diff options
Diffstat (limited to 'arch/sparc64/mm/init.c')
-rw-r--r-- | arch/sparc64/mm/init.c | 441 |
1 files changed, 206 insertions, 235 deletions
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c index 9f6ca624892..0d2e967c720 100644 --- a/arch/sparc64/mm/init.c +++ b/arch/sparc64/mm/init.c @@ -21,6 +21,7 @@ #include <linux/seq_file.h> #include <linux/kprobes.h> #include <linux/cache.h> +#include <linux/sort.h> #include <asm/head.h> #include <asm/system.h> @@ -41,7 +42,72 @@ extern void device_scan(void); -struct sparc_phys_banks sp_banks[SPARC_PHYS_BANKS]; +#define MAX_BANKS 32 + +static struct linux_prom64_registers pavail[MAX_BANKS] __initdata; +static struct linux_prom64_registers pavail_rescan[MAX_BANKS] __initdata; +static int pavail_ents __initdata; +static int pavail_rescan_ents __initdata; + +static int cmp_p64(const void *a, const void *b) +{ + const struct linux_prom64_registers *x = a, *y = b; + + if (x->phys_addr > y->phys_addr) + return 1; + if (x->phys_addr < y->phys_addr) + return -1; + return 0; +} + +static void __init read_obp_memory(const char *property, + struct linux_prom64_registers *regs, + int *num_ents) +{ + int node = prom_finddevice("/memory"); + int prop_size = prom_getproplen(node, property); + int ents, ret, i; + + ents = prop_size / sizeof(struct linux_prom64_registers); + if (ents > MAX_BANKS) { + prom_printf("The machine has more %s property entries than " + "this kernel can support (%d).\n", + property, MAX_BANKS); + prom_halt(); + } + + ret = prom_getproperty(node, property, (char *) regs, prop_size); + if (ret == -1) { + prom_printf("Couldn't get %s property from /memory.\n"); + prom_halt(); + } + + *num_ents = ents; + + /* Sanitize what we got from the firmware, by page aligning + * everything. + */ + for (i = 0; i < ents; i++) { + unsigned long base, size; + + base = regs[i].phys_addr; + size = regs[i].reg_size; + + size &= PAGE_MASK; + if (base & ~PAGE_MASK) { + unsigned long new_base = PAGE_ALIGN(base); + + size -= new_base - base; + if ((long) size < 0L) + size = 0UL; + base = new_base; + } + regs[i].phys_addr = base; + regs[i].reg_size = size; + } + sort(regs, ents, sizeof(struct linux_prom64_registers), + cmp_p64, NULL); +} unsigned long *sparc64_valid_addr_bitmap __read_mostly; @@ -67,6 +133,12 @@ extern unsigned int sparc_ramdisk_size; struct page *mem_map_zero __read_mostly; +unsigned int sparc64_highest_unlocked_tlb_ent __read_mostly; + +unsigned long sparc64_kern_pri_context __read_mostly; +unsigned long sparc64_kern_pri_nuc_bits __read_mostly; +unsigned long sparc64_kern_sec_context __read_mostly; + int bigkernel = 0; /* XXX Tune this... */ @@ -296,6 +368,7 @@ struct linux_prom_translation { unsigned long data; }; static struct linux_prom_translation prom_trans[512] __initdata; +static unsigned int prom_trans_ents __initdata; extern unsigned long prom_boot_page; extern void prom_remap(unsigned long physpage, unsigned long virtpage, int mmu_ihandle); @@ -309,57 +382,7 @@ unsigned long kern_locked_tte_data; unsigned long prom_pmd_phys __read_mostly; unsigned int swapper_pgd_zero __read_mostly; -/* Allocate power-of-2 aligned chunks