diff options
Diffstat (limited to 'arch/arc/mm')
-rw-r--r-- | arch/arc/mm/cache.c | 438 | ||||
-rw-r--r-- | arch/arc/mm/dma.c | 6 | ||||
-rw-r--r-- | arch/arc/mm/extable.c | 23 | ||||
-rw-r--r-- | arch/arc/mm/fault.c | 95 | ||||
-rw-r--r-- | arch/arc/mm/highmem.c | 84 | ||||
-rw-r--r-- | arch/arc/mm/init.c | 125 | ||||
-rw-r--r-- | arch/arc/mm/ioremap.c | 51 | ||||
-rw-r--r-- | arch/arc/mm/mmap.c | 41 | ||||
-rw-r--r-- | arch/arc/mm/tlb.c | 407 | ||||
-rw-r--r-- | arch/arc/mm/tlbex.S | 93 |
10 files changed, 356 insertions, 1007 deletions
diff --git a/arch/arc/mm/cache.c b/arch/arc/mm/cache.c index a2fbea3ee07c..9106ceac323c 100644 --- a/arch/arc/mm/cache.c +++ b/arch/arc/mm/cache.c @@ -28,6 +28,10 @@ int slc_enable = 1, ioc_enable = 1; unsigned long perip_base = ARC_UNCACHED_ADDR_SPACE; /* legacy value for boot */ unsigned long perip_end = 0xFFFFFFFF; /* legacy value */ +static struct cpuinfo_arc_cache { + unsigned int sz_k, line_len, colors; +} ic_info, dc_info, slc_info; + void (*_cache_line_loop_ic_fn)(phys_addr_t paddr, unsigned long vaddr, unsigned long sz, const int op, const int full_page); @@ -35,78 +39,24 @@ void (*__dma_cache_wback_inv)(phys_addr_t start, unsigned long sz); void (*__dma_cache_inv)(phys_addr_t start, unsigned long sz); void (*__dma_cache_wback)(phys_addr_t start, unsigned long sz); -char *arc_cache_mumbojumbo(int c, char *buf, int len) -{ - int n = 0; - struct cpuinfo_arc_cache *p; - -#define PR_CACHE(p, cfg, str) \ - if (!(p)->line_len) \ - n += scnprintf(buf + n, len - n, str"\t\t: N/A\n"); \ - else \ - n += scnprintf(buf + n, len - n, \ - str"\t\t: %uK, %dway/set, %uB Line, %s%s%s\n", \ - (p)->sz_k, (p)->assoc, (p)->line_len, \ - (p)->vipt ? "VIPT" : "PIPT", \ - (p)->alias ? " aliasing" : "", \ - IS_USED_CFG(cfg)); - - PR_CACHE(&cpuinfo_arc700[c].icache, CONFIG_ARC_HAS_ICACHE, "I-Cache"); - PR_CACHE(&cpuinfo_arc700[c].dcache, CONFIG_ARC_HAS_DCACHE, "D-Cache"); - - p = &cpuinfo_arc700[c].slc; - if (p->line_len) - n += scnprintf(buf + n, len - n, - "SLC\t\t: %uK, %uB Line%s\n", - p->sz_k, p->line_len, IS_USED_RUN(slc_enable)); - - n += scnprintf(buf + n, len - n, "Peripherals\t: %#lx%s%s\n", - perip_base, - IS_AVAIL3(ioc_exists, ioc_enable, ", IO-Coherency (per-device) ")); - - return buf; -} - -/* - * Read the Cache Build Confuration Registers, Decode them and save into - * the cpuinfo structure for later use. - * No Validation done here, simply read/convert the BCRs - */ -static void read_decode_cache_bcr_arcv2(int cpu) +static int read_decode_cache_bcr_arcv2(int c, char *buf, int len) { - struct cpuinfo_arc_cache *p_slc = &cpuinfo_arc700[cpu].slc; + struct cpuinfo_arc_cache *p_slc = &slc_info; + struct bcr_identity ident; struct bcr_generic sbcr; - - struct bcr_slc_cfg { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int pad:24, way:2, lsz:2, sz:4; -#else - unsigned int sz:4, lsz:2, way:2, pad:24; -#endif - } slc_cfg; - - struct bcr_clust_cfg { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int pad:7, c:1, num_entries:8, num_cores:8, ver:8; -#else - unsigned int ver:8, num_cores:8, num_entries:8, c:1, pad:7; -#endif - } cbcr; - - struct bcr_volatile { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int start:4, limit:4, pad:22, order:1, disable:1; -#else - unsigned int disable:1, order:1, pad:22, limit:4, start:4; -#endif - } vol; - + struct bcr_clust_cfg cbcr; + struct bcr_volatile vol; + int n = 0; READ_BCR(ARC_REG_SLC_BCR, sbcr); if (sbcr.ver) { + struct bcr_slc_cfg slc_cfg; READ_BCR(ARC_REG_SLC_CFG, slc_cfg); p_slc->sz_k = 128 << slc_cfg.sz; l2_line_sz = p_slc->line_len = (slc_cfg.lsz == 0) ? 128 : 64; + n += scnprintf(buf + n, len - n, + "SLC\t\t: %uK, %uB Line%s\n", + p_slc->sz_k, p_slc->line_len, IS_USED_RUN(slc_enable)); } READ_BCR(ARC_REG_CLUSTER_BCR, cbcr); @@ -129,70 +79,82 @@ static void read_decode_cache_bcr_arcv2(int cpu) ioc_enable = 0; } + READ_BCR(AUX_IDENTITY, ident); + /* HS 2.0 didn't have AUX_VOL */ - if (cpuinfo_arc700[cpu].core.family > 0x51) { + if (ident.family > 0x51) { READ_BCR(AUX_VOL, vol); perip_base = vol.start << 28; /* HS 3.0 has limit and strict-ordering fields */ - if (cpuinfo_arc700[cpu].core.family > 0x52) + if (ident.family > 0x52) perip_end = (vol.limit << 28) - 1; } + + n += scnprintf(buf + n, len - n, "Peripherals\t: %#lx%s%s\n", + perip_base, + IS_AVAIL3(ioc_exists, ioc_enable, ", IO-Coherency (per-device) ")); + + return n; } -void