diff options
Diffstat (limited to 'board-js2x/llfw')
-rw-r--r-- | board-js2x/llfw/Cboot.S | 18 | ||||
-rw-r--r-- | board-js2x/llfw/Makefile | 77 | ||||
-rw-r--r-- | board-js2x/llfw/board_io.S | 62 | ||||
-rw-r--r-- | board-js2x/llfw/hw.c | 124 | ||||
-rw-r--r-- | board-js2x/llfw/stage2.c | 271 | ||||
-rw-r--r-- | board-js2x/llfw/stage2.h | 23 | ||||
-rw-r--r-- | board-js2x/llfw/stage2.lds | 56 | ||||
-rw-r--r-- | board-js2x/llfw/stage2_head.S | 91 | ||||
-rw-r--r-- | board-js2x/llfw/stage_s.S | 43 | ||||
-rw-r--r-- | board-js2x/llfw/stage_s.lds | 34 | ||||
-rw-r--r-- | board-js2x/llfw/startup.S | 708 | ||||
-rw-r--r-- | board-js2x/llfw/u4mem.c | 4071 |
12 files changed, 5560 insertions, 18 deletions
diff --git a/board-js2x/llfw/Cboot.S b/board-js2x/llfw/Cboot.S new file mode 100644 index 0000000..3e37984 --- /dev/null +++ b/board-js2x/llfw/Cboot.S @@ -0,0 +1,18 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + + .org 0 + + /* Boot Information, hardcoded to ColdReset */ + .quad 1 + /* start address */ + .quad 0x100 diff --git a/board-js2x/llfw/Makefile b/board-js2x/llfw/Makefile new file mode 100644 index 0000000..fcb7eb4 --- /dev/null +++ b/board-js2x/llfw/Makefile @@ -0,0 +1,77 @@ +# ***************************************************************************** +# * Copyright (c) 2004, 2007 IBM Corporation +# * All rights reserved. +# * This program and the accompanying materials +# * are made available under the terms of the BSD License +# * which accompanies this distribution, and is available at +# * http://www.opensource.org/licenses/bsd-license.php +# * +# * Contributors: +# * IBM Corporation - initial implementation +# ****************************************************************************/ + +include ../../make.rules + +CPPFLAGS = -I$(INCLBRDDIR) -I$(INCLCMNDIR) -I$(INCLCMNDIR)/$(CPUARCH) \ + -I$(LIBCMNDIR)/libc/include +CFLAGS += -fno-builtin $(CPPFLAGS) -O2 -msoft-float $(MAMBO) $(PLATFORM) +CFLAGS += $(BOOT) $(IOCONF) -Wa,-mregnames $(RELEASE) $(CPUARCHDEF) -Wall +ASFLAGS = $(PLATFORM) $(BOOT) $(IOCONF) $(RELEASE)$(CPUARCHDEF) -Wa,-mregnames +LDFLAGS1 = -nostdlib -e__start -Tstage2.lds -N -Ttext=0x100 + + +STG1OBJ = startup.o boot_abort.o romfs.o hw.o io_generic.o board_io.o +#STG1OBJ += monitor.o ipmi_kcs.o mmu.o +STG1OBJ += stage2_head.o stage2.o comlib.o \ + romfs_wrap.o nvramlog.o + +all : stage1.js2x stage1.bimini stageS.bin + +stage1.js2x: $(STG1OBJ) Cboot.o u4maui.o $(LIBCMNDIR)/libc.a + $(LD) $(LDFLAGS1) -o stage1.elf $(STG1OBJ) u4maui.o $(LIBCMNDIR)/libc.a + $(OBJCOPY) -O binary stage1.elf stage1.js2x + +stage1.bimini: $(STG1OBJ) Cboot.o u4bimini.o $(LIBCMNDIR)/libc.a + $(LD) $(LDFLAGS1) -o stage1.elf $(STG1OBJ) u4bimini.o $(LIBCMNDIR)/libc.a + $(OBJCOPY) -O binary stage1.elf stage1.bimini + +stageS.bin: stage_s.o + $(LD) -nostdlib -N -Tstage_s.lds -o stage_s.elf stage_s.o + $(OBJCOPY) -O binary stage_s.elf stageS.bin + +u4bimini.o: u4mem.c + $(CC) $(CFLAGS) -DBOARD_BIMINI -o u4bimini.o -c u4mem.c + +u4maui.o: u4mem.c + $(CC) $(CFLAGS) -DBOARD_MAUI -o u4maui.o -c u4mem.c + +romfs.o: ../../llfw/romfs.S + $(CC) $(CFLAGS) -c ../../llfw/romfs.S + +ipmi_kcs.o: ../rtas/ipmi_kcs.c + $(CC) $(CFLAGS) -c ../rtas/ipmi_kcs.c + +boot_abort.o: ../../llfw/boot_abort.S + $(CC) $(CFLAGS) -c ../../llfw/boot_abort.S + +nvramlog.o: ../../llfw/nvramlog.S + $(CC) $(CFLAGS) -c ../../llfw/nvramlog.S + +include $(LLFWCMNDIR)/clib/Makefile.inc + +include $(LLFWCMNDIR)/io_generic/Makefile.inc + +romfs_wrap.o: ../../llfw/romfs_wrap.c + $(CC) $(CFLAGS) -c ../../llfw/romfs_wrap.c +Cboot.o: Cboot.S + $(CC) $(CFLAGS) -c $^ + $(OBJCOPY) -O binary Cboot.o Cboot.bin + +$(LIBCMNDIR)/libc.a: + $(MAKE) -C $(LIBCMNDIR) libc + +%.o: %.S + $(CC) $(CFLAGS) -c $^ + +clean: + rm -f *.o *.bin *.elf *.js2x *.bimini diff --git a/board-js2x/llfw/board_io.S b/board-js2x/llfw/board_io.S new file mode 100644 index 0000000..6e39dfc --- /dev/null +++ b/board-js2x/llfw/board_io.S @@ -0,0 +1,62 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +#include <macros.h> +#include <cpu.h> + + .text + +C_ENTRY(copy_from_flash) +# size in GPR3 (multiple of 64), from GPR4, to GPR5 + mflr 24 ; mtctr 3 ; addi 4,4,-64 ; addi 5,5,-64 +0: SETCI(r0) + ldu 16,64(4) ; ld 17,8(4) ; ld 18,16(4) ; ld 19,24(4) + ld 20,32(4) ; ld 21,40(4) ; ld 22,48(4) ; ld 23,56(4) + CLRCI(r0) + stdu 16,64(5) ; std 17,8(5) ; std 18,16(5) ; std 19,24(5) + std 20,32(5) ; std 21,40(5) ; std 22,48(5) ; std 23,56(5) + sync ; icbi 0,2 ; bdnz 0b ; sync ; isync ; mtlr 24 ; blr + +/**************************************************************************** + * prints one character to serial console + * + * Input: + * R3 - character + * + * Returns: - + * + * Modifies Registers: + * R3, R4, R5, R6, R7 + ****************************************************************************/ +ENTRY(io_putchar) + mflr r7 + + SETCI(r0) + + # always use serial1 + li 4,0x3f8 ; oris 4,4,0xf400 + + # print one char +0: lbz 0,5(4) ; andi. 0,0,0x20 ; beq 0b ; stb 3,0(4) ; eieio + + # also print char to serial2 if on a JS21 + # read ID register: only if it is a PC87427 (JS21) also use serial2 + addi 4,4,-0x3f8 + li 5,0x20 ; stb 5,0x2e(4) ; lbz 5,0x2f(4) ; cmpdi 5,0xf2 ; bne 1f + + addi 4,4,0x2f8 +0: lbz 0,5(4) ; andi. 0,0,0x20 ; beq 0b ; stb 3,0(4) ; eieio + +1: CLRCI(r0) + + mtlr r7 + blr diff --git a/board-js2x/llfw/hw.c b/board-js2x/llfw/hw.c new file mode 100644 index 0000000..95888dc --- /dev/null +++ b/board-js2x/llfw/hw.c @@ -0,0 +1,124 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +#include <cpu.h> +#include <types.h> +#include <hw.h> + +uint16_t +bswap16_load(uint64_t addr) +{ + unsigned int val; + set_ci(); + asm volatile ("lhbrx %0, 0, %1":"=r" (val):"r"(addr)); + clr_ci(); + return val; +} + +uint32_t +bswap32_load(uint64_t addr) +{ + unsigned int val; + set_ci(); + asm volatile ("lwbrx %0, 0, %1":"=r" (val):"r"(addr)); + clr_ci(); + return val; +} + +void +bswap16_store(uint64_t addr, uint16_t val) +{ + set_ci(); + asm volatile ("sthbrx %0, 0, %1"::"r" (val), "r"(addr)); + clr_ci(); +} + +void +bswap32_store(uint64_t addr, uint32_t val) +{ + set_ci(); + asm volatile ("stwbrx %0, 0, %1"::"r" (val), "r"(addr)); + clr_ci(); +} + +uint8_t +load8_ci(uint64_t addr) +{ + uint8_t val; + set_ci(); + val = *(uint8_t *) addr; + clr_ci(); + return val; +} + +uint16_t +load16_ci(uint64_t addr) +{ + uint16_t val; + set_ci(); + val = *(uint16_t *) addr; + clr_ci(); + return val; +} + +uint32_t +load32_ci(uint64_t addr) +{ + uint32_t val; + set_ci(); + val = *(uint32_t *) addr; + clr_ci(); + return val; +} + +uint64_t +load64_ci(uint64_t addr) +{ + uint64_t val; + set_ci(); + val = *(uint64_t *) addr; + clr_ci(); + return val; +} + + +void +store8_ci(uint64_t addr, uint8_t val) +{ + set_ci(); + *(uint8_t *) addr = val; + clr_ci(); +} + +void +store16_ci(uint64_t addr, uint16_t val) +{ + set_ci(); + *(uint16_t *) addr = val; + clr_ci(); +} + +void +store32_ci(uint64_t addr, uint32_t val) +{ + set_ci(); + *(uint32_t *) addr = val; + clr_ci(); +} + +void +store64_ci(uint64_t addr, uint64_t val) +{ + set_ci(); + *(uint64_t *) addr = val; + clr_ci(); +} diff --git a/board-js2x/llfw/stage2.c b/board-js2x/llfw/stage2.c new file mode 100644 index 0000000..b529230 --- /dev/null +++ b/board-js2x/llfw/stage2.c @@ -0,0 +1,271 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +#include <types.h> +#include <xvect.h> +#include <hw.h> +#include <stdio.h> +#include <romfs.h> +#include "memmap.h" +#include "stage2.h" +#include <termctrl.h> +#include "product.h" +#include "calculatecrc.h" +#include <cpu.h> +#include <string.h> + +uint64_t uart; +uint64_t gVecNum; +uint8_t u4Flag; + +uint64_t exception_stack_frame; + +typedef void (*pInterruptFunc_t) (void); + +pInterruptFunc_t vectorTable[0x2E << 1]; + +void c_memInit(uint64_t r3, uint64_t r4); + +void proceedInterrupt(); + +void +exception_forward(void) +{ + uint64_t val; + + if (*(uint64_t *) XVECT_M_HANDLER) { + proceedInterrupt(); + } + + printf("\r\n exception %llx ", gVecNum); + asm volatile ("mfsrr0 %0":"=r" (val):); + printf("\r\nSRR0 = %08llx%08llx ", val >> 32, val); + asm volatile ("mfsrr1 %0":"=r" (val):); + printf(" SRR1 = %08llx%08llx ", val >> 32, val); + + asm volatile ("mfsprg %0,2":"=r" (val):); + printf("\r\nSPRG2 = %08llx%08llx ", val >> 32, val); + asm volatile ("mfsprg %0,3":"=r" (val):); + printf(" SPRG3 = %08llx%08llx \r\n", val >> 32, val); + while (1); +} + +void +c_interrupt(uint64_t vecNum) +{ + gVecNum = vecNum; + if (vectorTable[vecNum >> 7]) { + vectorTable[vecNum >> 7] (); + } else { + exception_forward(); + } +} + +void +set_exceptionVector(int num, void *func) +{ + vectorTable[num >> 7] = (pInterruptFunc_t) func; +} + +static void +io_init(void) +{ + // read ID register: only if it is a PC87427, enable serial2 + store8_ci(0xf400002e, 0x20); + if (load8_ci(0xf400002f) != 0xf2) { + uart = 0xf40003f8; + u4Flag = 0; + } else { + uart = 0xf40002f8; + u4Flag = 1; + } +} + +int +io_getchar(char *ch) +{ + int retVal = 0; + if ((load8_ci(uart + 5) & 0x01)) { + *ch = load8_ci(uart); + retVal = 1; + } + return retVal; +} + +uint64_t +get_dec(void) +{ + return 0xdeadaffe; +} + +void +set_dec(uint64_t val) +{ +} + +uint64_t +tb_frequency(void) +{ + return 0; +} + +void copy_from_flash(uint64_t cnt, uint64_t src, uint64_t dest); + +const uint32_t CrcTableHigh[16] = { + 0x00000000, 0x4C11DB70, 0x9823B6E0, 0xD4326D90, + 0x34867077, 0x7897AB07, 0xACA5C697, 0xE0B41DE7, + 0x690CE0EE, 0x251D3B9E, 0xF12F560E, 0xBD3E8D7E, + 0x5D8A9099, 0x119B4BE9, 0xC5A92679, 0x89B8FD09 +}; +const uint32_t CrcTableLow[16] = { + 0x00000000, 0x04C11DB7, 0x09823B6E, 0x0D4326D9, + 0x130476DC, 0x17C56B6B, 0x1A864DB2, 0x1E475005, + 0x2608EDB8, 0x22C9F00F, 0x2F8AD6D6, 0x2B4BCB61, + 0x350C9B64, 0x31CD86D3, 0x3C8EA00A, 0x384FBDBD +}; + +unsigned long +check_flash_image(unsigned long rombase, unsigned long length, + unsigned long start_crc) +{ + + uint32_t AccumCRC = start_crc; + char val; + uint32_t Temp; + while (length-- > 0) { + val = load8_ci(rombase++); + Temp = ((AccumCRC >> 24) ^ val) & 0x000000ff; + AccumCRC <<= 8; + AccumCRC ^= CrcTableHigh[Temp / 16]; + AccumCRC ^= CrcTableLow[Temp % 16]; + } + + return AccumCRC; +} + +static void +load_file(uint64_t destAddr, char *name, uint64_t maxSize, uint64_t romfs_base) +{ + uint64_t *src, *dest, cnt; + struct romfs_lookup_t fileInfo; + c_romfs_lookup(name, romfs_base, &fileInfo); + if (maxSize) { + cnt = maxSize / 8; + } else { + cnt = (fileInfo.size_data + 7) / 8; + } + dest = (uint64_t *) destAddr; + src = (uint64_t *) fileInfo.addr_data; + while (cnt--) { + store64_ci((uint64_t) dest, *src); + dest++; + src++; + } + flush_cache((void *) destAddr, fileInfo.size_data); +} + +/*************************************************************************** + * Function: early_c_entry + * Input : start_addr + * + * Description: + **************************************************************************/ +void +early_c_entry(uint64_t start_addr) +{ + struct romfs_lookup_t fileInfo; + uint32_t crc; + void (*pofwStart) (uint64_t rombase); + uint64_t *boot_info; + uint64_t ofw_start = 0x80000; + exception_stack_frame = 0; + /* destination for the flash image; we copy it to RAM + * because from flash it is much too slow + * the flash is copied at 224MB - 4MB (max flash size) + * at 224MB starts SLOF + * at 256MB is the SLOF load-base */ + uint64_t romfs_base = 0xe000000 - 0x400000; + // romfs header values + struct stH *header = (struct stH *) (start_addr + 0x28); + //since we cannot read the fh_magic directly from flash as a string, we need to copy it to memory + uint64_t magic_val = 0; + uint64_t startVal = 0; + uint64_t flashlen = 0; + + io_init(); + + flashlen = load64_ci((uint64_t) (&header->flashlen)); + + //copy fh_magic to magic_val since, we cannot use it as a string from flash + magic_val = load64_ci((uint64_t) (header->magic)); + + printf(" Check ROM = "); + if (strncmp((char *)&magic_val, FLASHFS_MAGIC, 8) == 0) { + // somehow, the first 8 bytes in flashfs are overwritten, if booting from drone... + // so if we find "IMG1" in the first 4 bytes, we skip the CRC check... + startVal = load64_ci((uint64_t) start_addr); + if (strncmp((char *)&startVal, "IMG1", 4) == 0) { + printf + ("start from RAM detected, skipping CRC check!\r\n"); + // for romfs accesses (c_romfs_lookup) to work, we must fix the first uint64_t to the value we expect... + store64_ci((uint64_t) start_addr, 0xd8); + } else { + //checking CRC in flash, we must use cache_inhibit + // since the crc is included as the last 32 bits in the image, the resulting crc should be 0 + crc = + check_flash_image((uint64_t) start_addr, + load64_ci((uint64_t) + (&header->flashlen)), + 0); + if (crc == 0) { + printf("OK\r\n"); + } else { + printf("failed!\r\n"); + while (1); + } + } + } else { + printf + ("failed (magic string is \"%.8s\" should be \"%.8s\")\r\n", + (char *)&magic_val, FLASHFS_MAGIC); + while (1); + } + + printf(" Press \"s\" to enter Open Firmware.\r\n\r\n"); + + if ((start_addr > 0xF0000000) && u4Flag) + u4memInit(); + + /* here we have real ram avail -> hopefully + * copy flash to ram; size is in 64 byte blocks */ + flashlen /= 64; + /* align it a bit */ + flashlen += 7; + flashlen &= ~7; + copy_from_flash(flashlen, start_addr, romfs_base); + /* takeover sometimes fails if the image running on the system + * has a different size; flushing the cache helps, because it is + * the right thing to do anyway */ + flush_cache((void *) romfs_base, flashlen * 64); + + c_romfs_lookup("bootinfo", romfs_base, &fileInfo); + boot_info = (uint64_t *) fileInfo.addr_data; + boot_info[1] = start_addr; + load_file(0x100, "xvect", 0x100, romfs_base); + load_file(SLAVELOOP_LOADBASE, "stageS", 0, romfs_base); + load_file(ofw_start, "ofw_main", 0, romfs_base); + pofwStart = (void (*)(uint64_t)) &ofw_start; + // re-enable the cursor + printf("%s%s", TERM_CTRL_RESET, TERM_CTRL_CRSON); + // never return + pofwStart(romfs_base); +} diff --git a/board-js2x/llfw/stage2.h b/board-js2x/llfw/stage2.h new file mode 100644 index 0000000..1c5205a --- /dev/null +++ b/board-js2x/llfw/stage2.h @@ -0,0 +1,23 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +#ifndef __STAGE2_H +#define __STAGE2_H + +#ifndef __ASSEMBLER__ + +#include <stddef.h> + +void u4memInit(void); + +#endif /* __ASSEMBLER__ */ +#endif /* __STAGE2_H */ diff --git a/board-js2x/llfw/stage2.lds b/board-js2x/llfw/stage2.lds new file mode 100644 index 0000000..703ef9d --- /dev/null +++ b/board-js2x/llfw/stage2.lds @@ -0,0 +1,56 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +OUTPUT_FORMAT("elf64-powerpc", "elf64-powerpc", "elf64-powerpc") +OUTPUT_ARCH(powerpc:common64) + +/* set the entry point */ +ENTRY ( __start ) + +SECTIONS { + .text : { + *(.text) + } + + . = ALIGN(8); + + .data : { + *(.data) + *(.rodata .rodata.