Index: linux-2.6.22/arch/mips/kernel/genex.S =================================================================== --- linux-2.6.22.orig/arch/mips/kernel/genex.S 2007-07-26 06:29:25.057170943 +0200 +++ linux-2.6.22/arch/mips/kernel/genex.S 2007-07-26 06:29:40.890073208 +0200 @@ -51,6 +51,10 @@ NESTED(except_vec3_generic, 0, sp) .set push .set noat +#ifdef CONFIG_BCM947XX + nop + nop +#endif #if R5432_CP0_INTERRUPT_WAR mfc0 k0, CP0_INDEX #endif Index: linux-2.6.22/arch/mips/mm/c-r4k.c =================================================================== --- linux-2.6.22.orig/arch/mips/mm/c-r4k.c 2007-07-26 06:29:40.826069560 +0200 +++ linux-2.6.22/arch/mips/mm/c-r4k.c 2007-07-26 06:32:45.956619550 +0200 @@ -29,6 +29,9 @@ #include /* for run_uncached() */ +/* For enabling BCM4710 cache workarounds */ +int bcm4710 = 0; + /* * Special Variant of smp_call_function for use by cache functions: * @@ -85,14 +88,21 @@ static inline void r4k_blast_dcache_page_dc32(unsigned long addr) { + unsigned long flags; + + local_irq_save(flags); R4600_HIT_CACHEOP_WAR_IMPL; blast_dcache32_page(addr); + local_irq_restore(flags); } static void __init r4k_blast_dcache_page_setup(void) { unsigned long dc_lsize = cpu_dcache_line_size(); + if (bcm4710) + r4k_blast_dcache_page = blast_dcache_page; + else if (dc_lsize == 0) r4k_blast_dcache_page = (void *)cache_noop; else if (dc_lsize == 16) @@ -107,6 +117,9 @@ { unsigned long dc_lsize = cpu_dcache_line_size(); + if (bcm4710) + r4k_blast_dcache_page_indexed = blast_dcache_page_indexed; + else if (dc_lsize == 0) r4k_blast_dcache_page_indexed = (void *)cache_noop; else if (dc_lsize == 16) @@ -121,6 +134,9 @@ { unsigned long dc_lsize = cpu_dcache_line_size(); + if (bcm4710) + r4k_blast_dcache = blast_dcache; + else if (dc_lsize == 0) r4k_blast_dcache = (void *)cache_noop; else if (dc_lsize == 16) @@ -202,8 +218,12 @@ static void (* r4k_blast_icache_page)(unsigned long addr); +static void r4k_flush_cache_all(void); static void __init r4k_blast_icache_page_setup(void) { +#ifdef CONFIG_BCM947XX + r4k_blast_icache_page = (void *)r4k_flush_cache_all; +#else unsigned long ic_lsize = cpu_icache_line_size(); if (ic_lsize == 0) @@ -214,6 +234,7 @@ r4k_blast_icache_page = blast_icache32_page; else if (ic_lsize == 64) r4k_blast_icache_page = blast_icache64_page; +#endif } @@ -221,6 +242,9 @@ static void __init r4k_blast_icache_page_indexed_setup(void) { +#ifdef CONFIG_BCM947XX + r4k_blast_icache_page_indexed = (void *)r4k_flush_cache_all; +#else unsigned long ic_lsize = cpu_icache_line_size(); if (ic_lsize == 0) @@ -239,6 +263,7 @@ blast_icache32_page_indexed; } else if (ic_lsize == 64) r4k_blast_icache_page_indexed = blast_icache64_page_indexed; +#endif } static void (* r4k_blast_icache)(void); @@ -322,12 +347,17 @@ */ static inline void local_r4k_flush_cache_all(void * args) { + unsigned long flags; + + local_irq_save(flags); r4k_blast_dcache(); + r4k_blast_icache(); + local_irq_restore(flags); } static void r4k_flush_cache_all(void) { - if (!cpu_has_dc_aliases) + if (!cpu_has_dc_aliases && cpu_use_kmap_coherent) return; r4k_on_each_cpu(local_r4k_flush_cache_all, NULL, 1, 1); @@ -335,6 +365,9 @@ static inline void local_r4k___flush_cache_all(void * args) { + unsigned long flags; + + local_irq_save(flags); r4k_blast_dcache(); r4k_blast_icache(); @@ -348,6 +381,7 @@ case CPU_R14000: r4k_blast_scache(); } + local_irq_restore(flags); } static void r4k___flush_cache_all(void) @@ -358,17 +392,21 @@ static inline void local_r4k_flush_cache_range(void * args) { struct vm_area_struct *vma = args; + unsigned long flags; if (!