 7.5ns) */ uint64_t tzqinit : 4; /**< Indicates tZQINIT constraints. Set TZQINIT (CSR field) = RNDUP[tZQINIT(ns)/(256*tCYC(ns))], where tZQINIT is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) */ uint64_t tdllk : 4; /**< Indicates tDLLK constraints. Set TDLLK (CSR field) = RNDUP[tDLLK(ns)/(256*tCYC(ns))], where tDLLK is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=2 (equivalent to 512) This parameter is used in self-refresh exit and assumed to be greater than tRFC */ uint64_t tmod : 4; /**< Indicates tMOD constraints. Set TMOD (CSR field) = RNDUP[tMOD(ns)/tCYC(ns)]-1, where tMOD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(12nCK, 15ns) */ uint64_t tmrd : 4; /**< Indicates tMRD constraints. Set TMRD (CSR field) = RNDUP[tMRD(ns)/tCYC(ns)]-1, where tMRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4nCK */ uint64_t txpr : 4; /**< Indicates tXPR constraints. Set TXPR (CSR field) = RNDUP[tXPR(ns)/(16*tCYC(ns))], where tXPR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(5nCK, tRFC+10ns) */ uint64_t tcke : 4; /**< Indicates tCKE constraints. Set TCKE (CSR field) = RNDUP[tCKE(ns)/tCYC(ns)]-1, where tCKE is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(3nCK, 7.5/5.625/5.625/5ns) */ uint64_t tzqcs : 4; /**< Indicates tZQCS constraints. Set TZQCS (CSR field) = RNDUP[tZQCS(ns)/(16*tCYC(ns))], where tZQCS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=4 (equivalent to 64) */ uint64_t tckeon : 10; /**< Reserved. Should be written to zero. */ #else uint64_t tckeon : 10; uint64_t tzqcs : 4; uint64_t tcke : 4; uint64_t txpr : 4; uint64_t tmrd : 4; uint64_t tmod : 4; uint64_t tdllk : 4; uint64_t tzqinit : 4; uint64_t trp : 4; uint64_t tcksre : 4; uint64_t reserved_46_63 : 18; #endif } cn63xxp1; }; typedef union cvmx_lmcx_timing_params0 cvmx_lmcx_timing_params0_t; /** * cvmx_lmc#_timing_params1 */ union cvmx_lmcx_timing_params1 { uint64_t u64; struct cvmx_lmcx_timing_params1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_47_63 : 17; uint64_t tras_ext : 1; /**< Indicates tRAS constraints. Set [TRAS_EXT[0:0], TRAS[4:0]] (CSR field) = RNDUP[tRAS(ns)/tCYC(ns)]-1, where tRAS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=35ns-9*tREFI - 000000: RESERVED - 000001: 2 tCYC - 000010: 3 tCYC - ... - 111111: 64 tCYC */ uint64_t txpdll : 5; /**< Indicates tXPDLL constraints. Set TXPDLL (CSR field) = RNDUP[tXPDLL(ns)/tCYC(ns)]-1, where tXPDLL is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(10nCK, 24ns) */ uint64_t tfaw : 5; /**< Indicates tFAW constraints. Set TFAW (CSR field) = RNDUP[tFAW(ns)/(4*tCYC(ns))], where tFAW is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=30-40ns */ uint64_t twldqsen : 4; /**< Indicates tWLDQSEN constraints. Set TWLDQSEN (CSR field) = RNDUP[tWLDQSEN(ns)/(4*tCYC(ns))], where tWLDQSEN is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(25nCK) */ uint64_t twlmrd : 4; /**< Indicates tWLMRD constraints. Set TWLMRD (CSR field) = RNDUP[tWLMRD(ns)/(4*tCYC(ns))], where tWLMRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(40nCK) */ uint64_t txp : 3; /**< Indicates tXP constraints. Set TXP (CSR field) = RNDUP[tXP(ns)/tCYC(ns)]-1, where tXP is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(3nCK, 