*) + *(.got1) + *(.sdata) + *(.opd) + } + + /* FIXME bss at end ??? */ + + . = ALIGN(8); + __bss_start = .; + .bss : { + *(.sbss) *(.scommon) + *(.dynbss) + *(.bss) + } + + . = ALIGN(8); + __bss_end = .; + __bss_size = (__bss_end - __bss_start); + + __toc_start = .; + .got : + { + *(.toc .got) + } + . = ALIGN(8); + __toc_end = .; + __stack_end = . ; +} diff --git a/board-js2x/llfw/stage2_head.S b/board-js2x/llfw/stage2_head.S new file mode 100644 index 0000000..ea1cf59 --- /dev/null +++ b/board-js2x/llfw/stage2_head.S @@ -0,0 +1,91 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +#include "macros.h" +#include "../../llfw/boot_abort.h" + +/*#################### defines #####################*/ +#define STACK_SIZE 0x4000 + +/*#################### code ########################*/ + .text + .globl .gluon_c_entry + .globl __toc_start + .globl __toc_end + .globl __stack_end + .globl __bss_start + .globl __bss_size + .globl __start + +ASM_ENTRY(__startC) + /* clear out bss section */ + LOAD64(r3, (__bss_start - 8)) + LOAD64(r4, __bss_size) + + /* divide __bss_size by 8 to get number */ + /* of dwords to clear */ + srwi. r4, r4, 3 + beq bsscdone + li r5, 0 + mtctr r4 +bssc: stdu r5, 8(r3) + bdnz bssc +bsscdone: + /* setup stack */ + LOAD64(r1, __stack_end + STACK_SIZE) + + /* save return address beside stack */ + addi r3, r1, 128 + mflr r0 + std r0, 0(r3) + + /* setup toc */ + bl toc_init + + /* ------------------------------------ */ + /* jump to c-code */ + /* r10 = cpu_init_slave address - r3 */ + /* r11 = slave_setup address - r4 */ + /* ------------------------------------ */ + mr r3, r10 + mr r4, r11 + bl .early_c_entry + + /* return to caller... */ + LOAD64(r1, __stack_end + STACK_SIZE) + addi r1, r1, 128 + ld r3, 0(r1) + mtlr r3 + blr + + /* #################################### */ + /* Basic Additional Functions */ + /* for extended lib functions see */ + /* external library */ + /* #################################### */ + .align 2 + + /* ------------------------------------ */ + /* updates toc in r2 */ + /* ------------------------------------ */ +ASM_ENTRY(toc_init) + LOAD64(r2, __toc_start) + addi r2,r2,0x4000 + addi r2,r2,0x4000 + blr + + /* ------------------------------------ */ + /* stores arg#1 in r27 and stops */ + /* ------------------------------------ */ +ENTRY(do_panic) +ENTRY(halt_sys) + BOOT_ABORT_R3HINT(ABORT_CANIO, ALTBOOT, msg_e_ierror); diff --git a/board-js2x/llfw/stage_s.S b/board-js2x/llfw/stage_s.S new file mode 100644 index 0000000..15eed93 --- /dev/null +++ b/board-js2x/llfw/stage_s.S @@ -0,0 +1,43 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + .text + .org 0 /* after loading to mem this is at slaveloop_base */ +# . = 0x3f00 loaded to this addr + .globl __stage_s_entry + +__stage_s_entry: + bl $+4 + mflr r4 + li r3, 0xff + not r3, r3 + and r4, r4, r3 + li 0,1 ; std 0,0xf8(4) + li 0,0 ; std 0,0xa0(4) +slaveloop: + ori 28,28,0x100 + lis 0,10 + mtctr 0 + bdnz $ # do some waiting, to prevent flooding the buses + lwz 0,0xa0(4) + cmpw 0,28 + bne $-20 # wait for our flag + + lwz 0,0x80(4) + lwz 3,0xc0(4) + mtctr 0 + bctr # jump to specified address, with specified GPR3 + +# .quad 0 +# .quad 0 + + + diff --git a/board-js2x/llfw/stage_s.lds b/board-js2x/llfw/stage_s.lds index d77a17e..989ca3d 100644 --- a/board-js2x/llfw/stage_s.lds +++ b/board-js2x/llfw/stage_s.lds @@ -1,24 +1,22 @@ -/* ============================================================================= */ -/* * Copyright (c) 2004, 2005 IBM Corporation -/* * All rights reserved. -/* * This program and the accompanying materials -/* * are made available under the terms of the BSD License -/* * which accompanies this distribution, and is available at -/* * http://www.opensource.org/licenses/bsd-license.php -/* * -/* * Contributors: -/* * IBM Corporation - initial implementation -/* ============================================================================= */ - +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ OUTPUT_FORMAT("elf64-powerpc", "elf64-powerpc", "elf64-powerpc") OUTPUT_ARCH(powerpc:common64) +ENTRY( __stage_s_entry ) + SECTIONS { - .startup 0x2000000 : { startup.o } - image = ALIGN(128); - .slof.loader image : { loader.o } - .slof.vectors image + 0x100 : { *(.slof.vectors) } - .slof.text image + 0x4100 : { *(.slof.text) } - .slof.data image + 0x6100 : { *(.slof.data) } + .text : { + *(.text) + } } diff --git a/board-js2x/llfw/startup.S b/board-js2x/llfw/startup.S new file mode 100644 index 0000000..59e1c3f --- /dev/null +++ b/board-js2x/llfw/startup.S @@ -0,0 +1,708 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ + +# SLOF for JS20/JS21 -- ROM boot code. +# Initial entry point, copy code from flash to cache, memory setup. +# Also sets up serial console and optimizes some settings. + +#include "termctrl.h" +#include <product.h> +#include <xvect.h> +#include <cpu.h> +#include <macros.h> +#include <southbridge.h> + + .text + .globl __start +__start: + /* put rombase in sprg1 ***********************/ + + bl postHeader + .long 0xDEADBEE0 + .long 0x0 /* size */ + .long 0x0 /* crc */ + .long relTag - __start +postHeader: + mflr r3 + li r4, 0x7fff + not r4, r4 + and r3, r3, r4 + mtsprg 1, r3 /* romfs base */ + bl _start + + .org 0x150 - 0x100 +__startSlave: + bl setup_cpu + bl set_ci_bit +# b slaveWithNumber + b slave + + .org 0x180 - 0x100 +__startMaster: + li 3,0 + mtsprg 1, r3 /* romfs base */ + bl setup_cpu + bl set_ci_bit + b master + + + /* FIXME: Also need 0280, 0380, 0f20, etc. */ + + .irp i, 0x0100,0x0180,0x0200,0x0280,0x0300,0x0380,0x0400,0x0500,0x0600,0x0700, \ + 0x0800,0x0900,0x0a00,0x0b00,0x0c00,0x0d00,0x0e00,0x0f00, \ + 0x1000,0x1100,0x1200,0x1300,0x1400,0x1500,0x1600,0x1700, \ + 0x1800,0x1900,0x1a00,0x1b00,0x1c00,0x1d00,0x1e00,0x1f00, \ + 0x2000,0x2100,0x2200,0x2300,0x2400,0x2500,0x2600,0x2700, \ + 0x2800,0x2900,0x2a00,0x2b00,0x2c00,0x2d00,0x2e00 + . = \i + + /* enable this if you get exceptions before the console works */ + /* this will allow using the hardware debugger to see where */ + /* it traps, and with what register values etc. */ + // b $ + + mtsprg 0, r0 + mfctr r0 + mtsprg 2,r0 + mflr r0 +// 10 + mtsprg 3,r0 + ld r0, (\i + 0x160)(0) + mtctr r0 + li r0, \i + 0x100 +// 20 + bctr + + . = \i + 0x60 + + .quad intHandler2C + + .endr + + + . = XVECT_M_HANDLER - 0x100 + .quad 0x00 + . = XVECT_S_HANDLER - 0x100 + + .quad 0 + + + + .org 0x4000 - 0x100 +_start: + # optimize HID register settings + bl setup_cpu + bl set_ci_bit + + # read semaphore, run as slave if not the first to do so + li 3,0 ; oris 3,3,0xf800 ; lwz 3,0x60(3) ; andi. 3,3,1 ; beq slave +master: + # setup flash, serial + bl setup_sio + + # early greet + li r3, 10 + bl putc + li 3,13 ; bl putc ; li 3,10 ; bl putc ; li 3,'S' ; bl putc + + + #do we run from ram ? + mfsprg r3, 1 /* rombase */ + cmpdi r3,0 /* rombase is 0 when running from RAM */ + + bne copy_to_cache + + # wait a bit, start scripts are slow... need to get all cores running! + lis 3,0x4000 ; mtctr 3 ; bdnz $ + + # copy 4MB from 0 to temp memory + lis 4,0x8 ; mtctr 4 ; lis 4,0x200 ; li 3,0 ; addi 4,4,-8 ; addi 3,3,-8 +0: ldu 5,8(3) ; stdu 5,8(4) ; bdnz 0b + + lis 4,0x200 + mtsprg 1, r4 + + lis 4,0x1 + lis 3,0x20 ; addi 3,3,0x200-8 ; + FLUSH_CACHE(r3, r4) + + lis 4,0x200 + addi 4,4,copy_to_cache@l + mtctr 4 + bctr + +# make all data accesses cache-inhibited +set_ci_bit: + SETCI(r0) + blr + +# make all data accesses cacheable +clr_ci_bit: + CLRCI(r0) + blr + +# write a character to the serial port +putc: +# always write to serial1 +0: lbz 0,5(13) ; andi. 0,0,0x20 ; beq 0b ; stb 3,0(13) ; eieio + +# read ID register: only if it is a PC87427 (JS21) also use serial2 + li 4,0 ; oris 4,4,0xf400 + li 5,0x20 ; stb 5,0x2e(4) ; lbz 5,0x2f(4); cmpdi 5,0xf2 ; bne 1f + + addi 4,4,0x2f8 +0: lbz 0,5(4) ; andi. 0,0,0x20 ; beq 0b ; stb 3,0(4) ; eieio + +1: blr + +# transfer from running from flash to running from cache +return_cacheable: + # find and set address to start running from cache, set msr value + mflr 3 ; rldicl 3,3,0,44 +jump_cacheable: + mtsrr0 3 ; + mfmsr 3 ; ori 3,3,0x1000 ; mtsrr1 3 # enable MCE, as well + + # set cacheable insn fetches, jump to cache + mfspr 3,HID1 ; rldicl 3,3,32,0 ; oris 3,3,0x0020 ; rldicl 3,3,32,0 + sync ; mtspr HID1,3 ; mtspr HID1,3 ; rfid ; b . + + + + +copy_to_cache: + # zero the whole cache + # also, invalidate the insn cache, to clear parity errors + # 128kB @ 0MB (boot code and vectors from 0x0 up to 0x20000) + li 4,0x400 ; mtctr 4 ; li 5,0x0 ; bl clr_ci_bit +0: dcbz 0,5 ; sync ; icbi 0,5 ; sync ; isync ; addi 5,5,0x80 ; bdnz 0b + + # 0x2000 to 0x100000/0x80000 (smaller on 970/970FX) + li 4,0x1C00 ; mfpvr 0 ; srdi 0,0,16 ; cmpdi 0,0x0044 ; bge 0f ; li 4,0xC00 +0: + mtctr 4 ; li 5,0x2000 +0: dcbz 0,5 ; sync ; isync ; addi 5,5,0x80 ; bdnz 0b ; bl set_ci_bit + + # find base address + bcl 20,31,$+4 ; mflr 31 ; rldicr 31,31,0,43 + + # copy 1kB from 0x4000 + li 4,0x80 ; mtctr 4 ; + li 5,0x3ff8 + addi 3,31,0x3ff8 +0: ldu 4,8(3) ; bl clr_ci_bit ; stdu 4,8(5) ; bl set_ci_bit ; bdnz 0b + # now start executing from cache -- insn cache is huge speed boost + + bl return_cacheable + + li 3,'L' ; bl putc + + # copy 128kB of flash to cache + li 4,0x800 ; mtctr 4 ; li 5,0x200-64 ; addi 3,31,0x200-64 ; +0: ldu 16,64(3) ; ld 17,8(3) ; ld 18,16(3) ; ld 19,24(3) + ld 20,32(3) ; ld 21,40(3) ; ld 22,48(3) ; ld 23,56(3) + bl clr_ci_bit + stdu 16,64(5) ; std 17,8(5) ; std 18,16(5) ; std 19,24(5) + std 20,32(5) ; std 21,40(5) ; std 22,48(5) ; std 23,56(5) + icbi 0,5 ; bl set_ci_bit ; bdnz 0b ; isync + + + li 3,'O' ; bl putc + + lis 4,0x20 + mfsprg r3,1 + cmpd r3,r4 + beq 1f + + // at 0xf8000000 we decide if it is u3 or u4 + li 4,0 ; oris 4,4,0xf800 ; lwz 3,0(4) ; srdi 3,3,4 ; cmpdi 3,3 ; bne 0f + bl setup_mem_u3 + bl setup_mem_size + b 1f +0: + +1: + li 3,'F' ; bl putc + + # setup nvram logging only when not running from RAM + mfsprg r3, 1 /* rombase */ + cmpdi r3, 0 /* rombase is 0 when running from RAM */ + beq 0f + + // at 0xf8000000 we decide if it is u3 or u4 + li r4, 0 + oris r4, r4, 0xf800 + lwz r3, 0(r4) + srdi r3, r3, 4 + cmpdi r3, 3 /* 3 means js20; no nvram logging on js20 */ + beq 0f + + bl io_log_init +0: + + #bl print_mem + + # data is cacheable by default from now on + bl clr_ci_bit + + + /* give live sign *****************************/ + bl 0f + .ascii TERM_CTRL_RESET + .ascii TERM_CTRL_CRSOFF + .ascii " **********************************************************************" + .ascii "\r\n" + .ascii TERM_CTRL_BRIGHT + .ascii PRODUCT_NAME + .ascii " Starting\r\n" + .ascii TERM_CTRL_RESET + .ascii " Build Date = ", __DATE__, " ", __TIME__ + .ascii "\r\n" + .ascii " FW Version = " , RELEASE + .ascii "\r\n\0" + .align 2 +0: mflr r3 + bl io_print + + # go! + li r3,__startC@l + mtctr r3 + mfsprg r10, 1 + bctrl + +relTag: + .ascii RELEASE + .ascii "\0" + .align 2 + +slave: + + # get cpu number + li 3,0 ; oris 3,3,0xf800 ; lwz 28,0x50(3) + +slaveWithNumber: + # create our slave loop address + sldi 3,28,24 ; oris 3,3,0x3000 + + # invalidate the insn cache, to clear parity errors + # clear the L2 cache as well, to get ECC right + li 4,0x2000 ; mfpvr 0 ; srdi 0,0,16 ; cmpdi 0,0x0044 ; bge 0f ; li 4,0x1000 +0: mtctr 4 ; mr 5,3 ; bl clr_ci_bit + +0: dcbz 0,5 ; sync ; icbi 0,5 ; sync ; isync ; addi 5,5,0x80 ; bdnz 0b + + + # write a "b $" insn in there + lis 4,0x4800 ; stw 4,0(3) +/* + mr 5,3 + + # jump there + bl set_ci_bit + li 13,0 ; oris 13,13,0xf400 + # device address + addi 13,13,0x2f8 + li 3,'O' ; add 3,3,28 ; bl putc + bl clr_ci_bit + mr 3,5 +*/ + b jump_cacheable + + + + +# allow the flash chip to be accessed faster +# initialize the 16550-compatible uart on serial port 1 of the sio +setup_sio: + + # i/o base address + li 3,0 ; oris 3,3,0xf400 + + # i/o base address + li 3,0 ; oris 3,3,0xf400 + + # put x-bus in turbo mode + li 4,0xf1 ; stb 4,0x400(3) ; eieio + + + # select sio serial1 + li 4,7 ; stb 4,0x2e(3) ; eieio ; li 4,3 ; stb 4,0x2f(3) ; eieio + + # set base address to 3f8 + li 4,0x60 ; stb 4,0x2e(3) ; eieio ; li 4,3 ; stb 4,0x2f(3) ; eieio + + # enable device + li 4,0x30 ; stb 4,0x2e(3) ; eieio ; li 4,1 ; stb 4,0x2f(3) ; eieio + + # read ID register: only if it is a PC87427, enable serial2 + li 4,0x20 ; stb 4,0x2e(3) ; eieio ; lbz 4,0x2f(3) ; cmpdi 4,0xf2 ; bne 0f + + # select sio serial2 + li 4,7 ; stb 4,0x2e(3) ; eieio ; li 4,2 ; stb 4,0x2f(3) ; eieio + + # set base address to 2f8 + li 4,0x60 ; stb 4,0x2e(3) ; eieio ; li 4,2 ; stb 4,0x2f(3) ; eieio + + # enable device + li 4,0x30 ; stb 4,0x2e(3) ; eieio ; li 4,1 ; stb 4,0x2f(3) ; eieio + + # uart @0x2f8 + addi 3,3,0x2f8 + + # disable interrupts, fifo off + li 4,0 ; stb 4,1(3) ; eieio ; stb 4,2(3) ; eieio + + # set serial speed + li 4,0x80 ; stb 4,3(3) ; eieio + li 4,115200/19200 ; stb 4,0(3) ; eieio ; li 4,0 ; stb 4,1(3) ; eieio + + # set 8-N-1, set RTS and DTR + li 4,3 ; stb 4,3(3) ; eieio ; stb 4,4(3) ; eieio + + eieio + + addi 3,3,-0x2f8 + + # uart @0x3f8 +0: addi 3,3,0x3f8 + + # disable interrupts, fifo off + li 4,0 ; stb 4,1(3) ; eieio ; stb 4,2(3) ; eieio + + # set serial speed + li 4,0x80 ; stb 4,3(3) ; eieio + li 4,115200/19200 ; stb 4,0(3) ; eieio ; li 4,0 ; stb 4,1(3) ; eieio + + # set 8-N-1, set RTS and DTR + li 4,3 ; stb 4,3(3) ; eieio ; stb 4,4(3) ; eieio + + eieio + + # save UART base for putc routine +0: mr 13,3 + + blr + + + + +# set the HID registers of the 970 for optimally executing from flash +setup_cpu: + + /* clear all the HV cruft */ + li r0, 0 + sync + mtspr HID4, r0 + isync + + /* enable dpm, disable attn insn, enable external mce + * first, try external time base; if clock doesn't run, switch to + * internal */ + li r0, 1 /* do the setup for external timebase */ + rldicl r0, r0, 44, 0 /* bit 19 has to be set */ + oris r0, r0, 0x8000 /* Enable external machine check */ + /* interrupts (preferred state */ + /* equals `1'). */ + sync + mtspr HID0, r0 + isync + + mftb r3 /* read the timebase */ + li r1, 0x4000 /* wait long enough for the external */ + mtctr r1 /* timebase (14MHz) to tick a bit */ + bdnz $ /* 0x4000 seems to be enough (for now) */ + mftb r4 /* read the timebase a second time */ + cmpld r3, r4 /* see if it changed */ + bne 0f + /* timebase did not change, do the setup for internal */ + rldicl r0, r0, 19, 1 + rldicl r0, r0, 45, 0 + sync + mtspr HID0, r0 + isync + +0: + /* enable insn prefetch, speculative table walks */ + mfspr r0, HID1 + rldicl r0, r0, 20, 0 + ori r0, r0, 0x1002 + mfsprg r3, 1 /* read rombase */ + cmpdi r3, 0 /* check if running from ram */ + bne 0f + /* running from ram */ + /* Enable instruction fetch cacheability control */ + ori r0, r0, 0x200 +0: + rldicl r0, r0, 44, 0 + sync + mtspr HID1, r0 + isync + + /* enable cache parity */ + mfspr r0, HID4 + oris r0, r0, 0xfff0 + xoris r0, r0, 0xfff0 + sync + mtspr HID4, r0 + isync + + /* exception offset at 0 */ + li r3, 0 + mtspr HIOR, r3 + + blr + +C_ENTRY(proceedInterrupt) + + ld r3,exception_stack_frame@got(r2) + ld r1,0(r3) + + .irp i, 2,3,4,5,6,7,8,9,10,11,12,13,14,15,16, \ + 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, \ + 27, 28, 29, 30, 31 + ld r\i, 0x30+\i*8 (r1) + .endr + + ld r14,0x138(r1); + mtsrr0 r14 + + ld r14,0x140(r1); + mtsrr1 r14 + + ld r14,0x148(r1); + mtcr r14 + + + ld 0,XVECT_M_HANDLER(0) + mtctr 0 + + ld r0,0x30(r1); # restore vector number + ld r1,0x38(r1); + + bctr + +intHandler2C: + mtctr r1 # save old stack pointer + lis r1,0x4 + stdu r1, -0x160(r1) + .irp i, 2,3,4,5,6,7,8,9,10,11,12,13,14,15,16, \ + 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, \ + 27, 28, 29, 30, 31 + std r\i, 0x30+\i*8 (r1) + .endr + + std r0,0x30(r1); # save vector number + + mfctr r14 + std r14,0x38(r1); # save old r1 + + mfsrr0 r14 + std r14,0x138(r1); + + mfsrr1 r14 + std r14,0x140(r1); + + mfcr r14 + std r14,0x148(r1); + + mfxer r14 + std r14,0x150(r1); + + bl toc_init + + ld r3,exception_stack_frame@got(r2) + std r1,0(r3) + + + mr r3,r0 + bl .c_interrupt + + ld r14,0x138(r1); + mtsrr0 r14 + + ld r14,0x140(r1); + mtsrr1 r14 + + ld r14,0x148(r1); + mtcr r14 + + ld r14,0x150(r1); + mtxer r14 + + + .irp i, 2,3,4,5,6,7,8,9,10,11,12,13,14,15,16, \ + 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, \ + 27, 28, 29, 30, 