(cpu_context(smp_processor_id(), vma->vm_mm))) return; + local_irq_save(flags); r4k_blast_dcache(); + r4k_blast_icache(); + local_irq_restore(flags); } static void r4k_flush_cache_range(struct vm_area_struct *vma, unsigned long start, unsigned long end) { - if (!cpu_has_dc_aliases) + if (!cpu_has_dc_aliases && cpu_use_kmap_coherent) return; r4k_on_each_cpu(local_r4k_flush_cache_range, vma, 1, 1); @@ -377,6 +415,7 @@ static inline void local_r4k_flush_cache_mm(void * args) { struct mm_struct *mm = args; + unsigned long flags; if (!cpu_context(smp_processor_id(), mm)) return; @@ -395,12 +434,15 @@ return; } + local_irq_save(flags); r4k_blast_dcache(); + r4k_blast_icache(); + local_irq_restore(flags); } static void r4k_flush_cache_mm(struct mm_struct *mm) { - if (!cpu_has_dc_aliases) + if (!cpu_has_dc_aliases && cpu_use_kmap_coherent) return; r4k_on_each_cpu(local_r4k_flush_cache_mm, mm, 1, 1); @@ -420,6 +462,7 @@ unsigned long paddr = fcp_args->pfn << PAGE_SHIFT; int exec = vma->vm_flags & VM_EXEC; struct mm_struct *mm = vma->vm_mm; + unsigned long flags; pgd_t *pgdp; pud_t *pudp; pmd_t *pmdp; @@ -451,8 +494,9 @@ * for every cache flush operation. So we do indexed flushes * in that case, which doesn't overly flush the cache too much. */ + local_irq_save(flags); if ((mm == current->active_mm) && (pte_val(*ptep) & _PAGE_VALID)) { - if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) { + if (!cpu_use_kmap_coherent || cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) { r4k_blast_dcache_page(addr); if (exec && !cpu_icache_snoops_remote_store) r4k_blast_scache_page(addr); @@ -460,14 +504,14 @@ if (exec) r4k_blast_icache_page(addr); - return; + goto done; } /* * Do indexed flush, too much work to get the (possible) TLB refills * to work correctly. */ - if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) { + if (!cpu_use_kmap_coherent || cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) { r4k_blast_dcache_page_indexed(cpu_has_pindexed_dcache ? paddr : addr); if (exec && !cpu_icache_snoops_remote_store) { @@ -483,6 +527,8 @@ } else r4k_blast_icache_page_indexed(addr); } +done: + local_irq_restore(flags); } static void r4k_flush_cache_page(struct vm_area_struct *vma, @@ -499,7 +545,11 @@ static inline void local_r4k_flush_data_cache_page(void * addr) { + unsigned long flags; + + local_irq_save(flags); r4k_blast_dcache_page((unsigned long) addr); + local_irq_restore(flags); } static void r4k_flush_data_cache_page(unsigned long addr) @@ -542,6 +592,9 @@ static void r4k_flush_icache_range(unsigned long start, unsigned long end) { +#ifdef CONFIG_BCM947XX + r4k_flush_cache_all(); +#else struct flush_icache_range_args args; args.start = start; @@ -549,12 +602,15 @@ r4k_on_each_cpu(local_r4k_flush_icache_range, &args, 1, 1); instruction_hazard(); +#endif } #ifdef CONFIG_DMA_NONCOHERENT static void r4k_dma_cache_wback_inv(unsigned long addr, unsigned long size) { + unsigned long flags; + /* Catch bad driver code */ BUG_ON(size == 0); @@ -571,18 +627,21 @@ * subset property so we have to flush the primary caches * explicitly */ + local_irq_save(flags); if (size >= dcache_size) { r4k_blast_dcache(); } else { R4600_HIT_CACHEOP_WAR_IMPL; blast_dcache_range(addr, addr + size); } - bc_wback_inv(addr, size); + local_irq_restore(flags); } static void r4k_dma_cache_inv(unsigned long addr, unsigned long size) { + unsigned long flags; + /* Catch bad driver code */ BUG_ON(size == 0); @@ -594,6 +653,7 @@ return; } + local_irq_save(flags); if (size >= dcache_size) { r4k_blast_dcache(); } else { @@ -602,6 +662,7 @@ } bc_inv(addr, size); + local_irq_restore(flags); } #endif /* CONFIG_DMA_NONCOHERENT */ @@ -616,8 +677,12 @@ unsigned long dc_lsize = cpu_dcache_line_size(); unsigned long sc_lsize = cpu_scache_line_size(); unsigned long addr = (unsigned long) arg; + unsigned long flags; + local_irq_save(flags); R4600_HIT_CACHEOP_WAR_IMPL; + BCM4710_PROTECTED_FILL_TLB(addr); + BCM4710_PROTECTED_FILL_TLB(addr + 4); if (dc_lsize) protected_writeback_dcache_line(addr & ~(dc_lsize - 1)); if (!cpu_icache_snoops_remote_store && scache_size) @@ -644,6 +709,7 @@ } if (MIPS_CACHE_SYNC_WAR) __asm__ __volatile__ ("sync"); + local_irq_restore(flags); } static void r4k_flush_cache_sigtramp(unsigned long addr) @@ -1144,6 +1210,17 @@ * silly idea of putting something else there ... */ switch (current_cpu_data.cputype) { + case CPU_BCM3302: + { + u32 cm; + cm = read_c0_diag(); + /* Enable icache */ + cm |= (1 << 31); + /* Enable dcache */ + cm |= (1 << 30); + write_c0_diag(cm); + } + break; case CPU_R4000PC: case CPU_R4000SC: case CPU_R4000MC: @@ -1174,6 +1251,15 @@ /* Default cache error handler for R4000 and R5000 family */ set_uncached_handler (0x100, &except_vec2_generic, 0x80); + /* Check if special workarounds are required */ +#ifdef CONFIG_BCM947XX + if (current_cpu_data.cputype == CPU_BCM4710 && (current_cpu_data.processor_id & 0xff) == 0) { + printk("Enabling BCM4710A0 cache workarounds.\n"); + bcm4710 = 1; + } else +#endif + bcm4710 = 0; + probe_pcache(); setup_scache(); @@ -1219,5 +1305,13 @@ build_clear_page(); build_copy_page(); local_r4k___flush_cache_all(NULL); +#ifdef CONFIG_BCM947XX + { + static void (*_coherency_setup)(void); + _coherency_setup = (void (*)(void)) KSEG1ADDR(coherency_setup); + _coherency_setup(); + } +#else coherency_setup(); +#endif } Index: linux-2.6.22/arch/mips/mm/tlbex.c =================================================================== --- linux-2.6.22.orig/arch/mips/mm/tlbex.c 2007-07-26 06:29:40.582055658 +0200 +++ linux-2.6.22/arch/mips/mm/tlbex.c 2007-07-26 06:32:45.964620005 +0200 @@ -1229,6 +1229,10 @@ #endif } +#ifdef CONFIG_BCM947XX +extern int bcm4710; +#endif + static void __init build_r4000_tlb_refill_handler(void) { u32 *p = tlb_handler; @@ -1243,6 +1247,10 @@ memset(relocs, 0, sizeof(relocs)); memset(final_handler, 0, sizeof(final_handler)); +#ifdef CONFIG_BCM947XX + i_nop(&p); +#endif + /* * create the plain linear handler */ @@ -1736,6 +1744,9 @@ memset(labels, 0, sizeof(labels)); memset(relocs, 0, sizeof(relocs)); +#ifdef CONFIG_BCM947XX + i_nop(&p); +#endif if (bcm1250_m3_war()) { i_MFC0(&p, K0, C0_BADVADDR); i_MFC0(&p, K1, C0_ENTRYHI); Index: linux-2.6.22/include/asm-mips/r4kcache.h =================================================================== --- linux-2.6.22.orig/include/asm-mips/r4kcache.h 2007-07-26 06:29:25.085172538 +0200 +++ linux-2.6.22/include/asm-mips/r4kcache.h 2007-07-26 06:29:40.938075943 +0200 @@ -17,6 +17,20 @@ #include #include +#ifdef CONFIG_BCM947XX +#include +#include +#define BCM4710_DUMMY_RREG() ((void) *((u8 *) KSEG1ADDR(SSB_ENUM_BASE + SSB_IMSTATE))) + +#define BCM4710_FILL_TLB(addr) (*(volatile unsigned long *)(addr)) +#define BCM4710_PROTECTED_FILL_TLB(addr) ({ unsigned long x; get_dbe(x, (volatile unsigned long *)(addr)); }) +#else +#define BCM4710_DUMMY_RREG() + +#define BCM4710_FILL_TLB(addr) +#define BCM4710_PROTECTED_FILL_TLB(addr) +#endif + /* * This macro return a properly sign-extended address suitable as base address * for indexed cache operations. Two issues here: @@ -150,6 +164,7 @@ static inline void flush_dcache_line_indexed(unsigned long addr) { __dflush_prologue + BCM4710_DUMMY_RREG(); cache_op(Index_Writeback_Inv_D, addr); __dflush_epilogue } @@ -169,6 +184,7 @@ static inline void flush_dcache_line(unsigned long addr) { __dflush_prologue + BCM4710_DUMMY_RREG(); cache_op(Hit_Writeback_Inv_D, addr); __dflush_epilogue } @@ -176,6 +192,7 @@ static inline void invalidate_dcache_line(unsigned long addr) { __dflush_prologue + BCM4710_DUMMY_RREG(); cache_op(Hit_Invalidate_D, addr); __dflush_epilogue } @@ -208,6 +225,7 @@ */ static inline void protected_flush_icache_line(unsigned long addr) { + BCM4710_DUMMY_RREG(); protected_cache_op(Hit_Invalidate_I, addr); } @@ -219,6 +237,7 @@ */ static inline void protected_writeback_dcache_line(unsigned long addr) { + BCM4710_DUMMY_RREG(); protected_cache_op(Hit_Writeback_Inv_D, addr); } @@ -339,8 +358,52 @@ : "r" (base), \ "i" (op)); +static inline void blast_dcache(void) +{ + unsigned long start = KSEG0; + unsigned long dcache_size = current_cpu_data.dcache.waysize * current_cpu_data.dcache.ways; + unsigned long end = (start + dcache_size); + + do { + BCM4710_DUMMY_RREG(); + cache_op(Index_Writeback_Inv_D, start); + start += current_cpu_data.dcache.linesz; + } while(start < end); +} + +static inline void blast_dcache_page(unsigned long page) +{ + unsigned long start = page; + unsigned long end = start + PAGE_SIZE; + + BCM4710_FILL_TLB(start); + do { + BCM4710_DUMMY_RREG(); + cache_op(Hit_Writeback_Inv_D, start); + start += current_cpu_data.dcache.linesz; + } while(start < end); +} + +static inline void blast_dcache_page_indexed(unsigned long page) +{ + unsigned long start = page; + unsigned long end = start + PAGE_SIZE; + unsigned long ws_inc = 1UL << current_cpu_data.dcache.waybit; + unsigned long ws_end = current_cpu_data.dcache.ways << + current_cpu_data.dcache.waybit; + unsigned long ws, addr; + for (ws = 0; ws < ws_end; ws += ws_inc) { + start = page + ws; + for (addr = start; addr < end; addr += current_cpu_data.dcache.linesz) { + BCM4710_DUMMY_RREG(); + cache_op(Index_Writeback_Inv_D, addr); + } + } +} + + /* build blast_xxx, blast_xxx_page, blast_xxx_page_indexed */ -#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize) \ +#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize, war) \ static inline void blast_##pfx##cache##lsize(void) \ { \ unsigned long start = INDEX_BASE; \ @@ -352,6 +415,7 @@ \ __##pfx##flush_prologue \ \ + war \ for (ws = 0; ws < ws_end; ws += ws_inc) \ for (addr = start; addr < end; addr += lsize * 32) \ cache##lsize##_unroll32(addr|ws,indexop); \ @@ -366,6 +430,7 @@ \ __##pfx##flush_prologue \ \ + war \ do { \ cache##lsize##_unroll32(start,hitop); \ start += lsize * 32; \ @@ -384,6 +449,8 @@ current_cpu_data.desc.waybit; \ unsigned long ws, addr; \ \ + war \ + \ __##pfx##flush_prologue \ \ for (ws = 0; ws < ws_end; ws += ws_inc) \ @@ -393,28 +460,30 @@ __##pfx##flush_epilogue \ } -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16) -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16) -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16) -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32) -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32) -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32) -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64) -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64) -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128) +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16, ) +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16, BCM4710_FILL_TLB(start);) +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16, ) +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32, ) +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32, BCM4710_FILL_TLB(start);) +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32, ) +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64, BCM4710_FILL_TLB(start);) +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64, ) +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128, ) /* build blast_xxx_range, protected_blast_xxx_range */ -#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot) \ +#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot, war, war2) \ static inline void prot##blast_##pfx##cache##_range(unsigned long start, \ unsigned long end) \ { \ unsigned long lsize = cpu_##desc##_line_size(); \ unsigned long addr = start & ~(lsize - 1); \ unsigned long aend = (end - 1) & ~(lsize - 1); \ + war \ \ __##pfx##flush_prologue \ \ while (1) { \ + war2 \ prot##cache_op(hitop, addr); \ if (addr == aend) \ break; \ @@ -424,13 +493,13 @@ __##pfx##flush_epilogue \ } -__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_) -__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_) -__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_) -__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, ) -__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, ) +__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_, BCM4710_PROTECTED_FILL_TLB(addr); BCM4710_PROTECTED_FILL_TLB(aend);, BCM4710_DUMMY_RREG();) +__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_,, ) +__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_,, ) +__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D,, BCM4710_FILL_TLB(addr); BCM4710_FILL_TLB(aend);, BCM4710_DUMMY_RREG();) +__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD,,, ) /* blast_inv_dcache_range */ -__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D, ) -__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD, ) +__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D,,,BCM4710_DUMMY_RREG();) +__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD,,, ) #endif /* _ASM_R4KCACHE_H */ Index: linux-2.6.22/include/asm-mips/stackframe.h =================================================================== --- linux-2.6.22.orig/include/asm-mips/stackframe.h 2007-07-26 06:29:25.093172994 +0200 +++ linux-2.6.22/include/asm-mips/stackframe.h 2007-07-26 06:29:40.962077312 +0200 @@ -350,6 +350,10 @@ .macro RESTORE_SP_AND_RET LONG_L sp, PT_R29(sp) .set mips3 +#ifdef CONFIG_BCM947XX + nop + nop +#endif eret .set mips0 .endm