7.5ns) */ uint64_t trrd : 3; /**< Indicates tRRD constraints. Set TRRD (CSR field) = RNDUP[tRRD(ns)/tCYC(ns)]-2, where tRRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(4nCK, 10ns) - 000: RESERVED - 001: 3 tCYC - ... - 110: 8 tCYC - 111: 9 tCYC */ uint64_t trfc : 5; /**< Indicates tRFC constraints. Set TRFC (CSR field) = RNDUP[tRFC(ns)/(8*tCYC(ns))], where tRFC is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=90-350ns - 00000: RESERVED - 00001: 8 tCYC - 00010: 16 tCYC - 00011: 24 tCYC - 00100: 32 tCYC - ... - 11110: 240 tCYC - 11111: 248 tCYC */ uint64_t twtr : 4; /**< Indicates tWTR constraints. Set TWTR (CSR field) = RNDUP[tWTR(ns)/tCYC(ns)]-1, where tWTR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(4nCK, 7.5ns) - 0000: RESERVED - 0001: 2 - ... - 0111: 8 - 1000-1111: RESERVED */ uint64_t trcd : 4; /**< Indicates tRCD constraints. Set TRCD (CSR field) = RNDUP[tRCD(ns)/tCYC(ns)], where tRCD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=10-15ns - 0000: RESERVED - 0001: 2 (2 is the smallest value allowed) - 0002: 2 - ... - 1001: 9 - 1010-1111: RESERVED In 2T mode, make this register TRCD-1, not going below 2. */ uint64_t tras : 5; /**< Indicates tRAS constraints. Set TRAS (CSR field) = RNDUP[tRAS(ns)/tCYC(ns)]-1, where tRAS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=35ns-9*tREFI - 00000: RESERVED - 00001: 2 tCYC - 00010: 3 tCYC - ... - 11111: 32 tCYC */ uint64_t tmprr : 4; /**< Indicates tMPRR constraints. Set TMPRR (CSR field) = RNDUP[tMPRR(ns)/tCYC(ns)]-1, where tMPRR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=1nCK */ #else uint64_t tmprr : 4; uint64_t tras : 5; uint64_t trcd : 4; uint64_t twtr : 4; uint64_t trfc : 5; uint64_t trrd : 3; uint64_t txp : 3; uint64_t twlmrd : 4; uint64_t twldqsen : 4; uint64_t tfaw : 5; uint64_t txpdll : 5; uint64_t tras_ext : 1; uint64_t reserved_47_63 : 17; #endif } s; struct cvmx_lmcx_timing_params1_s cn63xx; struct cvmx_lmcx_timing_params1_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_46_63 : 18; uint64_t txpdll : 5; /**< Indicates tXPDLL constraints. Set TXPDLL (CSR field) = RNDUP[tXPDLL(ns)/tCYC(ns)]-1, where tXPDLL is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(10nCK, 24ns) */ uint64_t tfaw : 5; /**< Indicates tFAW constraints. Set TFAW (CSR field) = RNDUP[tFAW(ns)/(4*tCYC(ns))], where tFAW is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=30-40ns */ uint64_t twldqsen : 4; /**< Indicates tWLDQSEN constraints. Set TWLDQSEN (CSR field) = RNDUP[tWLDQSEN(ns)/(4*tCYC(ns))], where tWLDQSEN is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(25nCK) */ uint64_t twlmrd : 4; /**< Indicates tWLMRD constraints. Set TWLMRD (CSR field) = RNDUP[tWLMRD(ns)/(4*tCYC(ns))], where tWLMRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(40nCK) */ uint64_t txp : 3; /**< Indicates tXP constraints. Set TXP (CSR field) = RNDUP[tXP(ns)/tCYC(ns)]-1, where tXP is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(3nCK, 7.5ns) */ uint64_t trrd : 3; /**< Indicates tRRD constraints. Set TRRD (CSR field) = RNDUP[tRRD(ns)/tCYC(ns)]-2, where tRRD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(4nCK, 10ns) - 000: RESERVED - 001: 3 tCYC - ... - 110: 8 tCYC - 111: 9 tCYC */ uint64_t trfc : 