31 + ld r\i, 0x30+\i*8 (r1) + .endr + + ld r1,0x38(r1); + + mfsprg r0,2 + mtctr r0 + mfsprg r0,3 + mtlr r0 + mfsprg r0,0 + rfid + +/* Set exception handler for given exception vector. + r3: exception vector offset + r4: exception handler +*/ + .globl .set_exception +.set_exception: + .globl set_exception +set_exception: + ld r4,0x0(r4) + .globl .set_exception_asm +.set_exception_asm: + .globl set_exception_asm +set_exception_asm: + std r4, 0x60(r3) # fixme diff 1f - 0b + blr + + +setup_mem_u3: + li 4,0x2000 ; oris 4,4,0xf800 + + # MemTimingParam -- CAS lat 2.5 / 4 (read-to-read / read-to-write) + lis 3,0x49e1 ; ori 3,3,0xa000 ; stw 3,0x50(4) + + # MRSRegCntl -- CAS lat 2.5 + li 3,0x6a ; stw 3,0xf0(4) + + # MemBusConfig -- 128 bit bus + lis 3,0x8500 ; stw 3,0x190(4) + + # CKDelAdj -- clock delay 75 + lis 3,0x12c3 ; ori 3,3,0x30cc ; stw 3,0x520(4) + + # IOModeCntl -- no termination on differential and 3-state drivers + lis 3,0x0350 ; stw 3,0x530(4) + + li 3,18 ; mtctr 3 ; addi 5,4,0x5f0 +0: # DQSDelAdj -- read delay offset -10 + lis 3,0x3d8f ; ori 3,3,0x6000 ; stwu 3,0x10(5) + + # DQSDataDelAdj -- write delay offset -32, write data delay offset +15 + lis 3,0x380e ; ori 3,3,0x003c ; stwu 3,0x10(5) + bdnz 0b + + # MemProgCntl -- set all + lis 3,0xc000 ; stw 3,0xe0(4) + + eieio + + blr + + +# read dimm SPDs, program memory size and type +setup_mem_size: + mflr 14 + + li 15,0 ; oris 15,15,0xf800 ; li 17,0 + li 3,0xa0 ; li 4,3 ; li 5,3 ; bl i2c_read + mr 16,4 ; cmpdi 3,0 ; beq 0f ; li 16,0 +0: li 3,0xa2 ; li 4,3 ; li 5,3 ; bl i2c_read + cmpd 16,4 ; bne 0f ; cmpdi 3,0 ; beq 1f +0: li 16,0x1e00 +1: #li 3,0xd ; bl print_byte ; li 3,0xa ; bl print_byte + #mr 3,16 ; bl print_hex + + #li 3,0x20 ; bl print_byte + sldi 3,16,7 ; add 3,3,16 ; rlwinm 3,3,10,0,6 ; subis 3,3,0x3c00 + stw 3,0x21c0(15) ; andi. 0,16,2 ; beq 0f ; stw 3,0x21e0(15) +0: #bl print_hex + sldi 3,16,8 ; add 3,3,16 ; rldicl 3,3,48,56 ; li 0,8 ; slw 3,0,3 + # slw, not sld, so that empty/bad banks translate into size 0 + stw 17,0x21d0(15) ; bl add17173 ; stw 17,0x21f0(15) + andi. 0,16,2 ; beq 0f ; bl add17173 +0: #bl print_hex + + li 3,0xa4 ; li 4,3 ; li 5,3 ; bl i2c_read + mr 16,4 ; cmpdi 3,0 ; beq 0f ; li 16,0 +0: li 3,0xa6 ; li 4,3 ; li 5,3 ; bl i2c_read + cmpd 16,4 ; bne 0f ; cmpdi 3,0 ; beq 1f +0: li 16,0x1e00 +1: #li 3,0xd ; bl print_byte ; li 3,0xa ; bl print_byte + #mr 3,16 ; bl print_hex + + #li 3,0x20 ; bl print_byte + sldi 3,16,7 ; add 3,3,16 ; rlwinm 3,3,10,0,6 ; subis 3,3,0x3c00 + stw 3,0x2200(15) ; andi. 0,16,2 ; beq 0f ; stw 3,0x2220(15) +0: #bl print_hex + sldi 3,16,8 ; add 3,3,16 ; rldicl 3,3,48,56 ; li 0,8 ; slw 3,0,3 + stw 17,0x2210(15) ; bl add17173 ; stw 17,0x2230(15) + andi. 0,16,2 ; beq 0f ; bl add17173 +0: #bl print_hex + #mr 3,17 ; bl print_hex + stw 17,0x2250(15) ; stw 17,0x2270(15) + stw 17,0x2290(15) ; stw 17,0x22b0(15) + + mtlr 14 + blr + + + + +# print GPR3 as 8-digit hex. uses GPR18,19 +print_hex: + mflr 18 ; mr 19,3 ; li 3,8 ; mtctr 3 +1: rlwinm 3,19,4,28,31 ; sldi 19,19,4 + cmpdi 3,0xa ; blt 0f ; addi 3,3,0x27 +0: addi 3,3,0x30 ; bl putc + bdnz 1b ; mtlr 18 ; blr + + +# i2c stuff uses GPR20..GPR24 + +# terminate any i2c transaction, at any point during that transaction +i2c_stop: +0: lwz 3,0x30(20) ; stw 3,0x30(20) ; andi. 3,3,4 ; beq 0b + mr 3,21 ; mr 4,22 ; mtlr 24 ; eieio ; blr + +# do a combined-mode read +# in: GPR3 = addr, GPR4 = subaddr, GPR5 = len +# out: GPR3 = error, GPR4 = result (right-aligned, msb) +i2c_read: + mflr 24 + li 20,0x1000 ; oris 20,20,0xf800 # uni-n i2c base + mr 21,3 ; mr 22,4 ; mr 23,5 # save params + li 4,0xc ; stw 4,0(20) # set mode (combined) + ori 4,21,1 ; stw 4,0x50(20) # set addr, read + stw 22,0x60(20) # set subaddr + li 4,2 ; stw 4,0x10(20) ; eieio # start address phase + li 21,1 # error + li 22,0 # result accumulator +0: lwz 3,0x30(20) ; andi. 3,3,2 ; beq 0b # wait until sent + lwz 3,0x20(20) ; andi. 3,3,2 ; beq i2c_stop # check result + li 4,1 ; cmpdi 23,1 ; bne 0f ; li 4,0 +0: stw 4,0x10(20) # AAK for next byte (or not) + li 4,2 ; stw 4,0x30(20) ; eieio # ack address phase +i2c_read_loop: + lwz 3,0x30(20) ; andi. 3,3,1 ; beq 1f # if byte recv'd: + subi 23,23,1 ; sldi 22,22,8 # shift byte accum + lwz 3,0x70(20) ; rlwimi 22,3,0,24,31 # get byte + cmpdi 23,0 ; bne 0f ; li 21,0 ; b i2c_stop # all done +0: li 4,1 ; cmpdi 23,1 ; bne 0f ; li 4,0 +0: stw 4,0x10(20) # AAK for next byte (or not) + li 4,1 ; stw 4,0x30(20) ; eieio # ack data phase +1: lwz 3,0x30(20) ; andi. 3,3,4 ; beq i2c_read_loop + li 4,0 ; stw 4,0x10(20) ; eieio ; b i2c_stop # stop bit received + +add17173: # add GPR3 into GPR17; if passing 2GB (0x10000000), add another 2GB. + lis 0,0x1000 ; cmpld 17,0 ; add 17,17,3 ; bgtlr + cmpld 17,0 ; blelr ; add 17,17,0 ; blr + +io_log_init: + LOAD64(r3, SB_NVRAM_adr) + b checkinitLog diff --git a/board-js2x/llfw/u4mem.c b/board-js2x/llfw/u4mem.c new file mode 100644 index 0000000..d0fc4cd --- /dev/null +++ b/board-js2x/llfw/u4mem.c @@ -0,0 +1,4071 @@ +/****************************************************************************** + * Copyright (c) 2004, 2007 IBM Corporation + * All rights reserved. + * This program and the accompanying materials + * are made available under the terms of the BSD License + * which accompanies this distribution, and is available at + * http://www.opensource.org/licenses/bsd-license.php + * + * Contributors: + * IBM Corporation - initial implementation + *****************************************************************************/ +#include <types.h> +#include <hw.h> +#include <stdio.h> +#include "stage2.h" +#include <cpu.h> +#include <string.h> + +/* + * compiler switches + ******************************************************************************* + */ +#define U4_DEBUG +#define U4_INFO +//#define U4_SHOW_REGS + +int io_getchar(char *); + +/* + * version info + */ +static const uint32_t VER = 2; +static const uint32_t SUBVER = 1; + +/* + * local macros + ******************************************************************************* + */ +// bit shifting in Motorola/IBM bit enumeration format (yaks...) +#define IBIT( nr ) ( (uint32_t) 0x80000000 >> (nr) ) +#define BIT( nr ) ( (uint32_t) 0x1 << (nr) ) + +/* + * macros to detect the current board layout + */ +#ifdef BOARD_BIMINI +#define IS_MAUI 0 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 0 ) +#define IS_BIMINI 1 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 1 ) +#define IS_KAUAI 0 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 2 ) +#endif + +#ifdef BOARD_MAUI +#define IS_MAUI 1 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 0 ) +#define IS_BIMINI 0 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 1 ) +#define IS_KAUAI 0 //( ( load8_ci( 0xf4000682 ) >> 4 ) == 2 ) +#endif + +/* + * local constants + ******************************************************************************* + */ + +/* + * u4 base address + */ +#define U4_BASE_ADDR ((uint64_t) 0xf8000000 ) +#define u4reg( reg ) (U4_BASE_ADDR + (uint64_t) (reg)) + +/* + * I2C registers + */ +#define I2C_MODE_R u4reg(0x1000) +#define I2C_CTRL_R u4reg(0x1010) +#define I2C_STAT_R u4reg(0x1020) +#define I2C_ISR_R u4reg(0x1030) +#define I2C_ADDR_R u4reg(0x1050) +#define I2C_SUBA_R u4reg(0x1060) +#define I2C_DATA_R u4reg(0x1070) + +/* + * clock control registers & needed bits/masks + */ +#define ClkCntl_R u4reg(0x0800) +#define PLL2Cntl_R u4reg(0x0860) + +/* + * clock control bits & masks + */ +#define CLK_DDR_CLK_MSK (IBIT(11) | IBIT(12) | IBIT(13)) + +/* + * memory controler registers + */ +#define RASTimer0_R u4reg(0x2030) +#define RASTimer1_R u4reg(0x2040) +#define CASTimer0_R u4reg(0x2050) +#define CASTimer1_R u4reg(0x2060) +#define MemRfshCntl_R u4reg(0x2070) +#define MemProgCntl_R u4reg(0x20b0) +#define Dm0Cnfg_R u4reg(0x2200) +#define Dm1Cnfg_R u4reg(0x2210) +#define Dm2Cnfg_R u4reg(0x2220) +#define Dm3Cnfg_R u4reg(0x2230) +#define MemWrQCnfg_R u4reg(0x2270) +#define MemArbWt_R u4reg(0x2280) +#define UsrCnfg_R u4reg(0x2290) +#define MemRdQCnfg_R u4reg(0x22a0) +#define MemQArb_R u4reg(0x22b0) +#define MemRWArb_R u4reg(0x22c0) +#define MemBusCnfg_R u4reg(0x22d0) +#define MemBusCnfg2_R u4reg(0x22e0) +#define ODTCntl_R u4reg(0x23a0) +#define MemModeCntl_R u4reg(0x2500) +#define MemPhyModeCntl_R u4reg(0x2880) +#define CKDelayL_R u4reg(0x2890) +#define CKDelayU_R u4reg(0x28a0) +#define IOPadCntl_R u4reg(0x29a0) +#define ByteWrClkDelC0B00_R u4reg(0x2800) +#define ByteWrClkDelC0B01_R u4reg(0x2810) +#define ByteWrClkDelC0B02_R u4reg(0x2820) +#define ByteWrClkDelC0B03_R u4reg(0x2830) +#define ByteWrClkDelC0B04_R u4reg(0x2900) +#define ByteWrClkDelC0B05_R u4reg(0x2910) +#define ByteWrClkDelC0B06_R u4reg(0x2920) +#define ByteWrClkDelC0B07_R u4reg(0x2930) +#define ByteWrClkDelC0B16_R u4reg(0x2980) +#define ByteWrClkDelC0B08_R u4reg(0x2a00) +#define ByteWrClkDelC0B09_R u4reg(0x2a10) +#define ByteWrClkDelC0B10_R u4reg(0x2a20) +#define ByteWrClkDelC0B11_R u4reg(0x2a30) +#define ByteWrClkDelC0B12_R u4reg(0x2b00) +#define ByteWrClkDelC0B13_R u4reg(0x2b10) +#define ByteWrClkDelC0B14_R u4reg(0x2b20) +#define ByteWrClkDelC0B15_R u4reg(0x2b30) +#define ByteWrClkDelC0B17_R u4reg(0x2b80) +#define ReadStrobeDelC0B00_R u4reg(0x2840) +#define ReadStrobeDelC0B01_R u4reg(0x2850) +#define ReadStrobeDelC0B02_R u4reg(0x2860) +#define ReadStrobeDelC0B03_R u4reg(0x2870) +#define ReadStrobeDelC0B04_R u4reg(0x2940) +#define ReadStrobeDelC0B05_R u4reg(0x2950) +#define ReadStrobeDelC0B06_R u4reg(0x2960) +#define ReadStrobeDelC0B07_R u4reg(0x2970) +#define ReadStrobeDelC0B16_R u4reg(0x2990) +#define ReadStrobeDelC0B08_R u4reg(0x2a40) +#define ReadStrobeDelC0B09_R u4reg(0x2a50) +#define ReadStrobeDelC0B10_R u4reg(0x2a60) +#define ReadStrobeDelC0B11_R u4reg(0x2a70) +#define ReadStrobeDelC0B12_R u4reg(0x2b40) +#define ReadStrobeDelC0B13_R u4reg(0x2b50) +#define ReadStrobeDelC0B14_R u4reg(0x2b60) +#define ReadStrobeDelC0B15_R u4reg(0x2b70) +#define ReadStrobeDelC0B17_R u4reg(0x2b90) +#define MemInit00_R u4reg(0x2100) +#define MemInit01_R u4reg(0x2110) +#define MemInit02_R u4reg(0x2120) +#define MemInit03_R u4reg(0x2130) +#define MemInit04_R u4reg(0x2140) +#define MemInit05_R u4reg(0x2150) +#define MemInit06_R u4reg(0x2160) +#define MemInit07_R u4reg(0x2170) +#define MemInit08_R u4reg(0x2180) +#define MemInit09_R u4reg(0x2190) +#define MemInit10_R u4reg(0x21a0) +#define MemInit11_R u4reg(0x21b0) +#define MemInit12_R u4reg(0x21c0) +#define MemInit13_R u4reg(0x21d0) +#define MemInit14_R u4reg(0x21e0) +#define MemInit15_R u4reg(0x21f0) +#define CalConf0_R u4reg(0x29b0) +#define CalConf1_R u4reg(0x29c0) +#define MeasStatusC0_R u4reg(0x28f0) +#define MeasStatusC1_R u4reg(0x29f0) +#define MeasStatusC2_R u4reg(0x2af0) +#define MeasStatusC3_R u4reg(0x2bf0) +#define CalC0_R u4reg(0x28e0) +#define CalC1_R u4reg(0x29e0) +#define CalC2_R u4reg(0x2ae0) +#define CalC3_R u4reg(0x2be0) +#define RstLdEnVerniersC0_R u4reg(0x28d0) +#define RstLdEnVerniersC1_R u4reg(0x29d0) +#define RstLdEnVerniersC2_R u4reg(0x2ad0) +#define RstLdEnVerniersC3_R u4reg(0x2bd0) +#define ExtMuxVernier0_R u4reg(0x28b0) +#define ExtMuxVernier1_R u4reg(0x28c0) +#define OCDCalCmd_R u4reg(0x2300) +#define OCDCalCntl_R u4reg(0x2310) +#define MCCR_R u4reg(0x2440) +#define MSRSR_R u4reg(0x2410) +#define MSRER_R u4reg(0x2420) +#define MSPR_R u4reg(0x2430) +#define MSCR_R u4reg(0x2400) +#define MEAR0_R u4reg(0x2460) +#define MEAR1_R u4reg(0x2470) +#define MESR_R u4reg(0x2480) +#define MRSRegCntl_R u4reg(0x20c0) +#define EMRSRegCntl_R u4reg(0x20d0) +#define APIMemRdCfg_R u4reg(0x30090) +#define APIExcp_R u4reg(0x300a0) + +/* + * common return values + */ +#define RET_OK 0 +#define RET_ERR -1 +#define RET_ACERR_CE -1 +#define RET_ACERR_UEWT -2 +#define RET_ACERR_UE -3 + +/* + * 'DIMM slot populated' indicator + */ +#define SL_POP 1 + +/* + * spd buffer size + */ +#define SPD_BUF_SIZE 0x40 + +/* + * maximum number of DIMM banks & DIMM groups + */ +#define NUM_SLOTS 8 +#define NUM_BANKS ( NUM_SLOTS / 2 ) +#define MAX_DGROUPS ( NUM_SLOTS / 2 ) +#define SLOT_ADJ() ( ( IS_MAUI ) ? NUM_SLOTS / 4 : NUM_SLOTS / 2 ) + +/* + * values needed for auto calibration + */ +#define MAX_DRANKS NUM_SLOTS +#define MAX_BLANE 18 +#define MAX_RMD 0xf + +/* + * maximum number of supported CAS latencies + */ +#define NUM_CL 3 + +/* + * min/max supported CL values by U4 + */ +#define U4_MIN_CL 3 +#define U4_MAX_CL 5 + +/* + * DIMM constants + */ +#define DIMM_TYPE_MSK BIT(0) +#define DIMM_ORG_x4 BIT(0) +#define DIMM_ORG_x8 BIT(1) +#define DIMM_ORG_x16 BIT(2) +#define DIMM_ORG_MIXx8x16 BIT(30) +#define DIMM_ORG_UNKNOWN 0 +#define DIMM_WIDTH 72 +#define DIMM_BURSTLEN_4 BIT(2) + +/* + * L2 cache size + */ +#define L2_CACHE_SIZE (uint32_t) 0x100000 + +/* + * scrub types + */ +#define IMMEDIATE_SCRUB IBIT(0) +#define IMMEDIATE_SCRUB_WITH_FILL ( IBIT(0) | IBIT(1) ) +#define BACKGROUND_SCRUB ( IBIT(1) | ( 0x29 << 16 ) ) + +/* + * I2C starting slave addresses of the DIMM banks + */ +#define I2C_START 0x50 + +/* + * Index to the speed dependend DIMM settings + */ +enum +{ + SPEED_IDX_400 = 0, + SPEED_IDX_533, + SPEED_IDX_667, + NUM_SPEED_IDX +}; + +/* + * number of read/write strobes of the U4 + */ +#define NUM_STROBES 18 + +/* + * 2GB hole definition + */ +static const uint64_t _2GB = (uint64_t) 0x80000000; + +/* + * local types + ******************************************************************************* + */ +/* + * DIMM definition + */ +typedef struct +{ + uint32_t m_pop_u32; // set if bank is populated + uint32_t m_bank_u32; // bank number + uint32_t m_clmsk_u32; // mask of supported CAS latencies + uint32_t m_clcnt_u32; // number of supporetd CAS latencies + uint32_t m_clval_pu32[NUM_CL]; // values of supporeted CAS latencies + uint32_t m_speed_pu32[NUM_CL]; // speed (Mhz) at CAS latency of same index + uint32_t m_size_u32; // chip size in Mb + uint32_t m_rank_u32; // # of ranks, total size = chip size*rank + uint32_t m_orgmsk_u32; // data organisation (x4, x8, x16) (mask) + uint32_t m_orgval_u32; // data organisation (value) + uint32_t m_width_u32; // data width + uint32_t m_ecc_u32; // set if ecc + uint32_t m_type_u32; // rdimm or udimm + uint32_t m_burst_u32; // supported burst lengths + uint32_t m_bankcnt_u32; // number of banks + + /* + * the following timing values are all in 1/100ns + */ + uint32_t m_tCK_pu32[NUM_CL]; + uint32_t m_tRAS_u32; + uint32_t m_tRTP_u32; + uint32_t m_tRP_u32; + uint32_t m_tWR_u32; + uint32_t m_tRRD_u32; + uint32_t m_tRC_u32; + uint32_t m_tRCD_u32; + uint32_t m_tWTR_u32; + uint32_t m_tREF_u32; + uint32_t m_tRFC_u32; +} dimm_t; + +/* + * DIMM group definition + */ +typedef struct +{ + uint32_t m_size_u32; // group size in MB + uint32_t m_start_u32; // in 128Mb granularity + uint32_t m_end_u32; // in 128Mb granularity + uint32_t m_ss_u32; // single sided/double sided + uint32_t m_csmode_u32; // selected CS mode for this group + uint32_t m_add2g_u32; + uint32_t m_sub2g_u32; + uint32_t m_memmd_u32; // selected mem mode for this group + uint32_t m_dcnt_u32; // number of DIMMs in group + dimm_t *m_dptr[NUM_SLOTS]; +} dgroup_t; + +/* + * auto calibration result structure + */ +typedef struct +{ + uint32_t m_MemBusCnfg_u32; + uint32_t m_MemBusCnfg2_u32; + uint32_t m_RstLdEnVerniers_pu32[4]; +} auto_calib_t; + +/* + * ECC error structure + */ +typedef struct +{ + int32_t m_err_i32; + uint32_t m_uecnt_u32; // number of uncorrectable errors + uint32_t m_cecnt_u32; // number of correctable errors + uint32_t m_rank_u32; // erroneous rank + uint32_t m_col_u32; // erroneous column + uint32_t