from the end of the - * kernel image. Return physical address. - */ -static inline unsigned long early_alloc_phys(unsigned long size) -{ - unsigned long base; - - BUILD_BUG_ON(size & (size - 1)); - - kern_size = (kern_size + (size - 1)) & ~(size - 1); - base = kern_base + kern_size; - kern_size += size; - - return base; -} - -static inline unsigned long load_phys32(unsigned long pa) -{ - unsigned long val; - - __asm__ __volatile__("lduwa [%1] %2, %0" - : "=&r" (val) - : "r" (pa), "i" (ASI_PHYS_USE_EC)); - - return val; -} - -static inline unsigned long load_phys64(unsigned long pa) -{ - unsigned long val; - - __asm__ __volatile__("ldxa [%1] %2, %0" - : "=&r" (val) - : "r" (pa), "i" (ASI_PHYS_USE_EC)); - - return val; -} - -static inline void store_phys32(unsigned long pa, unsigned long val) -{ - __asm__ __volatile__("stwa %0, [%1] %2" - : /* no outputs */ - : "r" (val), "r" (pa), "i" (ASI_PHYS_USE_EC)); -} - -static inline void store_phys64(unsigned long pa, unsigned long val) -{ - __asm__ __volatile__("stxa %0, [%1] %2" - : /* no outputs */ - : "r" (val), "r" (pa), "i" (ASI_PHYS_USE_EC)); -} +static pmd_t *prompmd __read_mostly; #define BASE_PAGE_SIZE 8192 @@ -369,34 +392,28 @@ static inline void store_phys64(unsigned long pa, unsigned long val) */ unsigned long prom_virt_to_phys(unsigned long promva, int *error) { - unsigned long pmd_phys = (prom_pmd_phys + - ((promva >> 23) & 0x7ff) * sizeof(pmd_t)); - unsigned long pte_phys; - pmd_t pmd_ent; - pte_t pte_ent; + pmd_t *pmdp = prompmd + ((promva >> 23) & 0x7ff); + pte_t *ptep; unsigned long base; - pmd_val(pmd_ent) = load_phys32(pmd_phys); - if (pmd_none(pmd_ent)) { + if (pmd_none(*pmdp)) { if (error) *error = 1; return 0; } - - pte_phys = (unsigned long)pmd_val(pmd_ent) << 11UL; - pte_phys += ((promva >> 13) & 0x3ff) * sizeof(pte_t); - pte_val(pte_ent) = load_phys64(pte_phys); - if (!pte_present(pte_ent)) { + ptep = (pte_t *)__pmd_page(*pmdp) + ((promva >> 13) & 0x3ff); + if (!pte_present(*ptep)) { if (error) *error = 1; return 0; } if (error) { *error = 0; - return pte_val(pte_ent); + return pte_val(*ptep); } - base = pte_val(pte_ent) & _PAGE_PADDR; - return (base + (promva & (BASE_PAGE_SIZE - 1))); + base = pte_val(*ptep) & _PAGE_PADDR; + + return base + (promva & (BASE_PAGE_SIZE - 1)); } /* The obp translations are saved based on 8k pagesize, since obp can @@ -409,25 +426,20 @@ static void __init build_obp_range(unsigned long start, unsigned long end, unsig unsigned long vaddr; for (vaddr = start; vaddr < end; vaddr += BASE_PAGE_SIZE) { - unsigned long val, pte_phys, pmd_phys; - pmd_t pmd_ent; - int i; - - pmd_phys = (prom_pmd_phys + - (((vaddr >> 23) & 0x7ff) * sizeof(pmd_t))); - pmd_val(pmd_ent) = load_phys32(pmd_phys); - if (pmd_none(pmd_ent)) { - pte_phys = early_alloc_phys(BASE_PAGE_SIZE); - - for (i = 0; i < BASE_PAGE_SIZE / sizeof(pte_t); i++) - store_phys64(pte_phys+i*sizeof(pte_t),0); + unsigned long val; + pmd_t *pmd; + pte_t *pte; - pmd_val(pmd_ent) = pte_phys >> 11UL; - store_phys32(pmd_phys, pmd_val(pmd_ent)); + pmd = prompmd + ((vaddr >> 23) & 0x7ff); + if (pmd_none(*pmd)) { + pte = __alloc_bootmem(BASE_PAGE_SIZE, BASE_PAGE_SIZE, + PAGE_SIZE); + if (!pte) + prom_halt(); + memset(pte, 0, BASE_PAGE_SIZE); + pmd_set(pmd, pte); } - - pte_phys = (unsigned long)pmd_val(pmd_ent) << 11UL; - pte_phys += (((vaddr >> 13) & 0x3ff) * sizeof(pte_t)); + pte = (pte_t *) __pmd_page(*pmd) + ((vaddr >> 13) & 0x3ff); val = data; @@ -435,7 +447,8 @@ static void __init build_obp_range(unsigned long start, unsigned long end, unsig if (tlb_type == spitfire) val &= ~0x0003fe0000000000UL; - store_phys64(pte_phys, val | _PAGE_MODIFIED); + set_pte_at(&init_mm, vaddr, pte, + __pte(val | _PAGE_MODIFIED)); data += BASE_PAGE_SIZE; } @@ -448,13 +461,17 @@ static inline int in_obp_range(unsigned long vaddr) } #define OBP_PMD_SIZE 2048 -static void __init build_obp_pgtable(int prom_trans_ents) +static void __init build_obp_pgtable(void) { unsigned long i; - prom_pmd_phys = early_alloc_phys(OBP_PMD_SIZE); - for (i = 0; i < OBP_PMD_SIZE; i += 4) - store_phys32(prom_pmd_phys + i, 0); + prompmd = __alloc_bootmem(OBP_PMD_SIZE, OBP_PMD_SIZE, PAGE_SIZE); + if (!prompmd) + prom_halt(); + + memset(prompmd, 0, OBP_PMD_SIZE); + + prom_pmd_phys = __pa(prompmd); for (i = 0; i < prom_trans_ents; i++) { unsigned long start, end; @@ -474,7 +491,7 @@ static void __init build_obp_pgtable(int prom_trans_ents) /* Read OBP translations property into 'prom_trans[]'. * Return the number of entries. */ -static int __init read_obp_translations(void) +static void __init read_obp_translations(void) { int n, node; @@ -495,8 +512,10 @@ static int __init read_obp_translations(void) prom_printf("prom_mappings: Couldn't get property.\n"); prom_halt(); } + n = n / sizeof(struct linux_prom_translation); - return n; + + prom_trans_ents = n; } static void __init remap_kernel(void) @@ -516,28 +535,38 @@ static void __init remap_kernel(void) prom_dtlb_load(tlb_ent, tte_data, tte_vaddr); prom_itlb_load(tlb_ent, tte_data, tte_vaddr); if (bigkernel) { - prom_dtlb_load(tlb_ent - 1, + tlb_ent -= 1; + prom_dtlb_load(tlb_ent, tte_data + 0x400000, tte_vaddr + 0x400000); - prom_itlb_load(tlb_ent - 1, + prom_itlb_load(tlb_ent, tte_data + 0x400000, tte_vaddr + 0x400000); } + sparc64_highest_unlocked_tlb_ent = tlb_ent - 1; + if (tlb_type == cheetah_plus) { + sparc64_kern_pri_context = (CTX_CHEETAH_PLUS_CTX0 | + CTX_CHEETAH_PLUS_NUC); + sparc64_kern_pri_nuc_bits = CTX_CHEETAH_PLUS_NUC; + sparc64_kern_sec_context = CTX_CHEETAH_PLUS_CTX0; + } } -static void __init inherit_prom_mappings(void) -{ - int n; - n = read_obp_translations(); - build_obp_pgtable(n); +static void __init inherit_prom_mappings_pre(void) +{ + read_obp_translations(); /* Now fixup OBP's idea about where we really are mapped. */ prom_printf("Remapping the kernel... "); remap_kernel(); prom_printf("done.