5; /**< Indicates tRFC constraints. Set TRFC (CSR field) = RNDUP[tRFC(ns)/(8*tCYC(ns))], where tRFC is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=90-350ns - 00000: RESERVED - 00001: 8 tCYC - 00010: 16 tCYC - 00011: 24 tCYC - 00100: 32 tCYC - ... - 11110: 240 tCYC - 11111: 248 tCYC */ uint64_t twtr : 4; /**< Indicates tWTR constraints. Set TWTR (CSR field) = RNDUP[tWTR(ns)/tCYC(ns)]-1, where tWTR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=max(4nCK, 7.5ns) - 0000: RESERVED - 0001: 2 - ... - 0111: 8 - 1000-1111: RESERVED */ uint64_t trcd : 4; /**< Indicates tRCD constraints. Set TRCD (CSR field) = RNDUP[tRCD(ns)/tCYC(ns)], where tRCD is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=10-15ns - 0000: RESERVED - 0001: 2 (2 is the smallest value allowed) - 0002: 2 - ... - 1001: 9 - 1010-1111: RESERVED In 2T mode, make this register TRCD-1, not going below 2. */ uint64_t tras : 5; /**< Indicates tRAS constraints. Set TRAS (CSR field) = RNDUP[tRAS(ns)/tCYC(ns)]-1, where tRAS is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=35ns-9*tREFI - 00000: RESERVED - 00001: 2 tCYC - 00010: 3 tCYC - ... - 11111: 32 tCYC */ uint64_t tmprr : 4; /**< Indicates tMPRR constraints. Set TMPRR (CSR field) = RNDUP[tMPRR(ns)/tCYC(ns)]-1, where tMPRR is from the DDR3 spec, and tCYC(ns) is the DDR clock frequency (not data rate). TYP=1nCK */ #else uint64_t tmprr : 4; uint64_t tras : 5; uint64_t trcd : 4; uint64_t twtr : 4; uint64_t trfc : 5; uint64_t trrd : 3; uint64_t txp : 3; uint64_t twlmrd : 4; uint64_t twldqsen : 4; uint64_t tfaw : 5; uint64_t txpdll : 5; uint64_t reserved_46_63 : 18; #endif } cn63xxp1; }; typedef union cvmx_lmcx_timing_params1 cvmx_lmcx_timing_params1_t; /** * cvmx_lmc#_tro_ctl * * LMC_TRO_CTL = LMC Temperature Ring Osc Control * This register is an assortment of various control fields needed to control the temperature ring oscillator * * Notes: * To bring up the temperature ring oscillator, write TRESET to 0, and follow by initializing RCLK_CNT to desired * value */ union cvmx_lmcx_tro_ctl { uint64_t u64; struct cvmx_lmcx_tro_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_33_63 : 31; uint64_t rclk_cnt : 32; /**< rclk counter */ uint64_t treset : 1; /**< Reset ring oscillator */ #else uint64_t treset : 1; uint64_t rclk_cnt : 32; uint64_t reserved_33_63 : 31; #endif } s; struct cvmx_lmcx_tro_ctl_s cn63xx; struct cvmx_lmcx_tro_ctl_s cn63xxp1; }; typedef union cvmx_lmcx_tro_ctl cvmx_lmcx_tro_ctl_t; /** * cvmx_lmc#_tro_stat * * LMC_TRO_STAT = LMC Temperature Ring Osc Status * This register is an assortment of various control fields needed to control the temperature ring oscillator */ union cvmx_lmcx_tro_stat { uint64_t u64; struct cvmx_lmcx_tro_stat_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t ring_cnt : 32; /**< ring counter */ #else uint64_t ring_cnt : 32; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_tro_stat_s cn63xx; struct cvmx_lmcx_tro_stat_s cn63xxp1; }; typedef union cvmx_lmcx_tro_stat cvmx_lmcx_tro_stat_t; /** * cvmx_lmc#_wlevel_ctl */ union cvmx_lmcx_wlevel_ctl { uint64_t u64; struct cvmx_lmcx_wlevel_ctl_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_22_63 : 42; uint64_t rtt_nom : 3; /**< RTT_NOM LMC writes a decoded value to MR1[Rtt_Nom] of the rank during write leveling. Per JEDEC DDR3 specifications, only values MR1[Rtt_Nom] = 1 (RQZ/4), 2 (RQZ/2), or 3 (RQZ/6) are allowed during write leveling with output buffer enabled. 