m_row_u32; // erroneous row + uint32_t m_bank_u32; // erroneous bank +} eccerror_t; + +/* + * U4 register setup structure + */ +typedef struct +{ + /* + * external MUX delays + */ + uint32_t RRMux; + uint32_t WRMux; + uint32_t WWMux; + uint32_t RWMux; + + /* + * default Wr/Rd Queue & Arbiter register settings + */ + uint32_t MemRdQCnfg; + uint32_t MemWrQCnfg; + uint32_t MemQArb; + uint32_t MemRWArb; + + /* + * misc fixed register values + */ + uint32_t ODTCntl; + uint32_t IOPadCntl; + uint32_t MemPhyModeCntl; + uint32_t OCDCalCntl; + uint32_t OCDCalCmd; + uint32_t CKDelayL; + uint32_t CKDelayU; + uint32_t MemBusCnfg; + uint32_t CAS1Dly0; + uint32_t CAS1Dly1; + uint32_t ByteWrClkDel[NUM_STROBES]; + uint32_t ReadStrobeDel[NUM_STROBES]; +} reg_statics_t; + +/* + * local variables + ******************************************************************************* + */ +static dimm_t m_dimm[NUM_SLOTS]; +static dimm_t m_gendimm; +static uint32_t m_dcnt_u32; +static dimm_t *m_dptr[NUM_SLOTS]; +static uint32_t m_bankoff_u32; +static uint32_t m_bankpop_u32[NUM_BANKS]; +static uint32_t m_dclidx_u32; +static uint32_t m_dgrcnt_u32; +static dgroup_t m_dgroup[MAX_DGROUPS]; +static dgroup_t *m_dgrptr[MAX_DGROUPS]; +static uint64_t m_memsize_u64; // memsize in bytes + +/* + * local functions + ******************************************************************************* + */ +static void +progbar( void ) +{ + static uint8_t bar[] = + { '|', '/', '-', '\\', 0 }; + static uint32_t idx = 0; + + printf( "\b%c", bar[idx] ); + + if( bar[++idx] == 0 ) { + idx = 0; + } + +} + +static void +or32_ci( uint64_t r, uint32_t m ) +{ + uint32_t v; + + v = load32_ci( r ); + v |= m; + store32_ci( r, v ); +} + +static void +and32_ci( uint64_t r, uint32_t m ) +{ + uint32_t v; + + v = load32_ci( r ); + v &= m; + store32_ci( r, v ); +} + +static void +dly( uint64_t volatile f_wait_u64 ) \ +{ + while( f_wait_u64 ) { + f_wait_u64--; + } +} + +/* + * local i2c access functions + */ +static void +i2c_term( void ) +{ + uint32_t l_stat_u32; + + /* + * clear out all pending int's and wait + * for the stop condition to occur + */ + do { + l_stat_u32 = load32_ci( I2C_ISR_R ); + store32_ci( I2C_ISR_R, l_stat_u32 ); + } while( ( l_stat_u32 & IBIT(29) ) == 0 ); + +} + +static int32_t +i2c_read( uint32_t f_addr_u32, uint32_t f_suba_u32, uint8_t *f_buf_pu08, uint32_t f_len_u32 ) +{ + uint32_t l_val_u32; + int32_t l_ret_i32 = 1; + + /* + * parameter check + */ + if( ( f_addr_u32 > (uint32_t) 0x7f ) || + ( f_suba_u32 > (uint32_t) 0xff ) || + ( f_len_u32 == (uint32_t) 0x00 ) ) { + return RET_ERR; + } + + /* + * set I2C Interface to combined mode + */ + store32_ci( I2C_MODE_R, IBIT(28) | IBIT(29) ); + + /* + * set address, subaddress & read mode + */ + store32_ci( I2C_ADDR_R, ( f_addr_u32 << 1 ) | (uint32_t) 0x1 ); + store32_ci( I2C_SUBA_R, f_suba_u32 ); + + /* + * start address transmission phase + */ + store32_ci( I2C_CTRL_R, IBIT(30) ); + + /* + * wait for address transmission to finish + */ + do { + l_val_u32 = load32_ci( I2C_ISR_R ); + } while( ( l_val_u32 & IBIT(30) ) == 0 ); + + /* + * check for success + */ + if( ( load32_ci( I2C_STAT_R ) & IBIT(30) ) == 0 ) { + i2c_term(); + return RET_ERR; + } else { + // send ack + store32_ci( I2C_CTRL_R, IBIT(31) ); + // clear int + store32_ci( I2C_ISR_R, IBIT(30) ); + } + + /* + * read data + */ + while( l_ret_i32 > 0 ) { + l_val_u32 = load32_ci( I2C_ISR_R ); + + if( ( l_val_u32 & IBIT(31) ) != 0 ) { + // data was received + *f_buf_pu08 = ( uint8_t ) load32_ci( I2C_DATA_R ); + + f_buf_pu08++; + f_len_u32--; + + /* + * continue when there is more data to read or + * exit if not + */ + if( f_len_u32 != 0 ) { + // send ack + store32_ci( I2C_CTRL_R, IBIT(31) ); + // clear int + store32_ci( I2C_ISR_R, IBIT(31) ); + } else { + // send nack + store32_ci( I2C_CTRL_R, 0 ); + // set exit flag + l_ret_i32 = RET_OK; + } + + } else if( ( l_val_u32 & IBIT(29) ) != 0 ) { + // early stop condition + // set exit flag + l_ret_i32 = RET_ERR; + } + + }; + + i2c_term(); + + return( l_ret_i32 ); +} + +static uint32_t +i2c_get_slot( uint32_t i2c_addr ) +{ + uint32_t slot; + + slot = ( i2c_addr - I2C_START ) / 2; + + if( ( i2c_addr & 0x1 ) != 0 ) { + slot += SLOT_ADJ(); + } + + return slot; +} + +/* + * 'serial presence detect' interpretation functions + */ +static uint32_t +ddr2_get_dimm_rank( uint8_t *f_spd_pu08 ) +{ + static const int RANK_IDX = (int) 5; + + return (uint32_t) ( f_spd_pu08[RANK_IDX] & 0x3 ) + 1; +} + +static uint32_t +ddr2_get_dimm_size( uint8_t *f_spd_pu08 ) +{ + static const int SIZE_IDX = (int) 31; + uint8_t l_smsk_u08; + uint32_t i; + + l_smsk_u08 = ( f_spd_pu08[SIZE_IDX] << 3 ) | + ( f_spd_pu08[SIZE_IDX] >> 5 ); + + for( i = 0; ( ( l_smsk_u08 & ( (uint8_t) 0x1 << i ) ) == 0 ) ; i++ ); + + return (uint32_t) 0x80 << i; +} + +static uint32_t +ddr2_get_dimm_type( uint8_t *f_spd_pu08 ) +{ + static const int TYPE_IDX = (int) 20; + + return (uint32_t) f_spd_pu08[TYPE_IDX] & DIMM_TYPE_MSK; +} + +static uint32_t +ddr2_get_dimm_org( uint8_t *f_spd_pu08, uint32_t /*out*/ *f_omsk_pu32 ) +{ + static const int ORG_IDX = (int) 13; + uint32_t l_ret_u32 = (uint32_t) f_spd_pu08[ORG_IDX]; + + if( l_ret_u32 == 4 ) { + *f_omsk_pu32 = DIMM_ORG_x4; + } else if( l_ret_u32 == 8 ) { + *f_omsk_pu32 = DIMM_ORG_x8; + *f_omsk_pu32 |= DIMM_ORG_MIXx8x16; + } else if( l_ret_u32 == 16 ) { + *f_omsk_pu32 = DIMM_ORG_x16; + *f_omsk_pu32 |= DIMM_ORG_MIXx8x16; + } else { + *f_omsk_pu32 = DIMM_ORG_UNKNOWN; + l_ret_u32 = (uint32_t) ~0; + } + + return l_ret_u32; +} + +static uint32_t +ddr2_get_dimm_width( uint8_t *f_spd_pu08 ) +{ + static const int WIDTH_IDX = (int) 6; + + return (uint32_t) f_spd_pu08[WIDTH_IDX]; +} + +static uint32_t +ddr2_get_dimm_ecc( uint8_t *f_spd_pu08 ) +{ + static const int ECC_IDX = (int) 11; + + return ( f_spd_pu08[ECC_IDX] & BIT(1) ) != 0; +} + +static uint32_t +ddr2_get_dimm_burstlen( uint8_t *f_spd_pu08 ) +{ + static const int BURST_IDX = (int) 16; + + return (uint32_t) f_spd_pu08[BURST_IDX]; +} + +static void +ddr2_get_dimm_speed( dimm_t *f_dimm, uint8_t *f_spd_pu08 ) +{ + static const int SPEED_IDX[] = { 25, 23, 9 }; + static const uint32_t NS[] = { 25, 33, 66, 75 }; + uint8_t l_tmp_u08; + uint32_t l_dspeed_u32; + uint32_t idx = 0; + uint32_t i; + + for( i = NUM_CL - f_dimm->m_clcnt_u32; i < NUM_CL; i++ ) { + l_tmp_u08 = f_spd_pu08[SPEED_IDX[i]]; + l_dspeed_u32 = (uint32_t) ( l_tmp_u08 >> 4 ) * 100; + l_tmp_u08 &= (uint8_t) 0xf; + + if( l_tmp_u08 >= (uint8_t) 10 ) { + l_dspeed_u32 += NS[l_tmp_u08 - 10]; + } else { + l_dspeed_u32 += (uint32_t) l_tmp_u08 * 10; + } + + f_dimm->m_tCK_pu32[idx] = l_dspeed_u32; + f_dimm->m_speed_pu32[idx] = (uint32_t) 2000000 / l_dspeed_u32; + f_dimm->m_speed_pu32[idx] += (uint32_t) 5; + f_dimm->m_speed_pu32[idx] /= (uint32_t) 10; + idx++; + } + +} + +static void +ddr2_get_dimm_timings( dimm_t *f_dimm, uint8_t *f_spd_pu08 ) +{ + static const uint32_t NS[] = { 00, 25, 33, 50, 66, 75, 00, 00 }; + static const uint32_t USMUL = (uint32_t) 390625; + static const int tREF_IDX = (int) 12; + static const int tRP_IDX = (int) 27; + static const int tRRD_IDX = (int) 28; + static const int tRCD_IDX = (int) 29; + static const int tRAS_IDX = (int) 30; + static const int tWR_IDX = (int) 36; + static const int tWTR_IDX = (int) 37; + static const int tRTP_IDX = (int) 38; + static const int tRC_IDX = (int) 41; // & 40 + static const int tRFC_IDX = (int) 42; // & 40 + + uint32_t l_tmp_u32; + + f_dimm->m_tRP_u32 = (uint32_t) f_spd_pu08[tRP_IDX] * 25; + f_dimm->m_tRRD_u32 = (uint32_t) f_spd_pu08[tRRD_IDX] * 25; + f_dimm->m_tRCD_u32 = (uint32_t) f_spd_pu08[tRCD_IDX] * 25; + f_dimm->m_tWR_u32 = (uint32_t) f_spd_pu08[tWR_IDX] * 25; + f_dimm->m_tWTR_u32 = (uint32_t) f_spd_pu08[tWTR_IDX] * 25; + f_dimm->m_tRTP_u32 = (uint32_t) f_spd_pu08[tRTP_IDX] * 25; + f_dimm->m_tRAS_u32 = (uint32_t) f_spd_pu08[tRAS_IDX] * 100; + + l_tmp_u32 = (uint32_t) ( f_spd_pu08[tRC_IDX - 1] >> 4 ); + l_tmp_u32 &= (uint32_t) 0x7; + f_dimm->m_tRC_u32 = (uint32_t) f_spd_pu08[tRC_IDX] * 100 + + NS[l_tmp_u32]; + + l_tmp_u32 = (uint32_t) f_spd_pu08[tRFC_IDX - 2]; + l_tmp_u32 &= (uint32_t) 0xf; + f_dimm->m_tRFC_u32 = (uint32_t) 256 * ( l_tmp_u32 & (uint32_t) 0x1 ); + f_dimm->m_tRFC_u32 += (uint32_t) f_spd_pu08[tRFC_IDX]; + f_dimm->m_tRFC_u32 *= 100; + l_tmp_u32 >>= 1; + f_dimm->m_tRFC_u32 += NS[l_tmp_u32]; + + l_tmp_u32 = (uint32_t) f_spd_pu08[tREF_IDX]; + l_tmp_u32 &= (uint32_t) 0x7f; + + if( l_tmp_u32 == 0 ) { + l_tmp_u32 = (uint32_t) 2; + } else if( l_tmp_u32 <= (uint32_t) 2 ) { + l_tmp_u32--; + } + + f_dimm->m_tREF_u32 = ( l_tmp_u32 + 1 ) * USMUL; +} + +static uint32_t +ddr2_get_banks( uint8_t *f_spd_pu08 ) +{ + static const int BANK_IDX = (int) 17; + + return (uint32_t) f_spd_pu08[BANK_IDX]; +} + +static uint32_t +ddr2_get_cl_mask( uint8_t *f_spd_pu08 ) +{ + static const int CL_IDX = (int) 18; + + return (uint32_t) f_spd_pu08[CL_IDX]; +} + +static void +ddr2_get_cl( dimm_t *f_dimm ) +{ + uint32_t l_clcnt_u32 = 0; + uint32_t i; + + for( i = 0; ( i < 8 ) && ( l_clcnt_u32 < NUM_CL ) ; i++ ) { + + if( ( f_dimm->m_clmsk_u32 & ( (uint32_t) 0x1 << i ) ) != 0 ) { + f_dimm->m_clval_pu32[l_clcnt_u32] = i; + l_clcnt_u32++; + } + + } + + f_dimm->m_clcnt_u32 = l_clcnt_u32; +} + +static uint32_t +ddr2_cl2speed( dimm_t *f_dimm, uint32_t f_cl_u32, uint32_t *f_tCK_pu32 ) +{ + uint32_t i; + + for( i = 0; ( f_dimm->m_clval_pu32[i] != f_cl_u32 ) && + ( i < NUM_CL ); i++ ); + + if( i == NUM_CL ) { + return (uint32_t) ~0; + } + + *f_tCK_pu32 = f_dimm->m_tCK_pu32[i]; + + return f_dimm->m_speed_pu32[i]; +} + +static void +ddr2_setupDIMM( dimm_t *f_dimm, uint32_t f_bank_u32, uint8_t *f_spd_pu08 ) +{ + f_dimm->m_pop_u32 = SL_POP; + f_dimm->m_bank_u32 = f_bank_u32; + f_dimm->m_size_u32 = ddr2_get_dimm_size( f_spd_pu08 ); + f_dimm->m_rank_u32 = ddr2_get_dimm_rank( f_spd_pu08 ); + f_dimm->m_type_u32 = ddr2_get_dimm_type( f_spd_pu08 ); + f_dimm->m_orgval_u32 = ddr2_get_dimm_org( f_spd_pu08, &f_dimm->m_orgmsk_u32 ); + f_dimm->m_width_u32 = ddr2_get_dimm_width( f_spd_pu08 ); + f_dimm->m_ecc_u32 = ddr2_get_dimm_ecc( f_spd_pu08 ); + f_dimm->m_burst_u32 = ddr2_get_dimm_burstlen( f_spd_pu08 ); + f_dimm->m_clmsk_u32 = ddr2_get_cl_mask( f_spd_pu08 ); + f_dimm->m_bankcnt_u32 = ddr2_get_banks( f_spd_pu08 ); + + ddr2_get_cl( f_dimm ); + ddr2_get_dimm_speed( f_dimm, f_spd_pu08 ); + ddr2_get_dimm_timings( f_dimm, f_spd_pu08 ); +} + +static int32_t +ddr2_checkSPD( uint8_t *f_spd_pu08 ) +{ + uint8_t crc = 0; + uint32_t i; + + for( i = 0; i < SPD_BUF_SIZE - 1; i++ ) { + crc += f_spd_pu08[i]; + } + + if( crc != f_spd_pu08[i] ) { + return RET_ERR; + } + + return RET_OK; +} + +static int32_t +ddr2_readSPDs( void ) +{ + static const uint32_t MAX_SPD_FAIL = 3; + uint8_t l_spdbuf_pu08[SPD_BUF_SIZE]; + uint32_t l_bankfail_u32 = 0; + uint32_t l_spdfail_u32 = 0; + int32_t l_i2c_i32 = RET_OK; + int32_t l_spd_i32 = RET_OK; + int32_t ret = RET_OK; + uint32_t i; + + /* + * read spd's and detect populated slots + */ + for( i = 0; i < NUM_SLOTS; i++ ) { + /* + * indicate slot as empty + */ + m_dimm[i].m_pop_u32 = 0; + + /* + * check whether bank is switched off + */ + if( ( m_bankoff_u32 & ( 0x1 << ( i / 2 ) ) ) != 0 ) { + continue; + } + + /* + * read SPD data + */ + + /* + * reset SPD fail counter + */ + l_spdfail_u32 = MAX_SPD_FAIL; + l_spd_i32 = RET_OK; + + while( l_spdfail_u32 != 0 ) { + l_i2c_i32 = i2c_read( I2C_START + i, 0x0, l_spdbuf_pu08, SPD_BUF_SIZE ); + + if( l_i2c_i32 == RET_OK ) { + l_spd_i32 = ddr2_checkSPD( l_spdbuf_pu08 ); + + if( l_spd_i32 == RET_OK ) { + l_spdfail_u32 = 0; + } else { + l_spdfail_u32--; + } + + } else { + l_spdfail_u32--; + } + + } + + if( l_spd_i32 != RET_OK ) { + #ifdef U4_INFO + printf( "\r\n [ERROR -> SPD read failure in slot %u]", + i2c_get_slot( I2C_START + i ) ); + #endif + + l_bankfail_u32 |= ( 0x1 << ( i / 2 ) ); + ret = RET_ERR; + } else if( l_i2c_i32 == RET_OK ) { + /* + * slot is populated + */ + ddr2_setupDIMM( &m_dimm[i], i / 2, l_spdbuf_pu08 ); + + m_dptr[m_dcnt_u32] = &m_dimm[i]; + m_dcnt_u32++; + } + + } + + if( ret != RET_OK ) { + m_bankoff_u32 |= l_bankfail_u32; + #ifdef U4_INFO + printf( "\r\n" ); + #endif + } + + return ret; +} + +static int32_t +ddr2_setupDIMMcfg( void ) +{ + uint32_t l_tmp_u32; + uint32_t l_tmp0_u32; + uint32_t l_tmp1_u32; + uint32_t i, j, e, b; + + /* + * check wether on board DIMM slot population is valid + */ + e = 0; + b = 0; + for( i = 0; i < NUM_SLOTS; i += 2 ) { + + switch( m_dimm[i].m_pop_u32 + m_dimm[i+1].m_pop_u32 ) { + case 0: { + m_bankpop_u32[i/2] = 0; + break; + } + + case 2 * SL_POP: { + m_bankpop_u32[i/2] = !0; + b++; + break; + } + + default: { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> only 1 DIMM installed in bank %u]", i/2 ); + #endif + e++; + } + + } + + } + + /* + * return on error + */ + if( e != 0 ) { + #ifdef U4_DEBUG + printf( "\r\n" ); + #endif + return RET_ERR; + } + + if( b == 0 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> no (functional) memory installed]\r\n" ); + #endif + return RET_ERR; + } + + /* + * check DIMM compatibility + * configuration is 128 bit data/128 bit bus + * -all DIMMs must be organized as x4 + * -all DIMMs must be 72 bit wide with ECC + * -all DIMMs must be registered DIMMs (RDIMMs) + * -paired DIMMs must have the same # of ranks, size & organization + */ + + /* + * check DIMM ranks & sizes + */ + e = 0; + for( i = 0; i < NUM_SLOTS; i += 2 ) { + + if( ( m_bankpop_u32[i/2] != 0 ) && + ( ( m_dimm[i].m_rank_u32 != m_dimm[i+1].m_rank_u32 ) || + ( m_dimm[i].m_size_u32 != m_dimm[i+1].m_size_u32 ) ) ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> installed DIMMs in bank %u have different ranks/sizes]", i/2 ); + #endif + e++; + } + + } + + /* + * return on error + */ + if( e != 0 ) { + #ifdef U4_DEBUG + printf( "\r\n" ); + #endif + return RET_ERR; + } + + /* + * check valid DIMM organisation (must be x4) + */ + e = 0; + for( i = 0; i < m_dcnt_u32; i++ ) { + + if( ( m_dptr[i]->m_orgmsk_u32 & DIMM_ORG_x4 ) == 0 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> wrong DIMM organisation in bank %u]", + m_dptr[i]->m_bank_u32 ); + #endif + e++; + } + + } + + /* + * return on error + */ + if( e != 0 ) { + #ifdef U4_DEBUG + printf( "\r\n" ); + #endif + return RET_ERR; + } + + e = (uint32_t) ~0; + for( i = 0; i < m_dcnt_u32; i++ ) { + e &= m_dptr[i]->m_type_u32; + } + + /* + * return on error + */ + if( e == 0 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> installed DIMMs are of different type]\r\n" ); + #endif + return RET_ERR; + } + + /* + * setup generic dimm + */ + m_gendimm.m_type_u32 = e; + + /* + * check valid width, ecc & burst length + */ + e = 0; + for( i = 0; i < m_dcnt_u32; i++ ) { + + if( m_dptr[i]->m_width_u32 != DIMM_WIDTH ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> invalid DIMM width in bank %u]", + m_dptr[i]->m_bank_u32 ); + #endif + e++; + } + + if( m_dptr[i]->m_ecc_u32 == 0 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> DIMM(s) do not support ECC in bank %u]", + m_dptr[i]->m_bank_u32 ); + #endif + e++; + } + + if( ( m_dptr[i]->m_burst_u32 & DIMM_BURSTLEN_4 ) == 0 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> DIMM(s) have invalid burst length in bank %u]", + m_dptr[i]->m_bank_u32 ); + #endif + e++; + } + + } + + /* + * return on error + */ + if( e != 0 ) { + #ifdef U4_DEBUG + printf( "\r\n" ); + #endif + return RET_ERR; + } + + /* + * setup generic dimm + */ + m_gendimm.m_width_u32 = m_dptr[0]->m_width_u32; + m_gendimm.m_ecc_u32 = m_dptr[0]->m_ecc_u32; + m_gendimm.m_burst_u32 = m_dptr[0]->m_burst_u32; + + /* + * success + */ + m_gendimm.m_pop_u32 = SL_POP; + + /* + * setup timing parameters + */ + + /* + * find smallest common CL value + */ + l_tmp_u32 = (uint32_t) ~0; + for( i = 0; i < m_dcnt_u32; i++ ) { + l_tmp_u32 &= m_dptr[i]->m_clmsk_u32; + } + + m_gendimm.m_clmsk_u32 = l_tmp_u32; + ddr2_get_cl( &m_gendimm ); + + /* + * find fastest common DIMM speed for all common CL values + */ + for( i = 0; i < m_gendimm.m_clcnt_u32; i++ ) { + m_gendimm.m_speed_pu32[i] = (uint32_t) ~0; + + for( j = 0; j < m_dcnt_u32; j++ ) { + l_tmp0_u32 = + ddr2_cl2speed( m_dptr[j], + m_gendimm.m_clval_pu32[i], + &l_tmp1_u32 ); + + if( m_gendimm.m_speed_pu32[i] > l_tmp0_u32 ) { + m_gendimm.m_speed_pu32[i] = l_tmp0_u32; + m_gendimm.m_tCK_pu32[i] = l_tmp1_u32; + } + + } + + } + + /* + * check wether cl values are supported by U4 + */ + for( i = 0; i < m_gendimm.m_clcnt_u32; i++ ) { + + if( ( m_gendimm.m_clval_pu32[i] >= U4_MIN_CL ) && + ( m_gendimm.m_clval_pu32[i] <= U4_MAX_CL ) ) { + break; + } + + } + + if( i == m_gendimm.m_clcnt_u32 ) { + #ifdef U4_DEBUG + printf( "\r\n [ERROR -> DIMM's CL values not supported]\r\n" ); + #endif + return RET_ERR; + } + + /* + * choose cl/speed values to use: prefer speed over CL + * i holds smallest supported cl value of u4 already + */ + l_tmp_u32 = 0; + while( i < m_gendimm.m_clcnt_u32 ) { + + if( l_tmp_u32 < m_gendimm.m_speed_pu32[i] ) { + l_tmp_u32 = m_gendimm.m_speed_pu32[i]; + m_dclidx_u32 = i; + } + + i++; + } + + /* + * choose largest number of banks + */ + m_gendimm.m_bankcnt_u32 = 0; + + for( i = 0; i < m_dcnt_u32; i++ ) { + + if( m_gendimm.m_bankcnt_u32 < m_dptr[i]->m_bankcnt_u32 ) { + m_gendimm.m_bankcnt_u32 = m_dptr[i]->m_bankcnt_u32; + } + + } + + /* + * setup fastest possible timing parameters for all DIMMs + */ + m_gendimm.m_tRP_u32 = 0; + m_gendimm.m_tRRD_u32 = 0; + m_gendimm.m_tRCD_u32 = 0; + m_gendimm.m_tWR_u32 = 0; + m_gendimm.m_tWTR_u32 = 0; + m_gendimm.m_tRTP_u32 = 0; + m_gendimm.m_tRAS_u32 = 0; + m_gendimm.m_tRC_u32 = 0; + m_gendimm.m_tRFC_u32 = 0; + m_gendimm.m_tREF_u32 = (uint32_t) ~0; + + for( i = 0; i < m_dcnt_u32; i++ ) { + + if( m_gendimm.m_tRP_u32 < m_dptr[i]->m_tRP_u32 ) { + m_gendimm.m_tRP_u32 = m_dptr[i]->m_tRP_u32; + } + + if( m_gendimm.m_tRRD_u32 < m_dptr[i]->m_tRRD_u32 ) { + m_gendimm.m_tRRD_u32 = m_dptr[i]->m_tRRD_u32; + } + + if( m_gendimm.m_tRCD_u32 < m_dptr[i]->m_tRCD_u32 ) { + m_gendimm.m_tRCD_u32 = m_dptr[i]->m_tRCD_u32; + } + + if( m_gendimm.m_tWR_u32 < m_dptr[i]->m_tWR_u32 ) { + m_gendimm.m_tWR_u32 = m_dptr[i]->m_tWR_u32; + } + + if( m_gendimm.m_tWTR_u32 < m_dptr[i]->m_tWTR_u32 ) { + m_gendimm.m_tWTR_u32 = m_dptr[i]->m_tWTR_u32; + } + + if( m_gendimm.m_tRTP_u32 < m_dptr[i]->m_tRTP_u32 ) { + m_gendimm.m_tRTP_u32 = m_dptr[i]->m_tRTP_u32; + } + + if( m_gendimm.m_tRAS_u32 < m_dptr[i]->m_tRAS_u32 ) { + m_gendimm.m_tRAS_u32 = m_dptr[i]->m_tRAS_u32; + } + + if( m_gendimm.m_tRC_u32 < m_dptr[i]->m_tRC_u32 ) { + m_gendimm.m_tRC_u32 = m_dptr[i]->m_tRC_u32; + } + + if( m_gendimm.m_tRFC_u32 < m_dptr[i]->m_tRFC_u32 ) { + m_gendimm.m_tRFC_u32 = m_dptr[i]->m_tRFC_u32; + } + + if( m_gendimm.m_tREF_u32 > m_dptr[i]->m_tREF_u32 ) { + m_gendimm.m_tREF_u32 = m_dptr[i]->m_tREF_u32; + } + + } + + return RET_OK; +} + +static void +u4_group2dimmsDS( dimm_t *f_dimm0, dimm_t *f_dimm1 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -at least 2 slots are populated + * -the 2 DIMMs are equal + * -DIMMs are double sided (2 ranks) + * + * RESULT: + * 1 group of 2 ranks (2 ranks/2 DIMMs) + * -> CS mode 1 (one double sided DIMM pair) + */ + l_dgr->m_size_u32 = 2 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 0; + l_dgr->m_csmode_u32 = 1; + l_dgr->m_dcnt_u32 = 2; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + + m_dgrcnt_u32++; +} + +static void +u4_group2dimmsSS( dimm_t *f_dimm0, dimm_t *f_dimm1 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -at least 2 slots are populated + * -the 2 DIMMs are equal + * -DIMMs are single sided (1 rank) + * + * RESULT: + * 1 group of 1 rank (1 rank/2 DIMMs) + * -> CS mode 0 (one single sided DIMM pair) + */ + l_dgr->m_size_u32 = 2 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 1; + l_dgr->m_csmode_u32 = 0; + l_dgr->m_dcnt_u32 = 2; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + + m_dgrcnt_u32++; +} + +static void +u4_group4dimmsDS( dimm_t *f_dimm0, dimm_t *f_dimm1, + dimm_t *f_dimm2, dimm_t *f_dimm3 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -4 slots are populated + * -all 4 DIMMs are equal + * -DIMMs are double sided (2 ranks) + * + * RESULT: + * 1 group of 4 ranks (2 ranks/2 DIMMs) + * -> CS mode 2 (two double sided DIMM pairs) + */ + l_dgr->m_size_u32 = 4 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 0; + l_dgr->m_csmode_u32 = 2; + l_dgr->m_dcnt_u32 = 4; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + l_dgr->m_dptr[2] = f_dimm2; + l_dgr->m_dptr[3] = f_dimm3; + + m_dgrcnt_u32++; +} + +static void +u4_group4dimmsSS( dimm_t *f_dimm0, dimm_t *f_dimm1, + dimm_t *f_dimm2, dimm_t *f_dimm3 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -4 slots are populated + * -all 4 DIMMs are equal + * -DIMMs are single sided (1 rank) + * + * RESULT: + * 1 group of 2 ranks (1 rank/2 DIMMs) + * -> CS mode 1 (two single sided DIMM pairs) + */ + l_dgr->m_size_u32 = 4 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 1; + l_dgr->m_csmode_u32 = 1; + l_dgr->m_dcnt_u32 = 4; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + l_dgr->m_dptr[2] = f_dimm2; + l_dgr->m_dptr[3] = f_dimm3; + + m_dgrcnt_u32++; +} + +static void +u4_group8dimmsDS( dimm_t *f_dimm0, dimm_t *f_dimm1, + dimm_t *f_dimm2, dimm_t *f_dimm3, + dimm_t *f_dimm4, dimm_t *f_dimm5, + dimm_t *f_dimm6, dimm_t *f_dimm7 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -8 slots are populated + * -all 8 DIMMs are equal + * -DIMMs are double sided (2 ranks) + * + * RESULT: + * 1 group of 8 ranks (2 ranks/2 DIMMs) + * -> CS mode 3 (four double sided DIMM pairs) + */ + l_dgr->m_size_u32 = 8 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 0; + l_dgr->m_csmode_u32 = 3; + l_dgr->m_dcnt_u32 = 8; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + l_dgr->m_dptr[2] = f_dimm2; + l_dgr->m_dptr[3] = f_dimm3; + l_dgr->m_dptr[4] = f_dimm4; + l_dgr->m_dptr[5] = f_dimm5; + l_dgr->m_dptr[6] = f_dimm6; + l_dgr->m_dptr[7] = f_dimm7; + + m_dgrcnt_u32++; +} + +static void +u4_group8dimmsSS( dimm_t *f_dimm0, dimm_t *f_dimm1, + dimm_t *f_dimm2, dimm_t *f_dimm3, + dimm_t *f_dimm4, dimm_t *f_dimm5, + dimm_t *f_dimm6, dimm_t *f_dimm7 ) +{ + dgroup_t *l_dgr = &m_dgroup[m_dgrcnt_u32]; + + /* + * known conditions at this point: + * -8 slots are populated + * -all 8 DIMMs are equal + * -DIMMs are single sided (1 rank) + * + * RESULT: + * 1 group of 4 ranks (1 rank/2 DIMMs) + * -> CS mode 2 (four single sided DIMM pairs) + */ + l_dgr->m_size_u32 = 8 * ( f_dimm0->m_size_u32 * f_dimm0->m_rank_u32 ); + l_dgr->m_ss_u32 = 1; + l_dgr->m_csmode_u32 = 2; + l_dgr->m_dcnt_u32 = 8; + l_dgr->m_dptr[0] = f_dimm0; + l_dgr->m_dptr[1] = f_dimm1; + l_dgr->m_dptr[2] = f_dimm2; + l_dgr->m_dptr[3] = f_dimm3; + l_dgr->m_dptr[4] = f_dimm4; + l_dgr->m_dptr[5] = f_dimm5; + l_dgr->m_dptr[6] = f_dimm6; + l_dgr->m_dptr[7] = f_dimm7; + + m_dgrcnt_u32++; +} + +static int32_t +u4_Dcmp( dimm_t *f_dimm0, dimm_t *f_dimm1 ) +{ + + if( ( f_dimm0->m_size_u32 == f_dimm1->m_size_u32 ) && + ( f_dimm0->m_rank_u32 == f_dimm1->m_rank_u32 ) ) { + return RET_OK; + } + + return RET_ERR; +} + +static void +u4_group1banks( uint32_t *bidx ) +{ + uint32_t didx = 2 * bidx[0]; + + /* + * known conditions at this point: + * -either DIMMs 0 & 4 or + * DIMMs 1 & 5 or + * DIMMs 2 & 6 or + * DIMMs 3 & 7 are populated + * -3 (bimini)/1 (maui) pair of slots is empty + * -installed DIMMs are equal + */ + + /* + * double/single sided setup + */ + if( m_dimm[didx].m_rank_u32 == 1 ) { + u4_group2dimmsSS( &m_dimm[didx], &m_dimm[didx+1] ); + } else { + u4_group2dimmsDS( &m_dimm[didx], &m_dimm[didx+1] ); + } + +} + +static void +u4_group2banks( uint32_t *bidx ) +{ + uint32_t didx0 = 2 * bidx[0]; + uint32_t didx1 = 2 * bidx[1]; + + /* + * known conditions at this point: + * -4 slots are populated + */ + + /* + * check wether DIMM banks may be grouped + */ + if( ( ( ( bidx[0] + bidx[1] ) & 0x1 ) != 0 ) && + ( u4_Dcmp( &m_dimm[didx0], &m_dimm[didx1] ) == 0 ) ) { + /* + * double/single sided setup + * NOTE: at this point all DIMMs have the same amount + * of ranks, therefore only the # of ranks on DIMM 0 is checked + */ + if( m_dimm[didx0].m_rank_u32 == 1 ) { + u4_group4dimmsSS( &m_dimm[didx0], &m_dimm[didx0+1], + &m_dimm[didx1], &m_dimm[didx1+1]); + } else { + u4_group4dimmsDS( &m_dimm[didx0], &m_dimm[didx0+1], + &m_dimm[didx1], &m_dimm[didx1+1]); + } + + } else { + u4_group1banks( &bidx[0] ); + u4_group1banks( &bidx[1] ); + } + +} + +static void +u4_group3banks( uint32_t *bidx ) +{ + + if( ( bidx[0] == 0 ) && ( bidx[1] == 1 ) ) { + u4_group2banks( &bidx[0] ); + u4_group1banks( &bidx[2] ); + } else if( ( bidx[1] == 2 ) && ( bidx[2] == 3 ) ) { + u4_group2banks( &bidx[1] ); + u4_group1banks( &bidx[0] ); + } + +} + +static void +u4_group4banks( uint32_t *bidx ) +{ + uint32_t didx0 = 2 * bidx[0]; + uint32_t didx1 = 2 * bidx[1]; + uint32_t didx2 = 2 * bidx[2]; + uint32_t didx3 = 2 * bidx[3]; + + if( ( u4_Dcmp( &m_dimm[didx0], &m_dimm[didx1] ) == RET_OK ) && + ( u4_Dcmp( &m_dimm[didx2], &m_dimm[didx3] ) == RET_OK ) && + ( u4_Dcmp( &m_dimm[didx0], &m_dimm[didx2] ) == RET_OK ) ) { + + if( m_dimm[didx0].m_rank_u32 == 1 ) { + u4_group8dimmsSS( &m_dimm[didx0], &m_dimm[didx0+1], + &m_dimm[didx1], &m_dimm[didx1+1], + &m_dimm[didx2], &m_dimm[didx2+1], + &m_dimm[didx3], &m_dimm[didx3+1] ); + } else { + u4_group8dimmsDS( &m_dimm[didx0], &m_dimm[didx0+1], + &m_dimm[didx1], &m_dimm[didx1+1], + &m_dimm[didx2], &m_dimm[didx2+1], + &m_dimm[didx3], &m_dimm[didx3+1] ); + } + + } else { + u4_group2banks( &bidx[0] ); + u4_group2banks( &bidx[2] ); + } + +} + +static void +u4_sortDIMMgroups( void ) +{ + uint32_t i, j; + + /* + * setup global group pointers + */ + for( i = 0; i < m_dgrcnt_u32; i++ ) { + m_dgrptr[i] = &m_dgroup[i]; + } + + /* + * use a simple bubble sort to sort groups by size (descending) + */ + for( i = 0; i < ( m_dgrcnt_u32 - 1 ); i++ ) { + + for( j = i + 1; j < m_dgrcnt_u32; j++ ) { + + if( m_dgrptr[i]->m_size_u32 < m_dgrptr[j]->m_size_u32 ) { + dgroup_t *l_sgr; + + l_sgr = m_dgrptr[i]; + m_dgrptr[i] = m_dgrptr[j]; + m_dgrptr[j] = l_sgr; + } + + } + + } + +} + +static void +u4_calcDIMMcnfg( void ) +{ + static const uint32_t _2GB = (uint32_t) 0x00800; + static const uint32_t _4GB = (uint32_t) 0x01000; + static const uint32_t _64GB = (uint32_t) 0x10000; + uint32_t l_start_u32 = (uint32_t) 0; + uint32_t l_end_u32 = (uint32_t) 0; + uint32_t l_add2g_u32 = (uint32_t) 1; + uint32_t l_sub2g_u32 = (uint32_t) 1; + uint32_t i; + + /* + * setup DIMM group parameters + */ + for( i = 0; i < m_dgrcnt_u32; i++ ) { + l_end_u32 = l_start_u32 + m_dgrptr[i]->m_size_u32; + + if( m_dgrptr[i]->m_size_u32 > _2GB ) { + + if( l_end_u32 < _64GB ) { + l_add2g_u32 = ( l_end_u32 >> 11 ); + } else { + l_add2g_u32 = 1; + } + + if( l_start_u32 == 0 ) { + l_sub2g_u32 = 1; + } else { + l_sub2g_u32 = ( l_start_u32 >> 11 ); + } + + } else if( l_add2g_u32 != 1 ) { + l_start_u32 += _2GB; + l_end_u32 += _2GB; + l_add2g_u32 = 1; + l_sub2g_u32 = 1; + } + + /* + * save values for the group + */ + m_dgrptr[i]->m_start_u32 = ( l_start_u32 >> 7 ); // = /128 + m_dgrptr[i]->m_end_u32 = ( l_end_u32 >> 7 ); + m_dgrptr[i]->m_add2g_u32 = l_add2g_u32; + m_dgrptr[i]->m_sub2g_u32 = l_sub2g_u32; + + /* + * continue with next group + */ + if( l_end_u32 != _2GB ) { + l_start_u32 = l_end_u32; + } else { + l_start_u32 = _4GB; + } + + } + +} + +static int32_t +u4_calcDIMMmemmode( void ) +{ + static const uint32_t MAX_ORG = (uint32_t) 0x10; + static const uint32_t MIN_BASE = (uint32_t) 0x80; + static const uint32_t MAX_MODE = (uint32_t) 0x10; + static const uint32_t MODE_ADD = (uint32_t) 0x04; + dimm_t *l_dptr; + uint32_t l_modeoffs_u32; + uint32_t l_sizebase_u32; + int32_t ret = RET_OK; + uint32_t i, j; + + /* + * loop through all DIMM groups and calculate memmode setting + */ + for( i = 0; i < m_dgrcnt_u32; i++ ) { + l_dptr = m_dgrptr[i]->m_dptr[0]; // all dimms in one group are equal! + + l_modeoffs_u32 = MAX_ORG / l_dptr->m_orgval_u32; + l_modeoffs_u32 /= (uint32_t) 2; + l_sizebase_u32 = ( MIN_BASE << l_modeoffs_u32 ); + + j = 0; + while( ( l_sizebase_u32 != l_dptr->m_size_u32 ) && + ( j < MAX_MODE ) ) { + l_sizebase_u32 <<= 1; + j += (uint32_t) MODE_ADD; + } + + // return on error + if( j >= MAX_MODE ) { + #ifdef U4_INFO + uint32_t b, k, l; + printf( "\r\n [ERROR -> unsupported memory type in bank(s)" ); + + l = 0; + for( k = 0; k < m_dgrptr[i]->m_dcnt_u32; k++ ) { + b = m_dgrptr[i]->m_dptr[k]->m_bank_u32; + + if( ( l & ( 1 << b ) ) == 0 ) { + printf( " %u", b ); + l |= ( 1 << b ); + } + + } + + printf( "]\r\n" ); + #endif + + ret = RET_ERR; + } else { + m_dgrptr[i]->m_memmd_u32 = l_modeoffs_u32 + j; + } + + } + + return ret; +} + +static void +u4_setupDIMMgroups( void ) +{ + static const uint64_t _1MB = (uint64_t) 0x100000; + uint32_t l_bcnt_u32; + uint32_t l_bidx_u32[NUM_BANKS]; + uint32_t i; + + /* + * calculate number of populated banks + * IMPORTANT: array must be in ascending order! + */ + l_bcnt_u32 = 0; + for( i = 0; i < NUM_BANKS; i++ ) { + + if( m_bankpop_u32[i] != 0 ) { + l_bidx_u32[l_bcnt_u32] = i; + l_bcnt_u32++; + } + + } + + switch( l_bcnt_u32 ) { + case 4: u4_group4banks( &l_bidx_u32[0] ); break; + case 3: u4_group3banks( &l_bidx_u32[0] ); break; + case 2: u4_group2banks( &l_bidx_u32[0] ); break; + case 1: u4_group1banks( &l_bidx_u32[0] ); break; + } + + /* + * sort DIMM groups by size (descending) + */ + u4_sortDIMMgroups(); + + /* + * calculate overall memory size in bytes + * (group size is in MB) + */ + m_memsize_u64 = 0; + for( i = 0; i < m_dgrcnt_u32; i++ ) { + m_memsize_u64 += (uint64_t) m_dgrptr[i]->m_size_u32 * _1MB; + } + +} + +static int32_t +u4_setup_core_clock( void ) +{ + static const uint32_t MCLK = (uint32_t) 266; + static const uint32_t CDIV = (uint32_t) 66; + static const uint32_t CMAX = (uint32_t) 7; + static const uint32_t MERR = (uint32_t) 10; + uint32_t volatile l_cclk_u32; + uint32_t volatile l_pll2_u32; + uint32_t i, s; + + #ifdef U4_INFO + printf( " [core clock reset: ]" ); + #endif + + /* + * calculate speed value + */ + s = m_gendimm.m_speed_pu32[m_dclidx_u32]; + s -= MCLK; + s /= CDIV; + + /* + * insert new core clock value + */ + l_cclk_u32 = load32_ci( ClkCntl_R ); + l_cclk_u32 &= ~CLK_DDR_CLK_MSK; + l_cclk_u32 |= ( s << 18 ); + + + // return on error + if( s > CMAX ) { + #ifdef U4_INFO + printf( "\b\b\b\bERR\r\n" ); + #endif + return RET_ERR; + } + + /* + * reset core clock + */ + store32_ci( ClkCntl_R, l_cclk_u32 ); + dly( 0x1000000 ); + or32_ci( PLL2Cntl_R, IBIT(0) ); + dly( 0x1000000 ); + + /* + * wait for reset to finish + */ + do { + l_pll2_u32 = load32_ci( PLL2Cntl_R ); + } while( ( l_pll2_u32 & IBIT(0) ) != 0 ); + + /* + * wait for stable PLL + */ + s = 0; + do { + l_pll2_u32 = ( load32_ci( PLL2Cntl_R ) & IBIT(2) ); + + for( i = 0; i < 4; i++ ) { + l_pll2_u32 &= ( load32_ci( PLL2Cntl_R ) & IBIT(2) ); + l_pll2_u32 &= ( load32_ci( PLL2Cntl_R ) & IBIT(2) ); + l_pll2_u32 &= ( load32_ci( PLL2Cntl_R ) & IBIT(2) ); + dly( 0x10000 ); + } + + } while( ( l_pll2_u32 == 0 ) && ( s++ < MERR ) ); + + if( s >= MERR ) { + #ifdef U4_INFO + printf( "\b\b\b\bERR\r\n" ); + #endif + return