\n"); +} +static void __init inherit_prom_mappings_post(void) +{ + build_obp_pgtable(); register_prom_callbacks(); } @@ -722,8 +751,8 @@ void inherit_locked_prom_mappings(int save_p) } } if (tlb_type == spitfire) { - int high = SPITFIRE_HIGHEST_LOCKED_TLBENT - bigkernel; - for (i = 0; i < high; i++) { + int high = sparc64_highest_unlocked_tlb_ent; + for (i = 0; i <= high; i++) { unsigned long data; /* Spitfire Errata #32 workaround */ @@ -811,9 +840,9 @@ void inherit_locked_prom_mappings(int save_p) } } } else if (tlb_type == cheetah || tlb_type == cheetah_plus) { - int high = CHEETAH_HIGHEST_LOCKED_TLBENT - bigkernel; + int high = sparc64_highest_unlocked_tlb_ent; - for (i = 0; i < high; i++) { + for (i = 0; i <= high; i++) { unsigned long data; data = cheetah_get_ldtlb_data(i); @@ -1206,14 +1235,14 @@ unsigned long __init bootmem_init(unsigned long *pages_avail) int i; #ifdef CONFIG_DEBUG_BOOTMEM - prom_printf("bootmem_init: Scan sp_banks, "); + prom_printf("bootmem_init: Scan pavail, "); #endif bytes_avail = 0UL; - for (i = 0; sp_banks[i].num_bytes != 0; i++) { - end_of_phys_memory = sp_banks[i].base_addr + - sp_banks[i].num_bytes; - bytes_avail += sp_banks[i].num_bytes; + for (i = 0; i < pavail_ents; i++) { + end_of_phys_memory = pavail[i].phys_addr + + pavail[i].reg_size; + bytes_avail += pavail[i].reg_size; if (cmdline_memory_size) { if (bytes_avail > cmdline_memory_size) { unsigned long slack = bytes_avail - cmdline_memory_size; @@ -1221,12 +1250,15 @@ unsigned long __init bootmem_init(unsigned long *pages_avail) bytes_avail -= slack; end_of_phys_memory -= slack; - sp_banks[i].num_bytes -= slack; - if (sp_banks[i].num_bytes == 0) { - sp_banks[i].base_addr = 0xdeadbeef; + pavail[i].reg_size -= slack; + if ((long)pavail[i].reg_size <= 0L) { + pavail[i].phys_addr = 0xdeadbeefUL; + pavail[i].reg_size = 0UL; + pavail_ents = i; } else { - sp_banks[i+1].num_bytes = 0; - sp_banks[i+1].base_addr = 0xdeadbeef; + pavail[i+1].reg_size = 0Ul; + pavail[i+1].phys_addr = 0xdeadbeefUL; + pavail_ents = i + 1; } break; } @@ -1280,12 +1312,12 @@ unsigned long __init bootmem_init(unsigned long *pages_avail) /* Now register the available physical memory with the * allocator. */ - for (i = 0; sp_banks[i].num_bytes != 0; i++) { + for (i = 0; i < pavail_ents; i++) { #ifdef CONFIG_DEBUG_BOOTMEM - prom_printf("free_bootmem(sp_banks:%d): base[%lx] size[%lx]\n", - i, sp_banks[i].base_addr, sp_banks[i].num_bytes); + prom_printf("free_bootmem(pavail:%d): base[%lx] size[%lx]\n", + i, pavail[i].phys_addr, pavail[i].reg_size); #endif - free_bootmem(sp_banks[i].base_addr, sp_banks[i].num_bytes); + free_bootmem(pavail[i].phys_addr, pavail[i].reg_size); } #ifdef CONFIG_BLK_DEV_INITRD @@ -1334,7 +1366,7 @@ static unsigned long kernel_map_range(unsigned long pstart, unsigned long pend, unsigned long alloc_bytes = 0UL; if ((vstart & ~PAGE_MASK) || (vend & ~PAGE_MASK)) { - prom_printf("kernel_map: Unaligned sp_banks[%lx:%lx]\n", + prom_printf("kernel_map: Unaligned physmem[%lx:%lx]\n", vstart, vend); prom_halt(); } @@ -1381,23 +1413,24 @@ static unsigned long kernel_map_range(unsigned long pstart, unsigned long pend, return alloc_bytes; } -extern struct linux_mlist_p1275 *prom_ptot_ptr; +static struct linux_prom64_registers pall[MAX_BANKS] __initdata; +static int pall_ents __initdata; + extern unsigned int kvmap_linear_patch[1]; static void __init kernel_physical_mapping_init(void) { - struct linux_mlist_p1275 *p = prom_ptot_ptr; - unsigned long mem_alloced = 0UL; + unsigned long i, mem_alloced = 0UL; + + read_obp_memory("reg", &pall[0], &pall_ents); - while (p) { + for (i = 0; i < pall_ents; i++) { unsigned long phys_start, phys_end; - phys_start = p->start_adr; - phys_end = phys_start + p->num_bytes; + phys_start = pall[i].phys_addr; + phys_end = phys_start + pall[i].reg_size; mem_alloced += kernel_map_range(phys_start, phys_end, PAGE_KERNEL); - - p = p->theres_more; } printk("Allocated %ld bytes for kernel page tables.