000 : LMC writes 001 (RZQ/4) to MR1[Rtt_Nom] 001 : LMC writes 010 (RZQ/2) to MR1[Rtt_Nom] 010 : LMC writes 011 (RZQ/6) to MR1[Rtt_Nom] 011 : LMC writes 100 (RZQ/12) to MR1[Rtt_Nom] 100 : LMC writes 101 (RZQ/8) to MR1[Rtt_Nom] 101 : LMC writes 110 (Rsvd) to MR1[Rtt_Nom] 110 : LMC writes 111 (Rsvd) to MR1[Rtt_Nom] 111 : LMC writes 000 (Disabled) to MR1[Rtt_Nom] */ uint64_t bitmask : 8; /**< Mask to select bit lanes on which write-leveling feedback is returned when OR_DIS is set to 1 */ uint64_t or_dis : 1; /**< Disable or'ing of bits in a byte lane when computing the write-leveling bitmask */ uint64_t sset : 1; /**< Run write-leveling on the current setting only. */ uint64_t lanemask : 9; /**< One-hot mask to select byte lane to be leveled by the write-leveling sequence Used with x16 parts where the upper and lower byte lanes need to be leveled independently */ #else uint64_t lanemask : 9; uint64_t sset : 1; uint64_t or_dis : 1; uint64_t bitmask : 8; uint64_t rtt_nom : 3; uint64_t reserved_22_63 : 42; #endif } s; struct cvmx_lmcx_wlevel_ctl_s cn63xx; struct cvmx_lmcx_wlevel_ctl_cn63xxp1 { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_10_63 : 54; uint64_t sset : 1; /**< Run write-leveling on the current setting only. */ uint64_t lanemask : 9; /**< One-hot mask to select byte lane to be leveled by the write-leveling sequence Used with x16 parts where the upper and lower byte lanes need to be leveled independently */ #else uint64_t lanemask : 9; uint64_t sset : 1; uint64_t reserved_10_63 : 54; #endif } cn63xxp1; }; typedef union cvmx_lmcx_wlevel_ctl cvmx_lmcx_wlevel_ctl_t; /** * cvmx_lmc#_wlevel_dbg * * Notes: * A given write of LMC*_WLEVEL_DBG returns the write-leveling pass/fail results for all possible * delay settings (i.e. the BITMASK) for only one byte in the last rank that the HW write-leveled. * LMC*_WLEVEL_DBG[BYTE] selects the particular byte. * To get these pass/fail results for another different rank, you must run the hardware write-leveling * again. For example, it is possible to get the BITMASK results for every byte of every rank * if you run write-leveling separately for each rank, probing LMC*_WLEVEL_DBG between each * write-leveling. */ union cvmx_lmcx_wlevel_dbg { uint64_t u64; struct cvmx_lmcx_wlevel_dbg_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_12_63 : 52; uint64_t bitmask : 8; /**< Bitmask generated during deskew settings sweep if LMCX_WLEVEL_CTL[SSET]=0 BITMASK[n]=0 means deskew setting n failed BITMASK[n]=1 means deskew setting n passed for 0 <= n <= 7 BITMASK contains the first 8 results of the total 16 collected by LMC during the write-leveling sequence else if LMCX_WLEVEL_CTL[SSET]=1 BITMASK[0]=0 means curr deskew setting failed BITMASK[0]=1 means curr deskew setting passed */ uint64_t byte : 4; /**< 0 <= BYTE <= 8 */ #else uint64_t byte : 4; uint64_t bitmask : 8; uint64_t reserved_12_63 : 52; #endif } s; struct cvmx_lmcx_wlevel_dbg_s cn63xx; struct cvmx_lmcx_wlevel_dbg_s cn63xxp1; }; typedef union cvmx_lmcx_wlevel_dbg