RET_ERR; + } + + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + #endif + + return RET_OK; +} + +static void +u4_auto_calib_init( void ) +{ + static const uint32_t SEQ[] = { + 0xb1000000, 0xd1000000, 0xd1000000, 0xd1000000, + 0xd1000000, 0xd1000000, 0xd1000000, 0xd1000000, + 0xd1000000, 0xd1000000, 0xd1000000, 0xd1000000, + 0xd1000000, 0xd1000000, 0xd1000400, 0x00000000, + }; + + uint64_t i; + uint32_t j; + + for( i = MemInit00_R, j = 0; i <= MemInit15_R; i += 0x10, j++ ) { + store32_ci( i, SEQ[j] ); + } + +} + +#if 0 +static uint32_t +u4_RSL_BLane( uint32_t f_Rank_u32, uint32_t f_BLane_u32 ) +{ + static const uint32_t MemProgCntl_V = (uint32_t) 0x80000500; + static const uint32_t CalConf0_V = (uint32_t) 0x0000aa10; + uint32_t l_MemProgCntl_u32; + uint32_t l_CalConf0_u32; + uint32_t l_MeasStat_u32; + uint32_t l_CalC_u32; + uint64_t MeasStat_R; + uint64_t CalC_R; + uint64_t VerC_R; + uint32_t shft; + uint32_t v; + + if( f_BLane_u32 < 4 ) { + MeasStat_R = MeasStatusC0_R; + CalC_R = CalC0_R; + VerC_R = RstLdEnVerniersC0_R; + } else if( f_BLane_u32 < 8 ) { + f_BLane_u32 -= 4; + MeasStat_R = MeasStatusC1_R; + CalC_R = CalC1_R; + VerC_R = RstLdEnVerniersC1_R; + } else if( f_BLane_u32 < 12 ) { + f_BLane_u32 -= 8; + MeasStat_R = MeasStatusC2_R; + CalC_R = CalC2_R; + VerC_R = RstLdEnVerniersC2_R; + } else if( f_BLane_u32 == 16 ) { + f_BLane_u32 = 4; + MeasStat_R = MeasStatusC1_R; + CalC_R = CalC1_R; + VerC_R = RstLdEnVerniersC1_R; + } else if( f_BLane_u32 == 17 ) { + f_BLane_u32 = 4; + MeasStat_R = MeasStatusC3_R; + CalC_R = CalC3_R; + VerC_R = RstLdEnVerniersC3_R; + } else { + f_BLane_u32 -= 12; + MeasStat_R = MeasStatusC3_R; + CalC_R = CalC3_R; + VerC_R = RstLdEnVerniersC3_R; + } + + shft = (uint32_t) 28 - ( f_BLane_u32 * 4 ); + + /* + * start auto calibration logic & wait for completion + */ + or32_ci( MeasStat_R, IBIT(0) ); + + do { + l_MeasStat_u32 = load32_ci( MeasStat_R ); + } while( ( l_MeasStat_u32 & IBIT(0) ) == 1 ); + + l_CalConf0_u32 = CalConf0_V; + store32_ci( CalConf0_R, l_CalConf0_u32 ); + + for( v = 0x000; v < (uint32_t) 0x100; v++ ) { + store32_ci( VerC_R, ( v << 24 ) | ( v << 16 ) ); + + l_MemProgCntl_u32 = MemProgCntl_V; + l_MemProgCntl_u32 |= + ( (uint32_t) 0x00800000 >> f_Rank_u32 ); + store32_ci( MemProgCntl_R, l_MemProgCntl_u32 ); + + do { + l_MemProgCntl_u32 = load32_ci( MemProgCntl_R ); + } while( ( l_MemProgCntl_u32 & IBIT(1) ) == 0 ); + + l_CalC_u32 = ( ( load32_ci( CalC_R ) >> shft ) & + (uint32_t) 0xf ); + + if( l_CalC_u32 != (uint32_t) 0xa ) { + v--; + break; + } + + } + + if( v == (uint32_t) 0x100 ) { + v = (uint32_t) ~1; + } + + return v; +} +#endif + +static uint32_t +u4_RMDF_BLane( uint32_t f_Rank_u32, uint32_t f_BLane_u32 ) +{ + static const uint32_t MemProgCntl_V = (uint32_t) 0x80000f00; + static const uint32_t CalConf0_V = (uint32_t) 0x0000ac10; + uint32_t l_MemProgCntl_u32; + uint32_t l_CalConf0_u32; + uint32_t l_MeasStat_u32; + uint32_t l_CalC_u32; + uint64_t MeasStat_R; + uint64_t CalC_R; + uint64_t VerC_R; + uint32_t shft; + uint32_t v; + + if( f_BLane_u32 < 4 ) { + MeasStat_R = MeasStatusC0_R; + CalC_R = CalC0_R; + VerC_R = RstLdEnVerniersC0_R; + } else if( f_BLane_u32 < 8 ) { + f_BLane_u32 -= 4; + MeasStat_R = MeasStatusC1_R; + CalC_R = CalC1_R; + VerC_R = RstLdEnVerniersC1_R; + } else if( f_BLane_u32 < 12 ) { + f_BLane_u32 -= 8; + MeasStat_R = MeasStatusC2_R; + CalC_R = CalC2_R; + VerC_R = RstLdEnVerniersC2_R; + } else if( f_BLane_u32 == 16 ) { + f_BLane_u32 = 4; + MeasStat_R = MeasStatusC1_R; + CalC_R = CalC1_R; + VerC_R = RstLdEnVerniersC1_R; + } else if( f_BLane_u32 == 17 ) { + f_BLane_u32 = 4; + MeasStat_R = MeasStatusC3_R; + CalC_R = CalC3_R; + VerC_R = RstLdEnVerniersC3_R; + } else { + f_BLane_u32 -= 12; + MeasStat_R = MeasStatusC3_R; + CalC_R = CalC3_R; + VerC_R = RstLdEnVerniersC3_R; + } + + shft = (uint32_t) 28 - ( f_BLane_u32 * 4 ); + + /* + * start auto calibration logic & wait for completion + */ + or32_ci( MeasStat_R, IBIT(0) ); + + do { + l_MeasStat_u32 = load32_ci( MeasStat_R ); + } while( ( l_MeasStat_u32 & IBIT(0) ) == 1 ); + + l_CalConf0_u32 = CalConf0_V; + l_CalConf0_u32 |= ( f_BLane_u32 << 5 ); + store32_ci( CalConf0_R, l_CalConf0_u32 ); + + for( v = 0x000; v < (uint32_t) 0x100; v++ ) { + store32_ci( VerC_R, ( v << 24 ) | ( v << 16 ) ); + + l_MemProgCntl_u32 = MemProgCntl_V; + l_MemProgCntl_u32 |= + ( (uint32_t) 0x00800000 >> f_Rank_u32 ); + store32_ci( MemProgCntl_R, l_MemProgCntl_u32 ); + + do { + l_MemProgCntl_u32 = load32_ci( MemProgCntl_R ); + } while( ( l_MemProgCntl_u32 & IBIT(1) ) == 0 ); + + l_CalC_u32 = ( ( load32_ci( CalC_R ) >> shft ) & + (uint32_t) 0xf ); + + if( l_CalC_u32 != (uint32_t) 0xa ) { + v--; + break; + } + + } + + if( v == (uint32_t) 0x100 ) { + v = (uint32_t) ~1; + } + + return v; +} + +static int32_t +u4_RMDF_Rank( uint32_t f_Rank_u32, + uint32_t *f_Buf_pu32 ) +{ + int32_t l_Err_pi32 = 0; + uint32_t b; + + for( b = 0; ( b < MAX_BLANE ) && ( l_Err_pi32 == 0 ); b++ ) { + f_Buf_pu32[b] = u4_RMDF_BLane( f_Rank_u32, b ); + + if( f_Buf_pu32[b] == (uint32_t) ~0 ) { + f_Buf_pu32[b] = 0; + l_Err_pi32++; + } else if( f_Buf_pu32[b] == (uint32_t) ~1 ) { + f_Buf_pu32[b] = (uint32_t) 0xff; + l_Err_pi32++; + } + + } + + return l_Err_pi32; +} + +static int32_t +u4_auto_calib_MemBus( auto_calib_t *f_ac_pt ) +{ + uint32_t RdMacDly, RdMacCnt; + uint32_t ResMuxDly, ResMuxCnt; + uint32_t RdPipeDly; + uint32_t l_Buf_pu32[MAX_DRANKS][MAX_BLANE]; + uint32_t l_Rnk_pu32[MAX_DRANKS]; + uint32_t l_Ver_u32; + int32_t l_Err_i32; + uint32_t bidx; + uint32_t n, r, b; + + /* + * read starting delays out of the MemBus register + */ + RdMacDly = ( load32_ci( MemBusCnfg_R ) >> 28 ) & 0xf; + ResMuxDly = ( load32_ci( MemBusCnfg_R ) >> 24 ) & 0xf; + + /* + * initialize ranks as not populated + */ + for( r = 0; r < MAX_DRANKS; r++ ) { + l_Rnk_pu32[r] = 0; + } + + /* + * run through every possible delays of + * RdMacDly, ResMuxDly & RdPipeDly until + * the first working configuration is found + */ + RdPipeDly = 0; + do { + and32_ci( MemBusCnfg2_R, ~0x3 ); + or32_ci( MemBusCnfg2_R, RdPipeDly ); + + RdMacCnt = RdMacDly; + ResMuxCnt = ResMuxDly; + + /* + * RdMacDly >= ResMuxDly + */ + do { + and32_ci( MemBusCnfg_R, ( 1 << 24 ) - 1 ); + or32_ci( MemBusCnfg_R, ( RdMacCnt << 28 ) | + ( ResMuxCnt << 24 ) ); + and32_ci( MemBusCnfg2_R, ( 1 << 28 ) - 1 ); + or32_ci( MemBusCnfg2_R, ( RdMacCnt << 28 ) ); + + /* + * check the current value for every installed + * DIMM on each side for every bytelane + */ + l_Err_i32 = 0; + for( n = 0; + ( n < NUM_SLOTS ) && + ( l_Err_i32 == 0 ); + n += 2 ) { + + if( m_dimm[n].m_pop_u32 ) { + /* + * run through all 18 bytelanes of every rank + */ + for( r = n; + ( r < n + m_dimm[n].m_rank_u32 ) && + ( l_Err_i32 == 0 ); + r++ ) { + l_Rnk_pu32[r] = 1; + + l_Err_i32 = + u4_RMDF_Rank( r, + &l_Buf_pu32[r][0] ); + } + + } + + } + + /* + * decrementation before exit is wanted! + */ + RdMacCnt--; + ResMuxCnt--; + } while( ( ResMuxCnt > 0 ) && + ( l_Err_i32 != 0 ) ); + + if( l_Err_i32 != 0 ) { + RdPipeDly++; + } + + } while( ( RdPipeDly < 4 ) && + ( l_Err_i32 != 0 ) ); + + /* + * if l_Err_pi32 == 0 the auto calibration passed ok + */ + if( l_Err_i32 != 0 ) { + return RET_ERR; + } + + /* + * insert delay values into return struct + */ + and32_ci( MemBusCnfg_R, ( 1 << 24 ) - 1 ); + or32_ci( MemBusCnfg_R, ( RdMacCnt << 28 ) | + ( ResMuxCnt << 24 ) ); + and32_ci( MemBusCnfg2_R, ( ( 1 << 28 ) - 1 ) & ~0x3 ); + or32_ci( MemBusCnfg2_R, ( RdMacCnt << 28 ) | RdPipeDly ); + + f_ac_pt->m_MemBusCnfg_u32 = load32_ci( MemBusCnfg_R ); + f_ac_pt->m_MemBusCnfg2_u32 = load32_ci( MemBusCnfg2_R ); + + /* + * calculate the average vernier setting for the + * bytelanes which share one vernier + */ + for( b = 0; b < MAX_BLANE - 2; b += 2 ) { + n = 0; + l_Ver_u32 = 0; + + for( r = 0; r < MAX_DRANKS; r++ ) { + /* + * calculation is done or populated ranks only + */ + if( l_Rnk_pu32[r] != 0 ) { + /* + * calculate average value + */ + l_Ver_u32 += l_Buf_pu32[r][b]; + l_Ver_u32 += l_Buf_pu32[r][b+1]; + n += 2; + + if( b == 4 ) { + l_Ver_u32 += l_Buf_pu32[r][16]; + n++; + } else if( b == 12 ) { + l_Ver_u32 += l_Buf_pu32[r][17]; + n++; + } + + } + + } + + /* + * average the values + */ + l_Ver_u32 /= n; + + /* + * set appropiate vernier register for + * the current bytelane + */ + bidx = ( b >> 2 ); + if( ( b & (uint32_t) 0x3 ) == 0 ) { + l_Ver_u32 <<= 24; + f_ac_pt->m_RstLdEnVerniers_pu32[bidx] = l_Ver_u32; + } else { + l_Ver_u32 <<= 16; + f_ac_pt->m_RstLdEnVerniers_pu32[bidx] |= l_Ver_u32; + } + + } + + return RET_OK; +} + +static int32_t +u4_auto_calib( auto_calib_t *f_ac_pt ) +{ + uint32_t l_MemBusCnfg_S; + uint32_t l_MemBusCnfg2_S; + uint32_t l_RstLdEnVerniers_S[4]; + int32_t l_Ret_i32; + + /* + * save manipulated registers + */ + l_MemBusCnfg_S = load32_ci( MemBusCnfg_R ); + l_MemBusCnfg2_S = load32_ci( MemBusCnfg2_R ); + l_RstLdEnVerniers_S[0] = load32_ci( RstLdEnVerniersC0_R ); + l_RstLdEnVerniers_S[1] = load32_ci( RstLdEnVerniersC1_R ); + l_RstLdEnVerniers_S[2] = load32_ci( RstLdEnVerniersC2_R ); + l_RstLdEnVerniers_S[3] = load32_ci( RstLdEnVerniersC3_R ); + + u4_auto_calib_init(); + l_Ret_i32 = u4_auto_calib_MemBus( f_ac_pt ); + + /* + * restore manipulated registers + */ + store32_ci( MemBusCnfg_R, l_MemBusCnfg_S ); + store32_ci( MemBusCnfg2_R, l_MemBusCnfg2_S ); + store32_ci( RstLdEnVerniersC0_R, l_RstLdEnVerniers_S[0] ); + store32_ci( RstLdEnVerniersC1_R, l_RstLdEnVerniers_S[1] ); + store32_ci( RstLdEnVerniersC2_R, l_RstLdEnVerniers_S[2] ); + store32_ci( RstLdEnVerniersC3_R, l_RstLdEnVerniers_S[3] ); + + return l_Ret_i32; +} + +static int32_t +u4_checkeccerr( eccerror_t *f_ecc_pt ) +{ + uint32_t l_val_u32; + int32_t ret = RET_OK; + + l_val_u32 = load32_ci( MESR_R ); + l_val_u32 >>= 29; + + if( ( l_val_u32 & (uint32_t) 0x7 ) != 0 ) { + + if( ( l_val_u32 & (uint32_t) 0x4 ) != 0 ) { + /* UE */ + ret = RET_ACERR_UE; + } else if( ( l_val_u32 & (uint32_t) 0x1 ) != 0 ) { + /* UEWT */ + ret = RET_ACERR_UEWT; + } else { + /* CE */ + ret = RET_ACERR_CE; + } + + } + + f_ecc_pt->m_err_i32 = ret; + + l_val_u32 = load32_ci( MEAR1_R ); + f_ecc_pt->m_uecnt_u32 = ( ( l_val_u32 >> 24 ) & (uint32_t) 0xff ); + f_ecc_pt->m_cecnt_u32 = ( ( l_val_u32 >> 16 ) & (uint32_t) 0xff ); + + l_val_u32 = load32_ci( MEAR0_R ); + f_ecc_pt->m_rank_u32 = ( ( l_val_u32 >> 29 ) & (uint32_t) 0x7 ); + f_ecc_pt->m_col_u32 = ( ( l_val_u32 >> 18 ) & (uint32_t) 0x7ff ); + f_ecc_pt->m_row_u32 = ( ( l_val_u32 >> 0 ) & (uint32_t) 0x7fff ); + f_ecc_pt->m_bank_u32 = ( ( l_val_u32 >> 15 ) & (uint32_t) 0x7 ); + + return ret; +} + +static uint32_t +u4_CalcScrubEnd( void ) +{ + uint64_t l_scrend_u64 = m_memsize_u64; + + /* + * check for memory hole at 2GB + */ + if( l_scrend_u64 > _2GB ) { + l_scrend_u64 += _2GB; + } + + l_scrend_u64 -= 0x40; + l_scrend_u64 /= 0x10; + + return( (uint32_t) l_scrend_u64 ); +} + +static int32_t +u4_Scrub( uint32_t f_scrub_u32, uint32_t f_pattern_u32, eccerror_t *f_eccerr_pt ) +{ + uint32_t i; + int32_t ret; + + /* + * setup scrub parameters + */ + store32_ci( MSCR_R, 0 ); // stop scrub + store32_ci( MSRSR_R, 0x0 ); // set start + store32_ci( MSRER_R, u4_CalcScrubEnd() ); // set end + store32_ci( MSPR_R, f_pattern_u32 ); // set pattern + + /* + * clear out ECC error registers + */ + store32_ci( MEAR0_R, 0x0 ); + store32_ci( MEAR1_R, 0x0 ); + store32_ci( MESR_R, 0x0 ); + + /* + * Setup Scrub Type + */ + store32_ci( MSCR_R, f_scrub_u32 ); + + if( f_scrub_u32 != BACKGROUND_SCRUB ) { + /* + * wait for scrub to complete + */ + do { + progbar(); + dly( 15000000 ); + i = load32_ci( MSCR_R ); + } while( ( i & f_scrub_u32 ) != 0 ); + + ret = u4_checkeccerr( f_eccerr_pt ); + } else { + ret = RET_OK; + } + + return ret; +} + +static eccerror_t +u4_InitialScrub( void ) +{ + eccerror_t l_eccerr_st[2]; + int32_t l_err_i32[2] = { 0, 0 }; + + l_err_i32[0] = u4_Scrub( IMMEDIATE_SCRUB_WITH_FILL, 0x0, &l_eccerr_st[0] ); + + if( l_err_i32[0] >= -1 /*CE*/ ) { + l_err_i32[1] = u4_Scrub( IMMEDIATE_SCRUB, 0x0, &l_eccerr_st[1] ); + } + + if( l_err_i32[0] < l_err_i32[1] ) { + return l_eccerr_st[0]; + } else { + return l_eccerr_st[1]; + } + +} + +/* + * RND: calculates Timer cycles from the given frequency + * devided by the clock frequency. Values are rounded + * up to the nearest integer value if the division is not even. + */ +#define RND( tXXX ) ( ( ( tXXX ) + tCK - 1 ) / tCK ) + +static void +u4_MemInitSequence( uint32_t tRP, uint32_t tWR, uint32_t tRFC, uint32_t CL, + uint32_t tCK, uint32_t TD ) +{ + /* + * DIMM init sequence + */ + static const uint32_t INI_SEQ[] = { + 0xa0000400, 0x80020000, 0x80030000, 0x80010404, + 0x8000100a, 0xa0000400, 0x90000000, 0x90000000, + 0x8ff0100a, 0x80010784, 0x80010404, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000 + }; + + uint32_t l_MemInit_u32; + uint64_t r; + uint32_t i; + + for( r = MemInit00_R, i = 0; r <= MemInit15_R; r += 0x10, i++ ) { + l_MemInit_u32 = INI_SEQ[i]; + + switch( i ) { + case 0: + case 5: { + l_MemInit_u32 |= ( ( RND( tRP ) - TD ) << 20 ); + break; + } + case 3: { + store32_ci( EMRSRegCntl_R, l_MemInit_u32 & + (uint32_t) 0xffff ); + break; + } + case 4: { + l_MemInit_u32 |= IBIT(23); + } + case 8: { + l_MemInit_u32 |= ( ( RND( tWR ) - 1 ) << 9 ); + l_MemInit_u32 |= ( CL << 4 ); + + store32_ci( MRSRegCntl_R, l_MemInit_u32 & + (uint32_t) 0xffff ); + break; + } + case 6: + case 7: { + l_MemInit_u32 |= ( ( RND( tRFC ) - TD ) << 20 ); + break; + } + + } + + store32_ci( r, l_MemInit_u32 ); + +#ifdef U4_SHOW_REGS + printf( "\r\nMemInit%02d (0x%04X): 0x%08X", i, (uint16_t) r, l_MemInit_u32 ); +#endif + } +#ifdef U4_SHOW_REGS + printf( "\r\n" ); +#endif + /* + * Kick off memory init sequence & wait for completion + */ + store32_ci( MemProgCntl_R, IBIT(0) ); + + do { + i = load32_ci( MemProgCntl_R ); + } while( ( i & IBIT(1) ) == 0 ); + +} + +/* + * static DIMM configuartion settings + */ +static reg_statics_t reg_statics_maui[NUM_SPEED_IDX] = { + { /* 400 Mhz */ + .RRMux = 1, + .WRMux = 1, + .WWMux = 1, + .RWMux = 1, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x60000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x34000000, + .CKDelayU = 0x34000000, + + .MemBusCnfg = 0x00000050 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 2 ) << 24 ) ), + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 533 Mhz */ + .RRMux = 1, + .WRMux = 1, + .WWMux = 1, + .RWMux = 1, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x60000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x18000000, + .CKDelayU = 0x18000000, + + .MemBusCnfg = 0x00002070 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 3 ) << 24 ) ), + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + + 0x12000000, 0x12000000, 0x12000000 , 0x12000000, + 0x12000000, 0x12000000, 0x12000000 , 0x12000000, + 0x12000000, 0x12000000, 0x12000000 , 0x12000000, + 0x12000000, 0x12000000, 0x12000000 , 0x12000000, + 0x12000000, 0x12000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000 , 0x00000000, + 0x00000000, 0x00000000, 0x00000000 , 0x00000000, + 0x00000000, 0x00000000, 0x00000000 , 0x00000000, + 0x00000000, 0x00000000, 0x00000000 , 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 667 Mhz */ + .RRMux = 1, + .WRMux = 1, + .WWMux = 1, + .RWMux = 3, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x60000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x0a000000, + .CKDelayU = 0x0a000000, + + .MemBusCnfg = 0x000040a0 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 3 ) << 24 ) ), + + .CAS1Dly0 = 2, + .CAS1Dly1 = 2, + + .ByteWrClkDel = { + + 0x12000000, 0x12000000, 0x12000000, 