read_decode_cache_bcr(void) +int arc_cache_mumbojumbo(int c, char *buf, int len) { - struct cpuinfo_arc_cache *p_ic, *p_dc; - unsigned int cpu = smp_processor_id(); - struct bcr_cache { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int pad:12, line_len:4, sz:4, config:4, ver:8; -#else - unsigned int ver:8, config:4, sz:4, line_len:4, pad:12; -#endif - } ibcr, dbcr; + struct cpuinfo_arc_cache *p_ic = &ic_info, *p_dc = &dc_info; + struct bcr_cache ibcr, dbcr; + int vipt, assoc; + int n = 0; - p_ic = &cpuinfo_arc700[cpu].icache; READ_BCR(ARC_REG_IC_BCR, ibcr); - if (!ibcr.ver) goto dc_chk; - if (ibcr.ver <= 3) { + if (is_isa_arcompact() && (ibcr.ver <= 3)) { BUG_ON(ibcr.config != 3); - p_ic->assoc = 2; /* Fixed to 2w set assoc */ - } else if (ibcr.ver >= 4) { - p_ic->assoc = 1 << ibcr.config; /* 1,2,4,8 */ + assoc = 2; /* Fixed to 2w set assoc */ + } else if (is_isa_arcv2() && (ibcr.ver >= 4)) { + assoc = 1 << ibcr.config; /* 1,2,4,8 */ } p_ic->line_len = 8 << ibcr.line_len; p_ic->sz_k = 1 << (ibcr.sz - 1); - p_ic->vipt = 1; - p_ic->alias = p_ic->sz_k/p_ic->assoc/TO_KB(PAGE_SIZE) > 1; + p_ic->colors = p_ic->sz_k/assoc/TO_KB(PAGE_SIZE); + + n += scnprintf(buf + n, len - n, + "I-Cache\t\t: %uK, %dway/set, %uB Line, VIPT%s%s\n", + p_ic->sz_k, assoc, p_ic->line_len, + p_ic->colors > 1 ? " aliasing" : "", + IS_USED_CFG(CONFIG_ARC_HAS_ICACHE)); dc_chk: - p_dc = &cpuinfo_arc700[cpu].dcache; READ_BCR(ARC_REG_DC_BCR, dbcr); - if (!dbcr.ver) goto slc_chk; - if (dbcr.ver <= 3) { + if (is_isa_arcompact() && (dbcr.ver <= 3)) { BUG_ON(dbcr.config != 2); - p_dc->assoc = 4; /* Fixed to 4w set assoc */ - p_dc->vipt = 1; - p_dc->alias = p_dc->sz_k/p_dc->assoc/TO_KB(PAGE_SIZE) > 1; - } else if (dbcr.ver >= 4) { - p_dc->assoc = 1 << dbcr.config; /* 1,2,4,8 */ - p_dc->vipt = 0; - p_dc->alias = 0; /* PIPT so can't VIPT alias */ + vipt = 1; + assoc = 4; /* Fixed to 4w set assoc */ + p_dc->colors = p_dc->sz_k/assoc/TO_KB(PAGE_SIZE); + } else if (is_isa_arcv2() && (dbcr.ver >= 4)) { + vipt = 0; + assoc = 1 << dbcr.config; /* 1,2,4,8 */ + p_dc->colors = 1; /* PIPT so can't VIPT alias */ } p_dc->line_len = 16 << dbcr.line_len; p_dc->sz_k = 1 << (dbcr.sz - 1); + n += scnprintf(buf + n, len - n, + "D-Cache\t\t: %uK, %dway/set, %uB Line, %s%s\n", + p_dc->sz_k, assoc, p_dc->line_len, + vipt ? "VIPT" : "PIPT", + IS_USED_CFG(CONFIG_ARC_HAS_DCACHE)); + slc_chk: if (is_isa_arcv2()) - read_decode_cache_bcr_arcv2(cpu); + n += read_decode_cache_bcr_arcv2(c, buf + n, len - n); + + return n; } /* @@ -205,93 +167,24 @@ slc_chk: #define OP_INV_IC 0x4 /* - * I-Cache Aliasing in ARC700 VIPT caches (MMU v1-v3) - * - * ARC VIPT I-cache uses vaddr to index into cache and paddr to match the tag. - * The orig Cache Management Module "CDU" only required paddr to invalidate a - * certain line since it sufficed as index in Non-Aliasing VIPT cache-geometry. - * Infact for distinct V1,V2,P: all of {V1-P},{V2-P},{P-P} would end up fetching - * the exact same line. + * Cache Flush programming model * - * However for larger Caches (way-size > page-size) - i.e. in Aliasing config, - * paddr alone could not be used to correctly index the cache. + * ARC700 MMUv3 I$ and D$ are both VIPT and can potentially alias. + * Programming model requires both paddr and vaddr irrespecive of aliasing + * considerations: + * - vaddr in {I,D}C_IV?L + * - paddr in {I,D}C_PTAG * - * ------------------ - * MMU v1/v2 (Fixed Page Size 8k) - * ------------------ - * The solution was to provide CDU with these additonal vaddr bits. These - * would be bits [x:13], x would depend on cache-geometry, 13 comes from - * standard page size of 8k. - * H/w folks chose [17:13] to be a future safe range, and moreso these 5 bits - * of vaddr could easily be "stuffed" in the paddr as bits [4:0] since the - * orig 5 bits of paddr were anyways ignored by CDU line ops, as they - * represent the offset within cache-line. The adv of using this "clumsy" - * interface for additional info was no new reg was needed in CDU programming - * model. + * In HS38x (MMUv4), D$ is PIPT, I$ is VIPT and can still alias. + * Programming model is different for aliasing vs. non-aliasing I$ + * - D$ / Non-aliasing I$: only paddr in {I,D}C_IV?L + * - Aliasing I$: same as ARC700 above (so MMUv3 routine used for MMUv4 I$) * - * 17:13 represented the max num of bits passable, actual bits needed were - * fewer, based on the num-of-aliases