cvmx_lmcx_wlevel_dbg_t; /** * cvmx_lmc#_wlevel_rank# * * Notes: * This is four CSRs per LMC, one per each rank. * * Deskew setting is measured in units of 1/8 CK, so the above BYTE* values can range over 4 CKs. * * Assuming LMC*_WLEVEL_CTL[SSET]=0, the BYTE*<2:0> values are not used during write-leveling, and * they are over-written by the hardware as part of the write-leveling sequence. (HW sets STATUS==3 * after HW write-leveling completes for the rank). SW needs to set BYTE*<4:3> bits. * * Each CSR may also be written by SW, but not while a write-leveling sequence is in progress. (HW sets STATUS==1 after a CSR write.) * * SW initiates a HW write-leveling sequence by programming LMC*_WLEVEL_CTL and writing RANKMASK and INIT_START=1 with SEQUENCE=6 in LMC*_CONFIG. * LMC will then step through and accumulate write leveling results for 8 unique delay settings (twice), starting at a delay of * LMC*_WLEVEL_RANKn[BYTE*<4:3>]*8 CK increasing by 1/8 CK each setting. HW will then set LMC*_WLEVEL_RANKi[BYTE*<2:0>] to indicate the * first write leveling result of '1' that followed a reslt of '0' during the sequence by searching for a '1100' pattern in the generated * bitmask, except that LMC will always write LMC*_WLEVEL_RANKi[BYTE*<0>]=0. If HW is unable to find a match for a '1100' pattern, then HW will * set LMC*_WLEVEL_RANKi[BYTE*<2:0>] to 4. * See LMC*_WLEVEL_CTL. * * LMC*_WLEVEL_RANKi values for ranks i without attached DRAM should be set such that * they do not increase the range of possible BYTE values for any byte * lane. The easiest way to do this is to set * LMC*_WLEVEL_RANKi = LMC*_WLEVEL_RANKj, * where j is some rank with attached DRAM whose LMC*_WLEVEL_RANKj is already fully initialized. */ union cvmx_lmcx_wlevel_rankx { uint64_t u64; struct cvmx_lmcx_wlevel_rankx_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_47_63 : 17; uint64_t status : 2; /**< Indicates status of the write-leveling and where the BYTE* programmings in <44:0> came from: 0 = BYTE* values are their reset value 1 = BYTE* values were set via a CSR write to this register 2 = write-leveling sequence currently in progress (BYTE* values are unpredictable) 3 = BYTE* values came from a complete write-leveling sequence, irrespective of which lanes are masked via LMC*WLEVEL_CTL[LANEMASK] */ uint64_t byte8 : 5; /**< Deskew setting Bit 0 of BYTE8 must be zero during normal operation. When ECC DRAM is not present (i.e. when DRAM is not attached to chip signals DDR_CBS_0_* and DDR_CB[7:0]), SW should write BYTE8 with a value that does not increase the range of possible BYTE* values. The easiest way to do this is to set LMC*_WLEVEL_RANK*[BYTE8] = LMC*_WLEVEL_RANK*[BYTE0] when there is no ECC DRAM, using the final BYTE0 value. */ uint64_t byte7 : 5; /**< Deskew setting Bit 0 of BYTE7 must be zero during normal operation */ uint64_t byte6 : 5; /**< Deskew setting Bit 0 of BYTE6 must be zero during normal operation */ uint64_t byte5 : 5; /**< Deskew setting Bit 0 of BYTE5 must be zero during normal operation */ uint64_t byte4 : 5; /**< Deskew setting Bit 0 of BYTE4 must be zero during normal operation */ uint64_t byte3 : 5; /**< Deskew setting Bit 0 of BYTE3 must be zero during normal operation */ uint64_t byte2 : 