0x12000000, + 0x12000000, 0x12000000, 0x12000000, 0x12000000, + 0x12000000, 0x12000000, 0x12000000, 0x12000000, + 0x12000000, 0x12000000, 0x12000000, 0x12000000, + 0x12000000, 0x12000000 +/* + 0x31000000, 0x31000000, 0x31000000, 0x31000000, + 0x31000000, 0x31000000, 0x31000000, 0x31000000, + 0x31000000, 0x31000000, 0x31000000, 0x31000000, + 0x31000000, 0x31000000, 0x31000000, 0x31000000, + 0x31000000, 0x31000000 +*/ + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + } +}; + +static reg_statics_t reg_statics_bimini[NUM_SPEED_IDX] = { + { /* 400 Mhz */ + .RRMux = 2, + .WRMux = 2, + .WWMux = 2, + .RWMux = 2, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x40000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x00000000, + .CKDelayU = 0x28000000, + + .MemBusCnfg = 0x00552070 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 2 ) << 24 ) ), + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 533 Mhz */ + .RRMux = 3, + .WRMux = 3, + .WWMux = 3, + .RWMux = 3, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x40000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x00000000, + .CKDelayU = 0x20000000, + + .MemBusCnfg = 0x00644190 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 3 ) << 24 ) ), + + .CAS1Dly0 = 2, + .CAS1Dly1 = 2, + + .ByteWrClkDel = { + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 667 Mhz */ + .RRMux = 3, + .WRMux = 3, + .WWMux = 3, + .RWMux = 3, + + .MemRdQCnfg = 0x20020820, + .MemWrQCnfg = 0x40041040, + .MemQArb = 0x00000000, + .MemRWArb = 0x30413cc0, + + .ODTCntl = 0x40000000, + .IOPadCntl = 0x001a4000, + .MemPhyModeCntl = 0x00000000, + .OCDCalCntl = 0x00000000, + .OCDCalCmd = 0x00000000, + + .CKDelayL = 0x00000000, + .CKDelayU = 0x00000000, + + .MemBusCnfg = 0x00666270 | + ( ( MAX_RMD << 28 ) | + ( ( MAX_RMD - 3 ) << 24 ) ), + + .CAS1Dly0 = 2, + .CAS1Dly1 = 2, + + .ByteWrClkDel = { + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000, 0x14000000, 0x14000000, + 0x14000000, 0x14000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + } +}; + +static reg_statics_t reg_statics_kauai[NUM_SPEED_IDX] = { + { /* 400 Mhz */ + .RRMux = 0, + .WRMux = 0, + .WWMux = 0, + .RWMux = 0, + + .MemRdQCnfg = 0, + .MemWrQCnfg = 0, + .MemQArb = 0, + .MemRWArb = 0, + + .ODTCntl = 0, + .IOPadCntl = 0, + .MemPhyModeCntl = 0, + .OCDCalCntl = 0, + .OCDCalCmd = 0, + + .CKDelayL = 0, + .CKDelayU = 0, + + .MemBusCnfg = 0, + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 533 Mhz */ + .RRMux = 0, + .WRMux = 0, + .WWMux = 0, + .RWMux = 0, + + .MemRdQCnfg = 0, + .MemWrQCnfg = 0, + .MemQArb = 0, + .MemRWArb = 0, + + .ODTCntl = 0, + .IOPadCntl = 0, + .MemPhyModeCntl = 0, + .OCDCalCntl = 0, + .OCDCalCmd = 0, + + .CKDelayL = 0, + .CKDelayU = 0, + + .MemBusCnfg = 0, + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + }, + { /* 667 Mhz */ + .RRMux = 0, + .WRMux = 0, + .WWMux = 0, + .RWMux = 0, + + .MemRdQCnfg = 0, + .MemWrQCnfg = 0, + .MemQArb = 0, + .MemRWArb = 0, + + .ODTCntl = 0, + .IOPadCntl = 0, + .MemPhyModeCntl = 0, + .OCDCalCntl = 0, + .OCDCalCmd = 0, + + .CKDelayL = 0, + .CKDelayU = 0, + + .MemBusCnfg = 0, + + .CAS1Dly0 = 0, + .CAS1Dly1 = 0, + + .ByteWrClkDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + }, + .ReadStrobeDel = { + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000, 0x00000000, 0x00000000, + 0x00000000, 0x00000000 + } + + } +}; + +static int32_t +u4_start( eccerror_t *f_ecc_pt ) +{ + /* + * maximum runs for auto calibration + */ + static const uint32_t MAX_ACERR = (uint32_t) 5; + + /* + * fixed u4/DIMM timer/timing values for calculation + */ + static const uint32_t TD = (uint32_t) 2; // u4 delay cycles for loading a timer + static const uint32_t AL = (uint32_t) 0; // additional latency (fix) + static const uint32_t BL = (uint32_t) 4; // burst length (fix) + + uint32_t SPEED = m_gendimm.m_speed_pu32[m_dclidx_u32]; + uint32_t CL = m_gendimm.m_clval_pu32[m_dclidx_u32]; + uint32_t RL = AL + CL; + uint32_t WL = RL - 1; + uint32_t tCK = m_gendimm.m_tCK_pu32[m_dclidx_u32]; + uint32_t tRAS = m_gendimm.m_tRAS_u32; + uint32_t tRTP = m_gendimm.m_tRTP_u32; + uint32_t tRP = m_gendimm.m_tRP_u32; + uint32_t tWR = m_gendimm.m_tWR_u32; + uint32_t tRRD = m_gendimm.m_tRRD_u32; + uint32_t tRC = m_gendimm.m_tRC_u32; + uint32_t tRCD = m_gendimm.m_tRCD_u32; + uint32_t tWTR = m_gendimm.m_tWTR_u32; + uint32_t tRFC = m_gendimm.m_tRFC_u32; + uint32_t tREF = m_gendimm.m_tREF_u32; + + reg_statics_t *rst; + + uint32_t l_RAS0_u32; + uint32_t l_RAS1_u32; + uint32_t l_CAS0_u32; + uint32_t l_CAS1_u32; + uint32_t l_MemRfshCntl_u32; + uint32_t l_UsrCnfg_u32; + uint32_t l_DmCnfg_u32; + + uint32_t l_MemArbWt_u32; + uint32_t l_MemRWArb_u32; + uint32_t l_MemBusCnfg_u32; + + auto_calib_t l_ac_st; + int32_t l_ac_i32; + uint32_t l_acerr_i32; + uint32_t sidx; + uint32_t i, j, t0, t1; + + /* + * set index for different 400/533/667 Mhz setup + */ + switch( SPEED ) { + case 400: + case 533: + case 667: { + sidx = SPEED; + sidx -= 400; + sidx /= 133; + break; + } + + default: { + #ifdef U4_DEBUG2 + printf( "\r\n-> DIMM speed of %03u not supported\r\n", + m_gendimm.m_speed_pu32[m_dclidx_u32] ); + #endif + return RET_ERR; + } + + } + + /* + * setup pointer to the static register settings + */ + if( IS_MAUI ) { + rst = ®_statics_maui[sidx]; + } else if( IS_BIMINI ) { + rst = ®_statics_bimini[sidx]; + } else if( IS_KAUAI ) { + rst = ®_statics_kauai[sidx]; + } + + /* + * Switch off Fast Path by default for all DIMMs + * running with more than 400Mhz + */ + if( SPEED == 400 ) { + or32_ci( APIMemRdCfg_R, IBIT(30) ); + #ifdef U4_INFO + printf( " [fastpath : ON]\r\n" ); + #endif + } else { + and32_ci( APIMemRdCfg_R, ~IBIT(30) ); + #ifdef U4_INFO + printf( " [fastpath : OFF]\r\n" ); + #endif + } + + + #ifdef U4_INFO + printf( " [register setup : ]" ); + #endif + + /* + * setup RAS/CAS timers2 + * NOTE: subtract TD from all values because of the delay + * caused by reloading timers (see spec) + */ + + /* + * RAS Timer 0 + */ + // TiAtP = RND(tRAS) -> RAS0[0:4] + l_RAS0_u32 = ( ( RND( tRAS ) - TD ) << 27 ); + // TiRtP = AL + BL/2 - 2 + RND(tRTP) -> RAS01[5:9] + l_RAS0_u32 |= ( ( AL + BL/2 - 2 + RND( tRTP ) - TD ) << 22 ); + // TiWtP = WL + BL/2 + RND(tWR) -> RAS0[10:14] + l_RAS0_u32 |= ( ( WL + BL/2 + RND( tWR ) - TD ) << 17 ); + // TiPtA = RND(tRP) -> RAS0[15:19] + l_RAS0_u32 |= ( ( RND( tRP ) - TD ) << 12 ); + // TiPAtA = RND(tRP) or + // RND(tRP) + 1 for 8 bank devices -> RAS0[20:24] + if( m_gendimm.m_bankcnt_u32 <= 4 ) { + l_RAS0_u32 |= ( ( RND( tRP ) - TD ) << 7 ); + } else { + l_RAS0_u32 |= ( ( RND( tRP ) + 1 - TD ) << 7 ); + } + + /* + * RAS Timer 1 + */ + // TiRAPtA = AL + BL/2 - 2 + RND(tRTP + tRP) -> RAS1[0:4] + l_RAS1_u32 = ( ( AL + BL/2 - 2 + RND( tRTP + tRP ) - TD ) << 27 ); + // TiWAPtA = CL + AL + BL/2 - 1 + RND(tWR + tRP) -> RAS1[5:9] + l_RAS1_u32 |= ( ( CL + AL + BL/2 - 1 + RND( tWR + tRP ) - TD ) << 22 ); + // TiAtARk = tRRD -> RAS1[10:14] + l_RAS1_u32 |= ( ( RND( tRRD ) - TD ) << 17 ); + // TiAtABk = tRC -> RAS1[15:19] + l_RAS1_u32 |= ( ( RND( tRC ) - TD ) << 12 ); + // TiAtRW = tRCD -> RAS1[20:24] + l_RAS1_u32 |= ( ( RND( tRCD ) - TD ) << 7 ); + // TiSAtARk Win = 4 * tRRD + 2 -> RAS1[25:29] + l_RAS1_u32 |= ( ( RND( 4 * tRRD ) + 2 - TD ) << 2 ); + + /* + * CAS Timer 0 + */ + // TiRtRRk = BL/2 -> CAS0[0:4] + l_CAS0_u32 = ( ( BL/2 - TD ) << 27 ); + // TiRtRDm = BL/2 + 1 -> CAS0[5:9] + l_CAS0_u32 |= ( ( BL/2 + 1 - TD ) << 22 ); + // TiRtRSy = BL/2 + RRMux -> CAS0[10:14] + l_CAS0_u32 |= ( ( BL/2 + rst->RRMux - TD ) << 17 ); + // TiWtRRk = CL - 1 + BL/2 + tWTR ->CAS0[15:19] + l_CAS0_u32 |= ( ( CL - 1 + BL/2 + RND( tWTR ) - TD ) << 12 ); + // TiWtRDm = BL/2 + 1 -> CAS0[20:24] + l_CAS0_u32 |= ( ( BL/2 + 1 - TD ) << 7 ); + // TiWtRSy = BL/2 + WRMux -> CAS0[25:29] + l_CAS0_u32 |= ( ( BL/2 + rst->WRMux - TD ) << 2 ); + + /* + * CAS Timer 1 + */ + // TiWtWRk = BL/2 -> CAS1[0:4] + l_CAS1_u32 = ( ( BL/2 - TD ) << 27 ); + // TiWtWDm = BL/2 + 1 -> CAS1[5:9] + l_CAS1_u32 |= ( ( BL/2 + 1 - TD ) << 22 ); + // TiWtWSy = BL/2 + WWMux -> CAS1[10:14] + l_CAS1_u32 |= ( ( BL/2 + rst->WWMux - TD ) << 17 ); + // TiRtWRk = BL/2 + 2 -> CAS1[15:19] + l_CAS1_u32 |= ( ( BL/2 + 2 + rst->CAS1Dly0 - TD ) << 12 ); + // TiRtWDm = BL/2 + 2 -> CAS1[20:24] + l_CAS1_u32 |= ( ( BL/2 + 2 + rst->CAS1Dly1 - TD ) << 7 ); + // TiRtWSy = BL/2 + RWMux + 1 -> CAS1[25:29] + l_CAS1_u32 |= ( ( BL/2 + rst->RWMux + 1 - TD ) << 2 ); + + store32_ci( RASTimer0_R, l_RAS0_u32 ); + store32_ci( RASTimer1_R, l_RAS1_u32 ); + store32_ci( CASTimer0_R, l_CAS0_u32 ); + store32_ci( CASTimer1_R, l_CAS1_u32 ); + + /* + * Mem Refresh Control register + */ + l_MemRfshCntl_u32 = ( ( ( tREF / tCK ) / 16 ) << 23 ); + l_MemRfshCntl_u32 |= ( ( RND( tRFC ) - TD ) << 8 ); + store32_ci( MemRfshCntl_R, l_MemRfshCntl_u32 ); + + /* + * setup DmXCnfg registers + */ + store32_ci( Dm0Cnfg_R, (uint32_t) 0x0 ); + store32_ci( Dm1Cnfg_R, (uint32_t) 0x0 ); + store32_ci( Dm2Cnfg_R, (uint32_t) 0x0 ); + store32_ci( Dm3Cnfg_R, (uint32_t) 0x0 ); + + /* + * create DmCnfg & UsrCnfg values out of group data + */ + l_UsrCnfg_u32 = 0; + for( i = 0; i < m_dgrcnt_u32; i++ ) { + l_DmCnfg_u32 = ( m_dgrptr[i]->m_add2g_u32 << 27 ); + l_DmCnfg_u32 |= ( m_dgrptr[i]->m_sub2g_u32 << 19 ); + l_DmCnfg_u32 |= ( m_dgrptr[i]->m_memmd_u32 << 12 ); + l_DmCnfg_u32 |= ( m_dgrptr[i]->m_start_u32 << 3 ); + l_DmCnfg_u32 |= ( m_dgrptr[i]->m_ss_u32 << 1 ); + l_DmCnfg_u32 |= IBIT(31); // enable bit + + /* + * write value into DmXCnfg registers + */ + for( j = 0; j < m_dgrptr[i]->m_dcnt_u32; j++ ) { + t0 = m_dgrptr[i]->m_dptr[j]->m_bank_u32; + t1 = Dm0Cnfg_R + 0x10 * t0; + + if( load32_ci( t1 ) == 0 ) { + store32_ci( t1, l_DmCnfg_u32 ); + l_UsrCnfg_u32 |= + ( m_dgrptr[i]->m_csmode_u32 << ( 30 - 2 * t0 ) ); + } + + } + + } + + /* + * setup UsrCnfg register + *- cs mode is selected above + *- Interleave on L2 cache line + *- Usually closed page policy + */ + l_UsrCnfg_u32 |= IBIT(8); // interleave on L2 cache line + l_UsrCnfg_u32 &= ~IBIT(9); // usually closed + l_UsrCnfg_u32 |= IBIT(10); + store32_ci( UsrCnfg_R, l_UsrCnfg_u32 ); + + /* + * Memory Arbiter Weight Register + */ + // CohWt -> MemAWt[0:1] + l_MemArbWt_u32 = ( (uint32_t) 1 << 30 ); + // NCohWt -> MemAWt[2:3] + l_MemArbWt_u32 |= ( (uint32_t) 1 << 28 ); + // ScrbWt -> MemAWt[4:5] + l_MemArbWt_u32 |= ( (uint32_t) 0 << 26 ); + store32_ci( MemArbWt_R, l_MemArbWt_u32 ); + + /* + * misc fixed register setup + */ + store32_ci( ODTCntl_R, rst->ODTCntl ); + store32_ci( IOPadCntl_R, rst->IOPadCntl ); + store32_ci( MemPhyModeCntl_R, rst->MemPhyModeCntl ); + store32_ci( OCDCalCntl_R, rst->OCDCalCntl ); + store32_ci( OCDCalCmd_R, rst->OCDCalCmd ); + + /* + * CK Delay registers + */ + store32_ci( CKDelayL_R, rst->CKDelayL ); + store32_ci( CKDelayU_R, rst->CKDelayU ); + + /* + * read/write strobe delays + */ + store32_ci( ByteWrClkDelC0B00_R, rst->ByteWrClkDel[ 0] ); + store32_ci( ByteWrClkDelC0B01_R, rst->ByteWrClkDel[ 1] ); + store32_ci( ByteWrClkDelC0B02_R, rst->ByteWrClkDel[ 2] ); + store32_ci( ByteWrClkDelC0B03_R, rst->ByteWrClkDel[ 3] ); + store32_ci( ByteWrClkDelC0B04_R, rst->ByteWrClkDel[ 4] ); + store32_ci( ByteWrClkDelC0B05_R, rst->ByteWrClkDel[ 5] ); + store32_ci( ByteWrClkDelC0B06_R, rst->ByteWrClkDel[ 6] ); + store32_ci( ByteWrClkDelC0B07_R, rst->ByteWrClkDel[ 7] ); + store32_ci( ByteWrClkDelC0B16_R, rst->ByteWrClkDel[16] ); + store32_ci( ByteWrClkDelC0B08_R, rst->ByteWrClkDel[ 8] ); + store32_ci( ByteWrClkDelC0B09_R, rst->ByteWrClkDel[ 9] ); + store32_ci( ByteWrClkDelC0B10_R, rst->ByteWrClkDel[10] ); + store32_ci( ByteWrClkDelC0B11_R, rst->ByteWrClkDel[11] ); + store32_ci( ByteWrClkDelC0B12_R, rst->ByteWrClkDel[12] ); + store32_ci( ByteWrClkDelC0B13_R, rst->ByteWrClkDel[13] ); + store32_ci( ByteWrClkDelC0B14_R, rst->ByteWrClkDel[14] ); + store32_ci( ByteWrClkDelC0B15_R, rst->ByteWrClkDel[15] ); + store32_ci( ByteWrClkDelC0B17_R, rst->ByteWrClkDel[17] ); + store32_ci( ReadStrobeDelC0B00_R, rst->ReadStrobeDel[ 0] ); + store32_ci( ReadStrobeDelC0B01_R, rst->ReadStrobeDel[ 1] ); + store32_ci( ReadStrobeDelC0B02_R, rst->ReadStrobeDel[ 2] ); + store32_ci( ReadStrobeDelC0B03_R, rst->ReadStrobeDel[ 3] ); + store32_ci( ReadStrobeDelC0B04_R, rst->ReadStrobeDel[ 4] ); + store32_ci( ReadStrobeDelC0B05_R, rst->ReadStrobeDel[ 5] ); + store32_ci( ReadStrobeDelC0B06_R, rst->ReadStrobeDel[ 6] ); + store32_ci( ReadStrobeDelC0B07_R, rst->ReadStrobeDel[ 7] ); + store32_ci( ReadStrobeDelC0B16_R, rst->ReadStrobeDel[16] ); + store32_ci( ReadStrobeDelC0B08_R, rst->ReadStrobeDel[ 8] ); + store32_ci( ReadStrobeDelC0B09_R, rst->ReadStrobeDel[ 9] ); + store32_ci( ReadStrobeDelC0B10_R, rst->ReadStrobeDel[10] ); + store32_ci( ReadStrobeDelC0B11_R, rst->ReadStrobeDel[11] ); + store32_ci( ReadStrobeDelC0B12_R, rst->ReadStrobeDel[12] ); + store32_ci( ReadStrobeDelC0B13_R, rst->ReadStrobeDel[13] ); + store32_ci( ReadStrobeDelC0B14_R, rst->ReadStrobeDel[14] ); + store32_ci( ReadStrobeDelC0B15_R, rst->ReadStrobeDel[15] ); + store32_ci( ReadStrobeDelC0B17_R, rst->ReadStrobeDel[17] ); + + /* + * Mem Bus Configuration + * initial setup used in auto calibration + * final values will be written after + * auto calibration has finished + */ + l_MemBusCnfg_u32 = rst->MemBusCnfg; + +/* values calculation has been dropped, static values are used instead + // WdbRqDly = 2 * (CL - 3) (registered DIMMs) -> MBC[16:19] + l_MemBusCnfg_u32 += ( ( 2 * ( CL - 3 ) ) << 12 ); + // RdOEOnDly = 0 (typically) + l_MemBusCnfg_u32 += ( ( 0 ) << 8 ); + // RdOEOffDly = (2 * CL) - 4 -> MBC[24:27] + // NOTE: formula is not working, changed to: + // RdOEOffDly = (2 * CL) - 1 + l_MemBusCnfg_u32 += ( ( ( 2 * CL ) - 1 ) << 4 ); +*/ + + store32_ci( MemBusCnfg_R, l_MemBusCnfg_u32 ); + store32_ci( MemBusCnfg2_R, rst->MemBusCnfg & (uint32_t) 0xf0000000 ); + + /* + * reset verniers registers + */ + store32_ci( RstLdEnVerniersC0_R, 0x0 ); + store32_ci( RstLdEnVerniersC1_R, 0x0 ); + store32_ci( RstLdEnVerniersC2_R, 0x0 ); + store32_ci( RstLdEnVerniersC3_R, 0x0 ); + store32_ci( ExtMuxVernier0_R, 0x0 ); + store32_ci( ExtMuxVernier1_R, 0x0 ); + + /* + * Queue Configuration + */ + store32_ci( MemRdQCnfg_R, rst->MemRdQCnfg ); + store32_ci( MemWrQCnfg_R, rst->MemWrQCnfg ); + store32_ci( MemQArb_R, rst->MemQArb ); + store32_ci( MemRWArb_R, rst->MemRWArb ); + + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + #endif + + /* + * start up clocks & wait for pll2 to stabilize + */ + #ifdef U4_INFO + printf( " [start DDR clock : ]" ); + #endif + + store32_ci( MemModeCntl_R, IBIT(0) | IBIT(8) ); + dly( 50000000 ); + + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + + #endif + + /* + * memory initialization sequence + */ + #ifdef U4_INFO + printf( " [memory init : ]" ); + #endif + u4_MemInitSequence( tRP, tWR, tRFC, CL, tCK, TD ); + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + #endif + + /* + * start ECC before auto calibration to enable ECC bytelane + */ + store32_ci( MCCR_R, IBIT(0) ); + dly( 15000000 ); + + /* + * start up auto calibration + */ + #ifdef U4_INFO + printf( " [auto calibration: ]\b" ); + #endif + + /* + * start auto calibration + */ + l_acerr_i32 = 0; + do { + progbar(); + + l_ac_i32 = u4_auto_calib( &l_ac_st ); + + if( l_ac_i32 != 0 ) { + l_acerr_i32++; + } + + dly( 15000000 ); + } while( ( l_ac_i32 != 0 ) && + ( l_acerr_i32 <= MAX_ACERR ) ); + + if( l_acerr_i32 > MAX_ACERR ) { + #ifdef U4_INFO + printf( "\b\b\bERR\r\n" ); + #endif + return RET_ERR; + } + + /* + * insert auto calibration results + */ + store32_ci( MemBusCnfg_R, l_ac_st.m_MemBusCnfg_u32 ); + store32_ci( MemBusCnfg2_R, l_ac_st.m_MemBusCnfg2_u32 ); + store32_ci( RstLdEnVerniersC0_R, l_ac_st.m_RstLdEnVerniers_pu32[0] ); + store32_ci( RstLdEnVerniersC1_R, l_ac_st.m_RstLdEnVerniers_pu32[1] ); + store32_ci( RstLdEnVerniersC2_R, l_ac_st.m_RstLdEnVerniers_pu32[2] ); + store32_ci( RstLdEnVerniersC3_R, l_ac_st.m_RstLdEnVerniers_pu32[3] ); + + /* + * insert final timing value into MemRWArb + */ + l_MemRWArb_u32 = ( ( l_ac_st.m_MemBusCnfg_u32 >> 28 /*RdMacDel*/) + 1 ); + l_MemRWArb_u32 *= 10; // needed for rounding + l_MemRWArb_u32 /= 2; // due to spec + l_MemRWArb_u32 += 9; // round up + l_MemRWArb_u32 /= 10; // value is rounded now + l_MemRWArb_u32 = l_MemRWArb_u32 + 6 - WL - TD; + l_MemRWArb_u32 |= rst->MemRWArb; + store32_ci( MemRWArb_R, l_MemRWArb_u32 ); + + progbar(); + dly( 15000000 ); + + /* + * do initial scrubbing + */ + *f_ecc_pt = u4_InitialScrub(); + + switch( f_ecc_pt->m_err_i32 ) { + case RET_OK: { + #ifdef U4_INFO + printf( "\b\bOK\r\n" ); + #endif + break; + } + + case RET_ACERR_CE: { + #ifdef U4_INFO + printf( "\b\b\b\bWEAK][correctable errors during scrub (%u)]\r\n", + f_ecc_pt->m_cecnt_u32 ); + #endif + break; + } + + case RET_ACERR_UEWT: + case RET_ACERR_UE: { + #ifdef U4_INFO + printf( "\b\b\bERR][uncorrectable errors during scrub (%u)]\r\n", + f_ecc_pt->m_uecnt_u32 ); + #endif + return RET_ACERR_UE; + } + + } + + /* + * start continuous background scrub + */ + #ifdef U4_INFO + printf( " [background scrub: ]" ); + #endif + + u4_Scrub( BACKGROUND_SCRUB, 0, NULL ); + + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + #endif + + /* + * finally clear API Exception register + * (read to clear) + */ + load32_ci( APIExcp_R ); + + return RET_OK; +} + +#undef RND + +void +u4_memtest(uint8_t argCnt, char *pArgs[], uint64_t flags) + +{ + #define TEND 99 + #define TCHK 100 + static const uint64_t _2GB = (uint64_t) 0x80000000; + static const uint64_t _start = (uint64_t) 0x08000000; // 128Mb + static const uint64_t _bsize = (uint64_t) 0x08000000; // 128MB + static const uint64_t _line = (uint64_t) 128; + static const uint64_t _256MB = (uint64_t) 0x10000000; + + static const uint64_t PATTERN[] = { + 0x9090909090909090, 0x0020002000200020, + 0x0c0c0c0c0c0c0c0c, 0x8080808080808080, + 0x1004010004001041, 0x0000000000000000 + }; + + uint64_t mend = (uint64_t) 0x200000000;//m_memsize_u64; + uint64_t numblocks = ( mend - _start ) / _bsize; // 128Mb blocks + uint64_t numlines = _bsize / _line; + uint64_t tstate = 0; + uint64_t tlast = 0; + uint64_t pidx = 0; + uint64_t rotr = 0; + uint64_t rotl = 0; + uint64_t block; + uint64_t line; + uint64_t addr; + uint64_t i; + uint64_t check = 0; + uint64_t dcnt; + uint64_t uerr = 0; + uint64_t cerr = 0; + uint64_t merr = 0; + char c; + + printf( "\n\nU4 memory test" ); + printf( "\n--------------" ); + + /* + * mask out UEC & CEC + */ + or32_ci( MCCR_R, IBIT(6) | IBIT(7) ); + + while( PATTERN[pidx] ) { + + switch( tstate ) + { + case 0: { + printf( "\npattern fill 0x%08X%08X: ", (uint32_t) (PATTERN[pidx] >> 32), (uint32_t) PATTERN[pidx] ); + + /* + * first switch lines, then blocks. This way the CPU + * is not able to cache data + */ + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + *( (uint64_t *) addr ) = PATTERN[pidx]; + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + printf( "*" ); + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + check = PATTERN[pidx]; + tlast = 0; + tstate = TCHK; + } break; + + case 1: { + uint64_t one; + + /* + * new check pattern + + */ + one = ( ( check & 0x1 ) != 0 ); + check >>= 1; + if( one ) { + check |= 0x8000000000000000; + } + + printf( "\nrotate right 0x%08X%08X: ", (uint32_t) (check >> 32), (uint32_t) check ); + + /* + * first switch lines, then blocks. This way the CPU + * is not able to cache data + */ + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + *( (uint64_t *) addr ) >>= 1; + + if( one ) { + *( (uint64_t *) addr ) |= + (uint64_t) 0x8000000000000000; + } + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + printf( "*" ); + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + tlast = 1; + tstate = TCHK; + } break; + + case 2: { + + if( rotr < 6 ) { + rotr++; + tstate = 1; + } else { + rotr = 0; + tstate = 3; + } + + } break; + + case 3: { + /* + * new check pattern + */ + check ^= (uint64_t) ~0; + + printf( "\ninverting 0x%08X%08X: ", (uint32_t) (check >> 32), (uint32_t) check ); + + /* + * first switch lines, then blocks. This way the CPU + * is not able to cache data + */ + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + *( (uint64_t *) addr ) ^= (uint64_t) ~0; + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + printf( "*" ); + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + tlast = 3; + tstate = TCHK; + } break; + + case 4: { + uint64_t one; + + /* + * new check pattern + */ + one = ( ( check & 0x8000000000000000 ) != 0 ); + check <<= 1; + if( one ) { + check |= 0x1; + } + + printf( "\nrotate left 0x%08X%08X: ", (uint32_t) (check >> 32), (uint32_t) check ); + + /* + * first switch lines, then blocks. This way the CPU + * is not able to cache data + */ + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + *( (uint64_t *) addr ) <<= 1; + + if( one ) { + *( (uint64_t *) addr ) |= + (uint64_t) 0x1; + } + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + printf( "*" ); + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + tlast = 4; + tstate = TCHK; + } break; + + case 5: { + + if( rotl < 6 ) { + rotl++; + tstate = 4; + } else { + rotl = 0; + tstate = 6; + } + + } break; + + case 6: { + /* + * new check pattern + */ + check *= ~check; + printf( "\nmultiply 0x%08X%08X: ", (uint32_t) (check >> 32), (uint32_t) check ); + + /* + * first switch lines, then blocks. This way the CPU + * is not able to cache data + */ + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + *( (uint64_t *) addr ) *= ~( *( (uint64_t *) addr ) ); + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + printf( "*" ); + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + tlast = TEND - 1; + tstate = TCHK; + } break; + + case TEND: { + pidx++; + tstate = 0; + } break; + + case TCHK: { + uint64_t err; + /* + * check data + */ + printf( "\nchecking : " ); + + for( line = 0, dcnt = 0; line < numlines; line++ ) { + + for( block = 0; block < numblocks; block++ ) { + + for( i = 0; i < _line; i += 8 ) { + addr = _start + + ( block * _bsize ) + + ( line * _line ) + + i; + + if( addr >= _2GB ) { + addr += _2GB; + } + + err = ( *( (uint64_t *) addr ) != check ); + + if( err ) { + merr++; + } + + /* + * print out a dot every 256Mb + */ + dcnt += 8; + if( dcnt == _256MB ) { + dcnt = 0; + + if( err ) { + printf( "X" ); + } else { + printf( "*" ); + } + + if( io_getchar( &c ) ) { + goto mtend; + } + + } + + } + + } + + } + + err = (uint64_t) load32_ci( MEAR1_R ); + uerr += ( err >> 24 ) & (uint64_t) 0xff; + cerr += ( err >> 16 ) & (uint64_t) 0xff; + + printf( " (UE: %02llX, CE: %02llX)", ( err >> 24 ) & (uint64_t) 0xff, ( err >> 16 ) & (uint64_t) 0xff ); + + tstate = tlast + 1; + tlast = TCHK; + } break; + + } + + } + +mtend: + printf( "\n\nmemory test results" ); + printf( "\n-------------------" ); + printf( "\nuncorrectable errors: %u", (uint32_t) uerr ); + printf( "\ncorrectable errors : %u", (uint32_t) cerr ); + printf( "\nread/write errors : %u\n", (uint32_t) merr ); + + and32_ci( MCCR_R, ~( IBIT(6) | IBIT(7) ) ); +} + +void +u4_dump(uint8_t argCnt, char *pArgs[], uint64_t flags) +{ + printf( "\r\n*** u4 register dump ***\r\n\n" ); + printf( "register (offset): value\r\n" ); + printf( "----------------------------------\r\n" ); + printf( "Clock Control (0x%04X): 0x%08X\r\n", (uint16_t) ClkCntl_R, load32_ci( ClkCntl_R ) ); + printf( "PLL2 Control (0x%04X): 0x%08X\r\n", (uint16_t) PLL2Cntl_R, load32_ci( PLL2Cntl_R ) ); + printf( "MemModeCntl (0x%04X): 0x%08X\r\n", (uint16_t) MemModeCntl_R, load32_ci( MemModeCntl_R ) ); + printf( "RASTimer0 (0x%04X): 0x%08X\r\n", (uint16_t) RASTimer0_R, load32_ci( RASTimer0_R ) ); + printf( "RASTimer1 (0x%04X): 0x%08X\r\n", (uint16_t) RASTimer1_R, load32_ci( RASTimer1_R ) ); + printf( "CASTimer0 (0x%04X): 0x%08X\r\n", (uint16_t) CASTimer0_R, load32_ci( CASTimer0_R ) ); + printf( "CASTimer1 (0x%04X): 0x%08X\r\n", (uint16_t) CASTimer1_R, load32_ci( CASTimer1_R ) ); + printf( "MemRfshCntl (0x%04X): 0x%08X\r\n", (uint16_t) MemRfshCntl_R, load32_ci( MemRfshCntl_R ) ); + printf( "Dm0Cnfg (0x%04X): 0x%08X\r\n", (uint16_t) Dm0Cnfg_R, load32_ci( Dm0Cnfg_R ) ); + printf( "Dm1Cnfg (0x%04X): 0x%08X\r\n", (uint16_t) Dm1Cnfg_R, load32_ci( Dm1Cnfg_R ) ); + printf( "Dm2Cnfg (0x%04X): 0x%08X\r\n", (uint16_t) Dm2Cnfg_R, load32_ci( Dm2Cnfg_R ) ); + printf( "Dm3Cnfg (0x%04X): 0x%08X\r\n", (uint16_t) Dm3Cnfg_R, load32_ci( Dm3Cnfg_R ) ); + printf( "UsrCnfg (0x%04X): 0x%08X\r\n", (uint16_t) UsrCnfg_R, load32_ci( UsrCnfg_R ) ); + printf( "MemArbWt (0x%04X): 0x%08X\r\n", (uint16_t) MemArbWt_R, load32_ci( MemArbWt_R ) ); + printf( "ODTCntl (0x%04X): 0x%08X\r\n", (uint16_t) ODTCntl_R, load32_ci( ODTCntl_R ) ); + printf( "IOPadCntl (0x%04X): 0x%08X\r\n", (uint16_t) IOPadCntl_R, load32_ci( IOPadCntl_R ) ); + printf( "MemPhyMode (0x%04X): 0x%08X\r\n", (uint16_t) MemPhyModeCntl_R, load32_ci( MemPhyModeCntl_R ) ); + printf( "OCDCalCntl (0x%04X): 0x%08X\r\n", (uint16_t) OCDCalCntl_R, load32_ci( OCDCalCntl_R ) ); + printf( "OCDCalCmd (0x%04X): 0x%08X\r\n", (uint16_t) OCDCalCmd_R, load32_ci( OCDCalCmd_R ) ); + printf( "CKDelayL (0x%04X): 0x%08X\r\n", (uint16_t) CKDelayL_R, load32_ci( CKDelayL_R ) ); + printf( "CKDelayH (0x%04X): 0x%08X\r\n", (uint16_t) CKDelayU_R, load32_ci( CKDelayU_R ) ); + printf( "MemBusCnfg (0x%04X): 0x%08X\r\n", (uint16_t) MemBusCnfg_R, load32_ci( MemBusCnfg_R ) ); + printf( "MemBusCnfg2 (0x%04X): 0x%08X\r\n", (uint16_t) MemBusCnfg2_R, load32_ci( MemBusCnfg2_R ) ); + printf( "MemRdQCnfg (0x%04X): 0x%08X\r\n", (uint16_t) MemRdQCnfg_R, load32_ci( MemRdQCnfg_R ) ); + printf( "MemWrQCnfg (0x%04X): 0x%08X\r\n", (uint16_t) MemWrQCnfg_R, load32_ci( MemWrQCnfg_R ) ); + printf( "MemQArb (0x%04X): 0x%08X\r\n", (uint16_t) MemQArb_R, load32_ci( MemQArb_R ) ); + printf( "MemRWArb (0x%04X): 0x%08X\r\n", (uint16_t) MemRWArb_R, load32_ci( MemRWArb_R ) ); + printf( "ByteWrClkDel (0x%04X): 0x%08X\r\n", (uint16_t) ByteWrClkDelC0B00_R, load32_ci( ByteWrClkDelC0B00_R ) ); + printf( "ReadStrobeDel (0x%04X): 0x%08X\r\n", (uint16_t) ReadStrobeDelC0B00_R, load32_ci( ReadStrobeDelC0B00_R ) ); + printf( "RstLdEnVerC0 (0x%04X): 0x%08X\r\n", (uint16_t) RstLdEnVerniersC0_R, load32_ci( RstLdEnVerniersC0_R ) ); + printf( "RstLdEnVerC1 (0x%04X): 0x%08X\r\n", (uint16_t) RstLdEnVerniersC1_R, load32_ci( RstLdEnVerniersC1_R ) ); + printf( "RstLdEnVerC2 (0x%04X): 0x%08X\r\n", (uint16_t) RstLdEnVerniersC2_R, load32_ci( RstLdEnVerniersC2_R ) ); + printf( "RstLdEnVerC3 (0x%04X): 0x%08X\r\n", (uint16_t) RstLdEnVerniersC3_R, load32_ci( RstLdEnVerniersC3_R ) ); + printf( "APIMemRdCfg (0x%04X): 0x%08X\r\n", (uint16_t) APIMemRdCfg_R, load32_ci( APIMemRdCfg_R ) ); + printf( "scrub start (0x%04X): 0x%08X\r\n", (uint16_t) MSRSR_R, load32_ci( MSRSR_R ) ); + printf( "scrub end (0x%04X): 0x%08X\r\n", (uint16_t) MSRER_R, load32_ci( MSRER_R ) ); +} + +int32_t +u4_memBegin( eccerror_t *f_ecc_pt ) +{ + int32_t i; + + #ifdef U4_INFO + printf( "\r\n" ); + printf( "U4 DDR2 memory controller setup V%u.%u\r\n", + VER, SUBVER ); + printf( "------------------------------------\r\n" ); + printf( "> detected board : " ); + + if( IS_MAUI ) { + printf( "MAUI" ); + } else if( IS_BIMINI ) { + printf( "BIMINI" ); + } else if( IS_KAUAI ) { + printf( "KAUAI" ); + } else { + printf( "unknown!" ); + return RET_ERR; + } + #endif + + do { + /* + * initialize variables + */ + m_memsize_u64 = 0; + m_dcnt_u32 = 0; + m_dgrcnt_u32 = 0; + m_dclidx_u32 = 0; + + for( i = 0; i < NUM_SLOTS; i++ ) { + m_dptr[i] = NULL; + memset( ( void * ) &m_dimm[i], 0, sizeof( dimm_t ) ); + } + + for( i = 0; i < MAX_DGROUPS; i++ ) { + m_dgrptr[i] = NULL; + memset( ( void * ) &m_dgroup[i], 0, sizeof( dimm_t ) ); + } + + /* + * start configuration + */ + #ifdef U4_INFO + printf( "\r\n> detected DIMM configuration : " ); + #endif + + i = ddr2_readSPDs(); + + if( i != RET_OK ) { + #ifdef U4_INFO + printf( "\r\n-------------------------------------------------------------" ); + printf( "\r\n switching off memory bank(s) due to SPD integrity failure" ); + printf( "\r\n-------------------------------------------------------------\r\n" ); + #endif + } + + } while( i != RET_OK ); + + /* + * check DIMM configuration + */ + if( ddr2_setupDIMMcfg() != RET_OK ) { + #ifdef U4_INFO + printf( "> initialization failure.\r\n" ); + #endif + return RET_ERR; + } + + /* + * create DIMM groups + */ + u4_setupDIMMgroups(); + + /* + * start configuration of u4 + */ + u4_calcDIMMcnfg(); + + if( u4_calcDIMMmemmode() != RET_OK ) { + #ifdef U4_INFO + printf( "> initialization failure.\r\n" ); + #endif + return RET_ERR; + } + + #ifdef U4_INFO + printf( "%uMb @ %uMhz, CL %u\r\n", + (uint32_t) ( m_memsize_u64 / 0x100000 ), + m_gendimm.m_speed_pu32[m_dclidx_u32], + m_gendimm.m_clval_pu32[m_dclidx_u32] ); + + printf( "> initializing memory :\r\n" ); + #endif + + if( u4_setup_core_clock() != RET_OK ) { + #ifdef U4_INFO + printf( "> initialization failure.\r\n" ); + #endif + return RET_ERR; + } + + i = u4_start( f_ecc_pt ); + if( i != RET_OK ) { + #ifdef U4_INFO + printf( "> initialization failure.\r\n" ); + #endif + return i; + } + + #ifdef U4_INFO + printf( " [flush cache : ]" ); + #endif + + flush_cache( 0x0, L2_CACHE_SIZE ); + + #ifdef U4_INFO + printf( "\b\b\bOK\r\n" ); + printf( "> initialization complete.\r\n" ); + #endif + +#ifdef U4_SHOW_REGS + u4_dump(0,0,0); +#endif + + return RET_OK; +} + +static int32_t scrubstarted = 0; + +void +u4_scrubStart(uint8_t argCnt, char *pArgs[], uint64_t flags ) +{ + scrubstarted = 1; + + /* + * setup scrub parameters + */ + store32_ci( MSCR_R, 0 ); // stop scrub + store32_ci( MSRSR_R, 0x0 ); // set start + store32_ci( MSRER_R, 0x1c ); // set end + store32_ci( MSPR_R, 0x0 ); // set pattern + + /* + * clear out ECC error registers + */ + store32_ci( MEAR0_R, 0x0 ); + store32_ci( MEAR1_R, 0x0 ); + store32_ci( MESR_R, 0x0 ); + + /* + * Setup Scrub Type + */ + store32_ci( MSCR_R, IBIT(1) ); + printf( "\r\nscrub started\r\n" ); +} + +void +u4_scrubEnd(uint8_t argCnt, char *pArgs[], uint64_t flags ) +{ + store32_ci( MSCR_R, 0 ); // stop scrub + scrubstarted = 0; + printf( "\r\nscrub stopped\r\n" ); +} + +void +u4_memwr(uint8_t argCnt, char *pArgs[], uint64_t flags ) +{ + uint32_t i; + uint32_t v = 0; + + for( i = 0; i < 0x200; i += 4 ) { + + if( ( i & 0xf ) == 0 ) { + v = ~v; + } + + store32_ci( i, v ); + } + +} + +void +u4memInit() +{ + static uint32_t l_isInit_u32 = 0; + eccerror_t l_ecc_t; + int32_t ret; + + /* + * do not initialize memory more than once + */ + if( l_isInit_u32 ) { + #ifdef U4_INFO + printf( "\r\n\nmemory already initialized\r\n" ); + #endif + return; + } else { + l_isInit_u32 = 1; + } + + /* + * enable all DIMM banks on first run + */ + m_bankoff_u32 = 0; + + do { + ret = u4_memBegin( &l_ecc_t ); + + if( ret < RET_ERR ) { + uint32_t l_bank_u32 = l_ecc_t.m_rank_u32 / 2; + printf( "\r\n-----------------------------------------------------" ); + printf( "\r\n switching off memory bank %u due to memory failure", l_bank_u32 ); + printf( "\r\n-----------------------------------------------------" ); + m_bankoff_u32 |= ( 1 << l_bank_u32 ); + } + + } while( ret < RET_ERR ); + +} + +void +monitorDDR2( uint8_t argCnt, char *pArgs[], uint64_t flags ) +{ + u4memInit(); +} |