possible. - * -for 2 alias possibility, only bit 13 needed (32K cache) - * -for 4 alias possibility, bits 14:13 needed (64K cache) - * - * ------------------ - * MMU v3 - * ------------------ - * This ver of MMU supports variable page sizes (1k-16k): although Linux will - * only support 8k (default), 16k and 4k. - * However from hardware perspective, smaller page sizes aggravate aliasing - * meaning more vaddr bits needed to disambiguate the cache-line-op ; - * the existing scheme of piggybacking won't work for certain configurations. - * Two new registers IC_PTAG and DC_PTAG inttoduced. - * "tag" bits are provided in PTAG, index bits in existing IVIL/IVDL/FLDL regs + * - If PAE40 is enabled, independent of aliasing considerations, the higher + * bits needs to be written into PTAG_HI */ static inline -void __cache_line_loop_v2(phys_addr_t paddr, unsigned long vaddr, - unsigned long sz, const int op, const int full_page) -{ - unsigned int aux_cmd; - int num_lines; - - if (op == OP_INV_IC) { - aux_cmd = ARC_REG_IC_IVIL; - } else { - /* d$ cmd: INV (discard or wback-n-discard) OR FLUSH (wback) */ - aux_cmd = op & OP_INV ? ARC_REG_DC_IVDL : ARC_REG_DC_FLDL; - } - - /* Ensure we properly floor/ceil the non-line aligned/sized requests - * and have @paddr - aligned to cache line and integral @num_lines. - * This however can be avoided for page sized since: - * -@paddr will be cache-line aligned already (being page aligned) - * -@sz will be integral multiple of line size (being page sized). - */ - if (!full_page) { - sz += paddr & ~CACHE_LINE_MASK; - paddr &= CACHE_LINE_MASK; - vaddr &= CACHE_LINE_MASK; - } - - num_lines = DIV_ROUND_UP(sz, L1_CACHE_BYTES); - - /* MMUv2 and before: paddr contains stuffed vaddrs bits */ - paddr |= (vaddr >> PAGE_SHIFT) & 0x1F; - - while (num_lines-- > 0) { - write_aux_reg(aux_cmd, paddr); - paddr += L1_CACHE_BYTES; - } -} - -/* - * For ARC700 MMUv3 I-cache and D-cache flushes - * - ARC700 programming model requires paddr and vaddr be passed in seperate - * AUX registers (*_IV*L and *_PTAG respectively) irrespective of whether the - * caches actually alias or not. - * - For HS38, only the aliasing I-cache configuration uses the PTAG reg - * (non aliasing I-cache version doesn't; while D-cache can't possibly alias) - */ -static inline void __cache_line_loop_v3(phys_addr_t paddr, unsigned long vaddr, unsigned long sz, const int op, const int full_page) { @@ -350,17 +243,6 @@ void __cache_line_loop_v3(phys_addr_t paddr, unsigned long vaddr, #ifndef USE_RGN_FLSH /* - * In HS38x (MMU v4), I-cache is VIPT (can alias), D-cache is PIPT - * Here's how cache ops are implemented - * - * - D-cache: only paddr needed (in DC_IVDL/DC_FLDL) - * - I-cache Non Aliasing: Despite VIPT, only paddr needed (in IC_IVIL) - * - I-cache Aliasing: Both vaddr and paddr needed (in IC_IVIL, IC_PTAG - * respectively, similar to MMU v3 programming model, hence - * __cache_line_loop_v3() is used) - * - * If PAE40 is enabled, independent of aliasing considerations, the higher bits - * needs to be written into PTAG_HI */ static inline void __cache_line_loop_v4(phys_addr_t paddr, unsigned long vaddr, @@ -460,11 +342,9 @@ void __cache_line_loop_v4(phys_addr_t paddr, unsigned long vaddr, #endif -#if (CONFIG_ARC_MMU_VER < 3) -#define __cache_line_loop __cache_line_loop_v2 -#elif (CONFIG_ARC_MMU_VER == 3) +#ifdef CONFIG_ARC_MMU_V3 #define __cache_line_loop __cache_line_loop_v3 -#elif (CONFIG_ARC_MMU_VER > 3) +#else #define __cache_line_loop __cache_line_loop_v4 #endif @@ -483,7 +363,7 @@ static inline void __before_dc_op(const int op) { if (op == OP_FLUSH_N_INV) { /* Dcache provides 2 cmd: FLUSH or INV - * INV inturn has sub-modes: DISCARD or FLUSH-BEFORE + * INV in turn has sub-modes: DISCARD or FLUSH-BEFORE * flush-n-inv is achieved by INV cmd but with IM=1 * So toggle INV sub-mode depending on op request and default */ @@ -663,7 +543,7 @@ static void __ic_line_inv_vaddr(phys_addr_t paddr, unsigned long vaddr, #endif /* CONFIG_ARC_HAS_ICACHE */ -noinline void slc_op_rgn(phys_addr_t paddr, unsigned long sz, const int op) +static noinline void slc_op_rgn(phys_addr_t paddr, unsigned long sz, const int op) { #ifdef CONFIG_ISA_ARCV2 /* @@ -726,7 +606,7 @@ noinline void slc_op_rgn(phys_addr_t paddr, unsigned long sz, const int op) #endif } -noinline void slc_op_line(phys_addr_t paddr, unsigned long sz, const int op) +static __maybe_unused