\n", @@ -1425,6 +1458,18 @@ void kernel_map_pages(struct page *page, int numpages, int enable) } #endif +unsigned long __init find_ecache_flush_span(unsigned long size) +{ + int i; + + for (i = 0; i < pavail_ents; i++) { + if (pavail[i].reg_size >= size) + return pavail[i].phys_addr; + } + + return ~0UL; +} + /* paging_init() sets up the page tables */ extern void cheetah_ecache_flush_init(void); @@ -1435,7 +1480,19 @@ pgd_t swapper_pg_dir[2048]; void __init paging_init(void) { unsigned long end_pfn, pages_avail, shift; - unsigned long real_end; + unsigned long real_end, i; + + /* Find available physical memory... */ + read_obp_memory("available", &pavail[0], &pavail_ents); + + phys_base = 0xffffffffffffffffUL; + for (i = 0; i < pavail_ents; i++) + phys_base = min(phys_base, pavail[i].phys_addr); + + pfn_base = phys_base >> PAGE_SHIFT; + + kern_base = (prom_boot_mapping_phys_low >> 22UL) << 22UL; + kern_size = (unsigned long)&_end - (unsigned long)KERNBASE; set_bit(0, mmu_context_bmap); @@ -1462,8 +1519,7 @@ void __init paging_init(void) swapper_pgd_zero = pgd_val(swapper_pg_dir[0]); - /* Inherit non-locked OBP mappings. */ - inherit_prom_mappings(); + inherit_prom_mappings_pre(); /* Ok, we can use our TLB miss and window trap handlers safely. * We need to do a quick peek here to see if we are on StarFire @@ -1474,15 +1530,23 @@ void __init paging_init(void) extern void setup_tba(int); setup_tba(this_is_starfire); } - - inherit_locked_prom_mappings(1); - __flush_tlb_all(); + /* Everything from this point forward, until we are done with + * inherit_prom_mappings_post(), must complete successfully + * without calling into the firmware. The firwmare page tables + * have not been built, but we are running on the Linux kernel's + * trap table. + */ + /* Setup bootmem... */ pages_avail = 0; last_valid_pfn = end_pfn = bootmem_init(&pages_avail); + inherit_prom_mappings_post(); + + inherit_locked_prom_mappings(1); + #ifdef CONFIG_DEBUG_PAGEALLOC kernel_physical_mapping_init(); #endif @@ -1507,128 +1571,35 @@ void __init paging_init(void) device_scan(); } -/* Ok, it seems that the prom can allocate some more memory chunks - * as a side effect of some prom calls we perform during the - * boot sequence. My most likely theory is that it is from the - * prom_set_traptable() call, and OBP is allocating a scratchpad - * for saving client program register state etc. - */ -static void __init sort_memlist(struct linux_mlist_p1275 *thislist) -{ - int swapi = 0; - int i, mitr; - unsigned long tmpaddr, tmpsize; - unsigned long lowest; - - for (i = 0; thislist[i].theres_more != 0; i++) { - lowest = thislist[i].start_adr; - for (mitr = i+1; thislist[mitr-1].theres_more != 0; mitr++) - if (thislist[mitr].start_adr < lowest) { - lowest = thislist[mitr].start_adr; - swapi = mitr; - } - if (lowest == thislist[i].start_adr) - continue; - tmpaddr = thislist[swapi].start_adr; - tmpsize = thislist[swapi].num_bytes; - for (mitr = swapi; mitr > i; mitr--) { - thislist[mitr].start_adr = thislist[mitr-1].start_adr; - thislist[mitr].num_bytes = thislist[mitr-1].num_bytes; - } - thislist[i].start_adr = tmpaddr; - thislist[i].num_bytes = tmpsize; - } -} - -void __init rescan_sp_banks(void) -{ - struct linux_prom64_registers memlist[64]; - struct linux_mlist_p1275 avail[64], *mlist; - unsigned long bytes, base_paddr; - int num_regs, node = prom_finddevice("/memory"); - int i; - - num_regs = prom_getproperty(node, "available", - (char *) memlist, sizeof(memlist)); - num_regs = (num_regs / sizeof(struct linux_prom64_registers)); - for (i = 0; i < num_regs; i++) { - avail[i].start_adr = memlist[i].phys_addr; - avail[i].num_bytes = memlist[i].reg_size; - avail[i].theres_more = &avail[i + 1]; - } - avail[i - 1].theres_more = NULL; - sort_memlist(avail); - - mlist = &avail[0]; - i = 0; - bytes = mlist->num_bytes; - base_paddr = mlist->start_adr; - - sp_banks[0].base_addr = base_paddr; - sp_banks[0].num_bytes = bytes; - - while (mlist->theres_more != NULL){ - i++; - mlist = mlist->theres_more; - bytes = mlist->num_bytes; - if (i >= SPARC_PHYS_BANKS-1) { - printk ("The machine has more banks than " - "this kernel can support\n" - "Increase the SPARC_PHYS_BANKS " - "setting (currently %d)\n", - SPARC_PHYS_BANKS); - i = SPARC_PHYS_BANKS-1; - break; - } - - sp_banks[i].base_addr = mlist->start_adr; - sp_banks[i].num_bytes = mlist->num_bytes; - } - - i++; - sp_banks[i].base_addr = 0xdeadbeefbeefdeadUL; - sp_banks[i].num_bytes = 0; - - for (i = 0; sp_banks[i].num_bytes != 0; i++) - sp_banks[i].num_bytes &= PAGE_MASK; -} - static void __init taint_real_pages(void) { - struct sparc_phys_banks saved_sp_banks[SPARC_PHYS_BANKS]; int i; - for (i = 0; i < SPARC_PHYS_BANKS; i++) { - saved_sp_banks[i].base_addr = - sp_banks[i].base_addr; - saved_sp_banks[i].num_bytes = - sp_banks[i].num_bytes; - } - - rescan_sp_banks(); + read_obp_memory("available", &pavail_rescan[0], &pavail_rescan_ents); - /* Find changes discovered in the sp_bank rescan and + /* Find changes discovered in the physmem available rescan and * reserve the lost portions in the bootmem maps. */ - for (i = 0; saved_sp_banks[i].num_bytes; i++) { + for (i = 0; i < pavail_ents; i++) { unsigned long old_start, old_end; - old_start = saved_sp_banks[i].base_addr; + old_start = pavail[i].phys_addr; old_end = old_start + - saved_sp_banks[i].num_bytes; + pavail[i].reg_size; while (old_start < old_end) { int n; - for (n = 0; sp_banks[n].num_bytes; n++) { + for (n = 0; pavail_rescan_ents; n++) { unsigned long new_start, new_end; - new_start = sp_banks[n].base_addr; - new_end = new_start + sp_banks[n].num_bytes; + new_start = pavail_rescan[n].phys_addr; + new_end = new_start + + pavail_rescan[n].reg_size; if (new_start <= old_start && new_end >= (old_start + PAGE_SIZE)) { - set_bit (old_start >> 22, - sparc64_valid_addr_bitmap); + set_bit(old_start >> 22, + sparc64_valid_addr_bitmap); goto do_next_page; } } |