5; /**< Deskew setting Bit 0 of BYTE2 must be zero during normal operation */ uint64_t byte1 : 5; /**< Deskew setting Bit 0 of BYTE1 must be zero during normal operation */ uint64_t byte0 : 5; /**< Deskew setting Bit 0 of BYTE0 must be zero during normal operation */ #else uint64_t byte0 : 5; uint64_t byte1 : 5; uint64_t byte2 : 5; uint64_t byte3 : 5; uint64_t byte4 : 5; uint64_t byte5 : 5; uint64_t byte6 : 5; uint64_t byte7 : 5; uint64_t byte8 : 5; uint64_t status : 2; uint64_t reserved_47_63 : 17; #endif } s; struct cvmx_lmcx_wlevel_rankx_s cn63xx; struct cvmx_lmcx_wlevel_rankx_s cn63xxp1; }; typedef union cvmx_lmcx_wlevel_rankx cvmx_lmcx_wlevel_rankx_t; /** * cvmx_lmc#_wodt_ctl0 * * LMC_WODT_CTL0 = LMC Write OnDieTermination control * See the description in LMC_WODT_CTL1. * * Notes: * Together, the LMC_WODT_CTL1 and LMC_WODT_CTL0 CSRs control the write ODT mask. See LMC_WODT_CTL1. * */ union cvmx_lmcx_wodt_ctl0 { uint64_t u64; struct cvmx_lmcx_wodt_ctl0_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_0_63 : 64; #else uint64_t reserved_0_63 : 64; #endif } s; struct cvmx_lmcx_wodt_ctl0_cn30xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t wodt_d1_r1 : 8; /**< Write ODT mask DIMM1, RANK1 */ uint64_t wodt_d1_r0 : 8; /**< Write ODT mask DIMM1, RANK0 */ uint64_t wodt_d0_r1 : 8; /**< Write ODT mask DIMM0, RANK1 */ uint64_t wodt_d0_r0 : 8; /**< Write ODT mask DIMM0, RANK0 */ #else uint64_t wodt_d0_r0 : 8; uint64_t wodt_d0_r1 : 8; uint64_t wodt_d1_r0 : 8; uint64_t wodt_d1_r1 : 8; uint64_t reserved_32_63 : 32; #endif } cn30xx; struct cvmx_lmcx_wodt_ctl0_cn30xx cn31xx; struct cvmx_lmcx_wodt_ctl0_cn38xx { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t wodt_hi3 : 4; /**< Write ODT mask for position 3, data[127:64] */ uint64_t wodt_hi2 : 4; /**< Write ODT mask for position 2, data[127:64] */ uint64_t wodt_hi1 : 4; /**< Write ODT mask for position 1, data[127:64] */ uint64_t wodt_hi0 : 4; /**< Write ODT mask for position 0, data[127:64] */ uint64_t wodt_lo3 : 4; /**< Write ODT mask for position 3, data[ 63: 0] */ uint64_t wodt_lo2 : 4; /**< Write ODT mask for position 2, data[ 63: 0] */ uint64_t wodt_lo1 : 4; /**< Write ODT mask for position 1, data[ 63: 0] */ uint64_t wodt_lo0 : 4; /**< Write ODT mask for position 0, data[ 63: 0] */ #else uint64_t wodt_lo0 : 4; uint64_t wodt_lo1 : 4; uint64_t wodt_lo2 : 4; uint64_t wodt_lo3 : 4; uint64_t wodt_hi0 : 4; uint64_t wodt_hi1 : 4; uint64_t wodt_hi2 : 4; uint64_t wodt_hi3 : 4; uint64_t reserved_32_63 : 32; #endif } cn38xx; struct cvmx_lmcx_wodt_ctl0_cn38xx cn38xxp2; struct cvmx_lmcx_wodt_ctl0_cn38xx cn50xx; struct cvmx_lmcx_wodt_ctl0_cn30xx cn52xx; struct cvmx_lmcx_wodt_ctl0_cn30xx cn52xxp1; struct cvmx_lmcx_wodt_ctl0_cn30xx cn56xx; struct cvmx_lmcx_wodt_ctl0_cn30xx cn56xxp1; struct cvmx_lmcx_wodt_ctl0_cn38xx cn58xx; struct cvmx_lmcx_wodt_ctl0_cn38xx cn58xxp1; }; typedef union cvmx_lmcx_wodt_ctl0 cvmx_lmcx_wodt_ctl0_t; /** * cvmx_lmc#_wodt_ctl1 * * LMC_WODT_CTL1 = LMC Write OnDieTermination control * System designers may desire to terminate DQ/DQS/DM lines for higher frequency DDR operations * (667MHz and faster), especially on a multi-rank system. DDR2 DQ/DM/DQS I/O's have built in * Termination resistor that can be turned