noinline void slc_op_line(phys_addr_t paddr, unsigned long sz, const int op) { #ifdef CONFIG_ISA_ARCV2 /* @@ -822,47 +702,16 @@ static inline void arc_slc_enable(void) * Exported APIs */ -/* - * Handle cache congruency of kernel and userspace mappings of page when kernel - * writes-to/reads-from - * - * The idea is to defer flushing of kernel mapping after a WRITE, possible if: - * -dcache is NOT aliasing, hence any U/K-mappings of page are congruent - * -U-mapping doesn't exist yet for page (finalised in update_mmu_cache) - * -In SMP, if hardware caches are coherent - * - * There's a corollary case, where kernel READs from a userspace mapped page. - * If the U-mapping is not congruent to to K-mapping, former needs flushing. - */ -void flush_dcache_page(struct page *page) +void flush_dcache_folio(struct folio *folio) { - struct address_space *mapping; - - if (!cache_is_vipt_aliasing()) { - clear_bit(PG_dc_clean, &page->flags); - return; - } - - /* don't handle anon pages here */ - mapping = page_mapping_file(page); - if (!mapping) - return; - - /* - * pagecache page, file not yet mapped to userspace - * Make a note that K-mapping is dirty - */ - if (!mapping_mapped(mapping)) { - clear_bit(PG_dc_clean, &page->flags); - } else if (page_mapcount(page)) { - - /* kernel reading from page with U-mapping */ - phys_addr_t paddr = (unsigned long)page_address(page); - unsigned long vaddr = page->index << PAGE_SHIFT; + clear_bit(PG_dc_clean, &folio->flags); + return; +} +EXPORT_SYMBOL(flush_dcache_folio); - if (addr_not_cache_congruent(paddr, vaddr)) - __flush_dcache_page(paddr, vaddr); - } +void flush_dcache_page(struct page *page) +{ + return flush_dcache_folio(page_folio(page)); } EXPORT_SYMBOL(flush_dcache_page); @@ -992,7 +841,7 @@ EXPORT_SYMBOL(flush_icache_range); * @vaddr is typically user vaddr (breakpoint) or kernel vaddr (vmalloc) * However in one instance, when called by kprobe (for a breakpt in * builtin kernel code) @vaddr will be paddr only, meaning CDU operation will - * use a paddr to index the cache (despite VIPT). This is fine since since a + * use a paddr to index the cache (despite VIPT). This is fine since a * builtin kernel page will not have any virtual mappings. * kprobe on loadable module will be kernel vaddr. */ @@ -1003,18 +852,18 @@ void __sync_icache_dcache(phys_addr_t paddr, unsigned long vaddr, int len) } /* wrapper to compile time eliminate alignment checks in flush loop */ -void __inv_icache_page(phys_addr_t paddr, unsigned long vaddr) +void __inv_icache_pages(phys_addr_t paddr, unsigned long vaddr, unsigned nr) { - __ic_line_inv_vaddr(paddr, vaddr, PAGE_SIZE); + __ic_line_inv_vaddr(paddr, vaddr, nr * PAGE_SIZE); } /* * wrapper to clearout kernel or userspace mappings of a page * For kernel mappings @vaddr == @paddr */ -void __flush_dcache_page(phys_addr_t paddr, unsigned long vaddr) +void __flush_dcache_pages(phys_addr_t paddr, unsigned long vaddr, unsigned nr) { - __dc_line_op(paddr, vaddr & PAGE_MASK, PAGE_SIZE, OP_FLUSH_N_INV); + __dc_line_op(paddr, vaddr & PAGE_MASK, nr * PAGE_SIZE, OP_FLUSH_N_INV); } noinline void flush_cache_all(void) @@ -1030,89 +879,18 @@ noinline void flush_cache_all(void) } -#ifdef CONFIG_ARC_CACHE_VIPT_ALIASING - -void flush_cache_mm(struct mm_struct *mm) -{ - flush_cache_all(); -} - -void flush_cache_page(struct vm_area_struct *vma, unsigned long u_vaddr, - unsigned long pfn) -{ - phys_addr_t paddr = pfn << PAGE_SHIFT; - - u_vaddr &= PAGE_MASK; - - __flush_dcache_page(paddr, u_vaddr); - - if (vma->vm_flags & VM_EXEC) - __inv_icache_page(paddr, u_vaddr); -} - -void flush_cache_range(struct vm_area_struct *vma, unsigned long start, - unsigned long end) -{ - flush_cache_all(); -} - -void flush_anon_page(struct vm_area_struct *vma, struct page *page, - unsigned long u_vaddr) -{ - /* TBD: do we really need to clear the kernel mapping */ - __flush_dcache_page((phys_addr_t)page_address(page), u_vaddr); - __flush_dcache_page((phys_addr_t)page_address(page), - (phys_addr_t)page_address(page)); - -} - -#endif - void copy_user_highpage(struct page *to, struct page *from, unsigned long u_vaddr, struct vm_area_struct *vma) { + struct folio *src = page_folio(from); + struct folio *dst = page_folio(to); void *kfrom = kmap_atomic(from); void *kto = kmap_atomic(to); - int clean_src_k_mappings = 0; - - /* - * If SRC page was already mapped in userspace AND it's U-mapping is - * not congruent with K-mapping, sync former to physical page so that - * K-mapping in memcpy below, sees the right data - * - * Note that while @u_vaddr refers to DST page's userspace vaddr, it is - * equally valid for SRC page as well - * - * For !VIPT cache, all of this gets compiled out as - * addr_not_cache_congruent() is 0 - */ - if (page_mapcount(from) && addr_not_cache_congruent(kfrom, u_vaddr)) { - __flush_dcache_page((unsigned long)kfrom, u_vaddr); - clean_src_k_mappings = 1; - } copy_page(kto, kfrom); - /* - * Mark DST page K-mapping as dirty for a later finalization by - * update_mmu_cache(). Although the finalization could have been done - * here as well (given that both vaddr/paddr are available). - * But update_mmu_cache() already has code to do that for other - * non copied user pages (e.g. read faults which wire in pagecache page - * directly). - */ - clear_bit(PG_dc_clean, &to->flags); - - /* - * if SRC was already usermapped and non-congruent to kernel mapping - * sync the kernel mapping back to physical page - */ - if (clean_src_k_mappings) { - __flush_dcache_page((unsigned long)kfrom, (unsigned long)kfrom); - set_bit(PG_dc_clean, &from->flags); - } else { - clear_bit(PG_dc_clean, &from->flags); - } + clear_bit(PG_dc_clean, &dst->flags); + clear_bit(PG_dc_clean, &src->flags); kunmap_atomic(kto); kunmap_atomic(kfrom); @@ -1120,10 +898,11 @@ void copy_user_highpage(struct page *to, struct page *from, void clear_user_page(void *to, unsigned long u_vaddr, struct page *page) { + struct folio *folio = page_folio(page); clear_page(to); - clear_bit(PG_dc_clean, &page->flags); + clear_bit(PG_dc_clean, &folio->flags); } - +EXPORT_SYMBOL(clear_user_page); /********************************************************************** * Explicit Cache flush request from user space via syscall @@ -1151,7 +930,7 @@ SYSCALL_DEFINE3(cacheflush, uint32_t, start, uint32_t, sz, uint32_t, flags) * 3. All Caches need to be disabled when setting up IOC to elide any in-flight * Coherency transactions */ -noinline void __init arc_ioc_setup(void) +static noinline void __init arc_ioc_setup(void) { unsigned int ioc_base, mem_sz; @@ -1213,12 +992,10 @@ noinline void __init arc_ioc_setup(void) * one core suffices for all * - IOC setup / dma callbacks only need to be done once */ -void __init arc_cache_init_master(void) +static noinline void __init arc_cache_init_master(void) { - unsigned int __maybe_unused cpu = smp_processor_id(); - if (IS_ENABLED(CONFIG_ARC_HAS_ICACHE)) { - struct cpuinfo_arc_cache *ic = &cpuinfo_arc700[cpu].icache; + struct cpuinfo_arc_cache *ic = &ic_info; if (!ic->line_len) panic("cache support enabled but non-existent cache\n"); @@ -1231,14 +1008,14 @@ void __init arc_cache_init_master(void) * In MMU v4 (HS38x) the aliasing icache config uses IVIL/PTAG * pair to provide vaddr/paddr respectively, just as in MMU v3 */ - if (is_isa_arcv2() && ic->alias) + if (is_isa_arcv2() && ic->colors > 1) _cache_line_loop_ic_fn = __cache_line_loop_v3; else _cache_line_loop_ic_fn = __cache_line_loop; } if (IS_ENABLED(CONFIG_ARC_HAS_DCACHE)) { - struct cpuinfo_arc_cache *dc = &cpuinfo_arc700[cpu].dcache; + struct cpuinfo_arc_cache *dc = &dc_info; if (!dc->line_len) panic("cache support enabled but non-existent cache\n"); @@ -1248,18 +1025,8 @@ void __init arc_cache_init_master(void) dc->line_len, L1_CACHE_BYTES); /* check for D-Cache aliasing on ARCompact: ARCv2 has PIPT */ - if (is_isa_arcompact()) { - int handled = IS_ENABLED(CONFIG_ARC_CACHE_VIPT_ALIASING); - int num_colors = dc->sz_k/dc->assoc/TO_KB(PAGE_SIZE); - - if (dc->alias) { - if (!handled) - panic("Enable CONFIG_ARC_CACHE_VIPT_ALIASING\n"); - if (CACHE_COLORS_NUM != num_colors) - panic("CACHE_COLORS_NUM not optimized for config\n"); - } else if (!dc->alias && handled) { - panic("Disable CONFIG_ARC_CACHE_VIPT_ALIASING\n"); - } + if (is_isa_arcompact() && dc->colors > 1) { + panic("Aliasing VIPT cache not supported\n"); } } @@ -1300,9 +1067,6 @@ void __init arc_cache_init_master(void) void __ref arc_cache_init(void) { unsigned int __maybe_unused cpu = smp_processor_id(); - char str[256]; - - pr_info("%s", arc_cache_mumbojumbo(0, str, sizeof(str))); if (!cpu) arc_cache_init_master(); diff --git a/arch/arc/mm/dma.c b/arch/arc/mm/dma.c index e947572a521e..197707bc7658 100644 --- a/arch/arc/mm/dma.c +++ b/arch/arc/mm/dma.c @@ -3,7 +3,7 @@ * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com) */ -#include <linux/dma-noncoherent.h> +#include <linux/dma-map-ops.h> #include <asm/cache.h> #include <asm/cacheflush.h> @@ -32,7 +32,7 @@ void arch_dma_prep_coherent(struct page *page, size_t size) /* * Cache operations depending on function and direction argument, inspired by - * https://lkml.org/lkml/2018/5/18/979 + * https://lore.kernel.org/lkml/20180518175004.GF17671@n2100.armlinux.org.uk * "dma_sync_*_for_cpu and direction=TO_DEVICE (was Re: [PATCH 02/20] * dma-mapping: provide a generic dma-noncoherent implementation)" * @@ -91,7 +91,7 @@ void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size, * Plug in direct dma map ops. */ void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size, - const struct iommu_ops *iommu, bool coherent) + bool coherent) { /* * IOC hardware snoops all DMA traffic keeping the caches consistent diff --git a/arch/arc/mm/extable.c b/arch/arc/mm/extable.c index b06b09ddf924..88fa3a4d4906 100644 --- a/arch/arc/mm/extable.c +++ b/arch/arc/mm/extable.c @@ -22,26 +22,3 @@ int fixup_exception(struct pt_regs *regs) return 0; } - -#ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE - -unsigned long arc_clear_user_noinline(void __user *to, - unsigned long n) -{ - return __arc_clear_user(to, n); -} -EXPORT_SYMBOL(arc_clear_user_noinline); - -long arc_strncpy_from_user_noinline(char *dst, const char __user *src, - long count) -{ - return __arc_strncpy_from_user(dst, src, count); -} -EXPORT_SYMBOL(arc_strncpy_from_user_noinline); - -long arc_strnlen_user_noinline(const char __user *src, long n) -{ - return __arc_strnlen_user(src, n); -} -EXPORT_SYMBOL(arc_strnlen_user_noinline); -#endif diff --git a/arch/arc/mm/fault.c b/arch/arc/mm/fault.c index fb86bc3e9b35..95119a5e7761 100644 --- a/arch/arc/mm/fault.c +++ b/arch/arc/mm/fault.c @@ -13,7 +13,7 @@ #include <linux/kdebug.h> #include <linux/perf_event.h> #include <linux/mm_types.h> -#include <asm/pgalloc.h> +#include <asm/entry.h> #include <asm/mmu.h> /* @@ -34,28 +34,34 @@ noinline static int handle_kernel_vaddr_fault(unsigned long address) pud_t *pud, *pud_k; pmd_t *pmd, *pmd_k; - pgd = pgd_offset_fast(current->active_mm, address); + pgd = pgd_offset(current->active_mm, address); pgd_k = pgd_offset_k(address); - if (!pgd_present(*pgd_k)) + if (pgd_none (*pgd_k)) goto bad_area; + if (!pgd_present(*pgd)) + set_pgd(pgd, *pgd_k); p4d = p4d_offset(pgd, address); p4d_k = p4d_offset(pgd_k, address); - if (!p4d_present(*p4d_k)) + if (p4d_none(*p4d_k)) goto bad_area; + if (!p4d_present(*p4d)) + set_p4d(p4d, *p4d_k); pud = pud_offset(p4d, address); pud_k = pud_offset(p4d_k, address); - if (!pud_present(*pud_k)) + if (pud_none(*pud_k)) goto bad_area; + if (!pud_present(*pud)) + set_pud(pud, *pud_k); pmd = pmd_offset(pud, address); pmd_k = pmd_offset(pud_k, address); - if (!pmd_present(*pmd_k)) + if (pmd_none(*pmd_k)) goto bad_area; - - set_pmd(pmd, *pmd_k); + if (!pmd_present(*pmd)) + set_pmd(pmd, *pmd_k); /* XXX: create the TLB entry here */ return 0; @@ -94,28 +100,23 @@ void do_page_fault(unsigned long address, struct pt_regs *regs) if (faulthandler_disabled() || !mm) goto no_context; - if (regs->ecr_cause & ECR_C_PROTV_STORE) /* ST/EX */ + if (regs->ecr.cause & ECR_C_PROTV_STORE) /* ST/EX */ write = 1; - else if ((regs->ecr_vec == ECR_V_PROTV) && - (regs->ecr_cause == ECR_C_PROTV_INST_FETCH)) + else if ((regs->ecr.vec == ECR_V_PROTV) && + (regs->ecr.cause == ECR_C_PROTV_INST_FETCH)) exec = 1; - flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; + flags = FAULT_FLAG_DEFAULT; if (user_mode(regs)) flags |= FAULT_FLAG_USER; if (write) flags |= FAULT_FLAG_WRITE; + perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); retry: - down_read(&mm->mmap_sem); - - vma = find_vma(mm, address); + vma = lock_mm_and_find_vma(mm, address, regs); if (!vma) - goto bad_area; - if (unlikely(address < vma->vm_start)) { - if (!