on or off by the controller, after meeting tAOND and tAOF * timing requirements. Each Rank has its own ODT pin that fans out to all the memory parts * in that DIMM. System designers may prefer different combinations of ODT ON's for read and write * into different ranks. Octeon supports full programmability by way of the mask register below. * Each Rank position has its own 8-bit programmable field. * When the controller does a write to that rank, it sets the 8 ODT pins to the MASK pins below. * For eg., When doing a write into Rank0, a system designer may desire to terminate the lines * with the resistor on Dimm0/Rank1. The mask WODT_D0_R0 would then be [00000010]. * If ODT feature is not desired, the DDR parts can be programmed to not look at these pins by * writing 0 in QS_DIC. Octeon drives the appropriate mask values on the ODT pins by default. * If this feature is not required, write 0 in this register. * * Notes: * Together, the LMC_WODT_CTL1 and LMC_WODT_CTL0 CSRs control the write ODT mask. * When a given RANK is selected, the WODT mask for that RANK is used. The resulting WODT mask is * driven to the DIMMs in the following manner: * BUNK_ENA=1 BUNK_ENA=0 * Mask[7] -> DIMM3, RANK1 DIMM3 * Mask[6] -> DIMM3, RANK0 * Mask[5] -> DIMM2, RANK1 DIMM2 * Mask[4] -> DIMM2, RANK0 * Mask[3] -> DIMM1, RANK1 DIMM1 * Mask[2] -> DIMM1, RANK0 * Mask[1] -> DIMM0, RANK1 DIMM0 * Mask[0] -> DIMM0, RANK0 */ union cvmx_lmcx_wodt_ctl1 { uint64_t u64; struct cvmx_lmcx_wodt_ctl1_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t reserved_32_63 : 32; uint64_t wodt_d3_r1 : 8; /**< Write ODT mask DIMM3, RANK1/DIMM3 in SingleRanked */ uint64_t wodt_d3_r0 : 8; /**< Write ODT mask DIMM3, RANK0 */ uint64_t wodt_d2_r1 : 8; /**< Write ODT mask DIMM2, RANK1/DIMM2 in SingleRanked */ uint64_t wodt_d2_r0 : 8; /**< Write ODT mask DIMM2, RANK0 */ #else uint64_t wodt_d2_r0 : 8; uint64_t wodt_d2_r1 : 8; uint64_t wodt_d3_r0 : 8; uint64_t wodt_d3_r1 : 8; uint64_t reserved_32_63 : 32; #endif } s; struct cvmx_lmcx_wodt_ctl1_s cn30xx; struct cvmx_lmcx_wodt_ctl1_s cn31xx; struct cvmx_lmcx_wodt_ctl1_s cn52xx; struct cvmx_lmcx_wodt_ctl1_s cn52xxp1; struct cvmx_lmcx_wodt_ctl1_s cn56xx; struct cvmx_lmcx_wodt_ctl1_s cn56xxp1; }; typedef union cvmx_lmcx_wodt_ctl1 cvmx_lmcx_wodt_ctl1_t; /** * cvmx_lmc#_wodt_mask * * LMC_WODT_MASK = LMC Write OnDieTermination mask * System designers may desire to terminate DQ/DQS lines for higher frequency DDR operations * especially on a multi-rank system. DDR3 DQ/DQS I/O's have built in * Termination resistor that can be turned on or off by the controller, after meeting tAOND and tAOF * timing requirements. Each Rank has its own ODT pin that fans out to all the memory parts * in that DIMM. System designers may prefer different combinations of ODT ON's for writes * into different ranks. Octeon supports full programmability by way of the mask register below. * Each Rank position has its own 8-bit programmable field. * When the controller does a write to that rank, it sets the 4 ODT pins to the MASK pins below. * For eg., When doing a write into Rank0, a system designer may desire to terminate the lines * with the resistor on DIMM0/Rank1. The mask WODT_D0_R0 would then be [00000010]. * Octeon drives the appropriate mask values on the ODT pins by default. If this feature is not * required, write 0 in this register. * * Notes: * When a given RANK is selected, the WODT mask for that RANK is used. The resulting WODT mask is * driven to the DIMMs in the following manner: * RANK_ENA=1 RANK_ENA=0 * Mask[3] -> DIMM1_ODT_1 MBZ * Mask[2] -> DIMM1_ODT_0 DIMM1_ODT_0 * Mask[1] -> DIMM0_ODT_1 MBZ * Mask[0] -> DIMM0_ODT_0 DIMM0_ODT_0 * * LMC always writes entire cache blocks and always writes them via two consecutive * write CAS operations to the same rank+bank+row spaced exactly 4 CK's apart. * When a WODT mask bit is set, LMC asserts the OCTEON ODT output * pin(s) starting the same CK as the first write CAS operation. Then, OCTEON * normally continues to assert the ODT output pin(s) for 9+LMC*_CONTROL[WODT_BPRCH] more CK's * - for a total of 10+LMC*_CONTROL[WODT_BPRCH] CK's for the entire cache block write - * through the second write CAS operation of the cache block, * satisfying the 6 CK DDR3 ODTH8 requirements. * But it is possible for OCTEON to issue two cache block writes separated by as few as * WtW = 8 or 9 (10 if LMC*_CONTROL[WODT_BPRCH]=1) CK's. In that case, OCTEON asserts the ODT output pin(s) * for the WODT mask of the first cache block write for WtW CK's, then asserts * the ODT output pin(s) for the WODT mask of the second cache block write for 10+LMC*_CONTROL[WODT_BPRCH] CK's * (or less if a third cache block write follows within 8 or 9 (or 10) CK's of this second cache block write). * Note that it may be necessary to force LMC to space back-to-back cache block writes * to different ranks apart by at least 10+LMC*_CONTROL[WODT_BPRCH] CK's to prevent DDR3 ODTH8 violations. */ union cvmx_lmcx_wodt_mask { uint64_t u64; struct cvmx_lmcx_wodt_mask_s { #if __BYTE_ORDER == __BIG_ENDIAN uint64_t wodt_d3_r1 : 8; /**< Write ODT mask DIMM3, RANK1/DIMM3 in SingleRanked *UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d3_r0 : 8; /**< Write ODT mask DIMM3, RANK0 *UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d2_r1 : 8; /**< Write ODT mask DIMM2, RANK1/DIMM2 in SingleRanked *UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d2_r0 : 8; /**< Write ODT mask DIMM2, RANK0 *UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d1_r1 : 8; /**< Write ODT mask DIMM1, RANK1/DIMM1 in SingleRanked if (!RANK_ENA) then WODT_D1_R1[3:0] MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d1_r0 : 8; /**< Write ODT mask DIMM1, RANK0 if (!RANK_ENA) then WODT_D1_R0[3,1] MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d0_r1 : 8; /**< Write ODT mask DIMM0, RANK1/DIMM0 in SingleRanked if (!RANK_ENA) then WODT_D0_R1[3:0] MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ uint64_t wodt_d0_r0 : 8; /**< Write ODT mask DIMM0, RANK0 if (!RANK_ENA) then WODT_D0_R0[3,1] MBZ *Upper 4 bits UNUSED IN 63xx, and MBZ* */ #else uint64_t wodt_d0_r0 : 8; uint64_t wodt_d0_r1 : 8; uint64_t wodt_d1_r0 : 8; uint64_t wodt_d1_r1 : 8; uint64_t wodt_d2_r0 : 8; uint64_t wodt_d2_r1 : 8; uint64_t wodt_d3_r0 : 8; uint64_t wodt_d3_r1 : 8; #endif } s; struct cvmx_lmcx_wodt_mask_s cn63xx; struct cvmx_lmcx_wodt_mask_s cn63xxp1; }; typedef union cvmx_lmcx_wodt_mask cvmx_lmcx_wodt_mask_t; #endif