(vma->vm_flags & VM_GROWSDOWN) || expand_stack(vma, address)) - goto bad_area; - } + goto bad_area_nosemaphore; /* * vm_area is good, now check permissions for this memory access @@ -131,56 +132,38 @@ retry: goto bad_area; } - fault = handle_mm_fault(vma, address, flags); + fault = handle_mm_fault(vma, address, flags, regs); + + /* Quick path to respond to signals */ + if (fault_signal_pending(fault, regs)) { + if (!user_mode(regs)) + goto no_context; + return; + } + + /* The fault is fully completed (including releasing mmap lock) */ + if (fault & VM_FAULT_COMPLETED) + return; /* - * Fault retry nuances + * Fault retry nuances, mmap_lock already relinquished by core mm */ if (unlikely(fault & VM_FAULT_RETRY)) { - - /* - * If fault needs to be retried, handle any pending signals - * first (by returning to user mode). - * mmap_sem already relinquished by core mm for RETRY case - */ - if (fatal_signal_pending(current)) { - if (!user_mode(regs)) - goto no_context; - return; - } - /* - * retry state machine - */ - if (flags & FAULT_FLAG_ALLOW_RETRY) { - flags &= ~FAULT_FLAG_ALLOW_RETRY; - flags |= FAULT_FLAG_TRIED; - goto retry; - } + flags |= FAULT_FLAG_TRIED; + goto retry; } bad_area: - up_read(&mm->mmap_sem); + mmap_read_unlock(mm); +bad_area_nosemaphore: /* * Major/minor page fault accounting * (in case of retry we only land here once) */ - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); - - if (likely(!(fault & VM_FAULT_ERROR))) { - if (fault & VM_FAULT_MAJOR) { - tsk->maj_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, - regs, address); - } else { - tsk->min_flt++; - perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, - regs, address); - } - + if (likely(!(fault & VM_FAULT_ERROR))) /* Normal return path: fault Handled Gracefully */ return; - } if (!user_mode(regs)) goto no_context; diff --git a/arch/arc/mm/highmem.c b/arch/arc/mm/highmem.c index fc8849e4f72e..c79912a6b196 100644 --- a/arch/arc/mm/highmem.c +++ b/arch/arc/mm/highmem.c @@ -6,8 +6,8 @@ #include <linux/memblock.h> #include <linux/export.h> #include <linux/highmem.h> +#include <linux/pgtable.h> #include <asm/processor.h> -#include <asm/pgtable.h> #include <asm/pgalloc.h> #include <asm/tlbflush.h> @@ -36,9 +36,8 @@ * This means each only has 1 PGDIR_SIZE worth of kvaddr mappings, which means * 2M of kvaddr space for typical config (8K page and 11:8:13 traversal split) * - * - fixmap anyhow needs a limited number of mappings. So 2M kvaddr == 256 PTE - * slots across NR_CPUS would be more than sufficient (generic code defines - * KM_TYPE_NR as 20). + * - The fixed KMAP slots for kmap_local/atomic() require KM_MAX_IDX slots per + * CPU. So the number of CPUs sharing a single PTE page is limited. * * - pkmap being preemptible, in theory could do with more than 256 concurrent * mappings. However, generic pkmap code: map_new_virtual(), doesn't traverse @@ -47,80 +46,12 @@ */ extern pte_t * pkmap_page_table; -static pte_t * fixmap_page_table; - -void *kmap(struct page *page) -{ - BUG_ON(in_interrupt()); - if (!PageHighMem(page)) - return page_address(page); - - return kmap_high(page); -} -EXPORT_SYMBOL(kmap); - -void *kmap_atomic(struct page *page) -{ - int idx, cpu_idx; - unsigned long vaddr; - - preempt_disable(); - pagefault_disable(); - if (!PageHighMem(page)) - return page_address(page); - - cpu_idx = kmap_atomic_idx_push(); - idx = cpu_idx + KM_TYPE_NR * smp_processor_id(); - vaddr = FIXMAP_ADDR(idx); - - set_pte_at(&init_mm, vaddr, fixmap_page_table + idx, - mk_pte(page, kmap_prot)); - - return (void *)vaddr; -} -EXPORT_SYMBOL(kmap_atomic); - -void __kunmap_atomic(void *kv) -{ - unsigned long kvaddr = (unsigned long)kv; - - if (kvaddr >= FIXMAP_BASE && kvaddr < (FIXMAP_BASE + FIXMAP_SIZE)) { - - /* - * Because preemption is disabled, this vaddr can be associated - * with the current allocated index. - * But in case of multiple live kmap_atomic(), it still relies on - * callers to unmap in right order. - */ - int cpu_idx = kmap_atomic_idx(); - int idx = cpu_idx + KM_TYPE_NR * smp_processor_id(); - - WARN_ON(kvaddr != FIXMAP_ADDR(idx)); - - pte_clear(&init_mm, kvaddr, fixmap_page_table + idx); - local_flush_tlb_kernel_range(kvaddr, kvaddr + PAGE_SIZE); - - kmap_atomic_idx_pop(); - } - - pagefault_enable(); - preempt_enable(); -} -EXPORT_SYMBOL(__kunmap_atomic); static noinline pte_t * __init alloc_kmap_pgtable(unsigned long kvaddr) { - pgd_t *pgd_k; - p4d_t *p4d_k; - pud_t *pud_k; - pmd_t *pmd_k; + pmd_t *pmd_k = pmd_off_k(kvaddr); pte_t *pte_k; - pgd_k = pgd_offset_k(kvaddr); - p4d_k = p4d_offset(pgd_k, kvaddr); - pud_k = pud_offset(p4d_k, kvaddr); - pmd_k = pmd_offset(pud_k, kvaddr); - pte_k = (pte_t *)memblock_alloc_low(PAGE_SIZE, PAGE_SIZE); if (!pte_k) panic("%s: Failed to allocate %lu bytes align=0x%lx\n", @@ -134,10 +65,9 @@ void __init kmap_init(void) { /* Due to recursive include hell, we can't do this in processor.h */ BUILD_BUG_ON(PAGE_OFFSET < (VMALLOC_END + FIXMAP_SIZE + PKMAP_SIZE)); + BUILD_BUG_ON(LAST_PKMAP > PTRS_PER_PTE); + BUILD_BUG_ON(FIX_KMAP_SLOTS > PTRS_PER_PTE); - BUILD_BUG_ON(KM_TYPE_NR > PTRS_PER_PTE); pkmap_page_table = alloc_kmap_pgtable(PKMAP_BASE); - - BUILD_BUG_ON(LAST_PKMAP > PTRS_PER_PTE); - fixmap_page_table = alloc_kmap_pgtable(FIXMAP_BASE); + alloc_kmap_pgtable(FIXMAP_BASE); } diff --git a/arch/arc/mm/init.c b/arch/arc/mm/init.c index 0920c969c466..6a71b23f1383 100644 --- a/arch/arc/mm/init.c +++ b/arch/arc/mm/init.c @@ -14,8 +14,8 @@ #include <linux/module.h> #include <linux/highmem.h> #include <asm/page.h> -#include <asm/pgalloc.h> #include <asm/sections.h> +#include <asm/setup.h> #include <asm/arcregs.h> pgd_t swapper_pg_dir[PTRS_PER_PGD] __aligned(PAGE_SIZE); @@ -27,13 +27,10 @@ static unsigned long low_mem_sz; #ifdef CONFIG_HIGHMEM static unsigned long min_high_pfn, max_high_pfn; -static u64 high_mem_start; -static u64 high_mem_sz; -#endif - -#ifdef CONFIG_DISCONTIGMEM -struct pglist_data node_data[MAX_NUMNODES] __read_mostly; -EXPORT_SYMBOL(node_data); +static phys_addr_t high_mem_start; +static phys_addr_t high_mem_sz; +unsigned long arch_pfn_offset; +EXPORT_SYMBOL(arch_pfn_offset); #endif long __init arc_get_mem_sz(void) @@ -63,11 +60,14 @@ void __init early_init_dt_add_memory_arch(u64 base, u64 size) low_mem_sz = size; in_use = 1; + memblock_add_node(base, size, 0, MEMBLOCK_NONE); } else { #ifdef CONFIG_HIGHMEM high_mem_start = base; high_mem_sz = size; in_use = 1; + memblock_add_node(base, size, 1, MEMBLOCK_NONE); + memblock_reserve(base, size); #endif } @@ -83,25 +83,16 @@ void __init early_init_dt_add_memory_arch(u64 base, u64 size) */ void __init setup_arch_memory(void) { - unsigned long zones_size[MAX_NR_ZONES]; - unsigned long zones_holes[MAX_NR_ZONES]; + unsigned long max_zone_pfn[MAX_NR_ZONES] = { 0 }; - init_mm.start_code = (unsigned long)_text; - init_mm.end_code = (unsigned long)_etext; - init_mm.end_data = (unsigned long)_edata; - init_mm.brk = (unsigned long)_end; + setup_initial_init_mm(_text, _etext, _edata, _end); /* first page of system - kernel .vector starts here */ - min_low_pfn = ARCH_PFN_OFFSET; + min_low_pfn = virt_to_pfn((void *)CONFIG_LINUX_RAM_BASE); /* Last usable page of low mem */ max_low_pfn = max_pfn = PFN_DOWN(low_mem_start + low_mem_sz); -#ifdef CONFIG_FLATMEM - /* pfn_valid() uses this */ - max_mapnr = max_low_pfn - min_low_pfn; -#endif - /*------------- bootmem allocator setup -----------------------*/ /* @@ -115,7 +106,6 @@ void __init setup_arch_memory(void) * the crash */ - memblock_add_node(low_mem_start, low_mem_sz, 0); memblock_reserve(CONFIG_LINUX_LINK_BASE, __pa(_end) - CONFIG_LINUX_LINK_BASE); @@ -133,54 +123,55 @@ void __init setup_arch_memory(void) memblock_dump_all(); /*----------------- node/zones setup --------------------------*/ - memset(zones_size, 0, sizeof(zones_size)); - memset(zones_holes, 0, sizeof(zones_holes)); - - zones_size[ZONE_NORMAL] = max_low_pfn - min_low_pfn; - zones_holes[ZONE_NORMAL] = 0; - - /* - * We can't use the helper free_area_init(zones[]) because it uses - * PAGE_OFFSET to compute the @min_low_pfn which would be wrong - * when our kernel doesn't start at PAGE_OFFSET, i.e. - * PAGE_OFFSET != CONFIG_LINUX_RAM_BASE - */ - free_area_init_node(0, /* node-id */ - zones_size, /* num pages per zone */ - min_low_pfn, /* first pfn of node */ - zones_holes); /* holes */ + max_zone_pfn[ZONE_NORMAL] = max_low_pfn; #ifdef CONFIG_HIGHMEM /* - * Populate a new node with highmem - * * On ARC (w/o PAE) HIGHMEM addresses are actually smaller (0 based) - * than addresses in normal ala low memory (0x8000_0000 based). + * than addresses in normal aka low memory (0x8000_0000 based). * Even with PAE, the huge peripheral space hole would waste a lot of - * mem with single mem_map[]. This warrants a mem_map per region design. - * Thus HIGHMEM on ARC is imlemented with DISCONTIGMEM. - * - * DISCONTIGMEM in turns requires multiple nodes. node 0 above is - * populated with normal memory zone while node 1 only has highmem + * mem with single contiguous mem_map[]. + * Thus when HIGHMEM on ARC is enabled the memory map corresponding + * to the hole is freed and ARC specific version of pfn_valid() + * handles the hole in the memory map. */ - node_set_online(1); min_high_pfn = PFN_DOWN(high_mem_start); max_high_pfn = PFN_DOWN(high_mem_start + high_mem_sz); - zones_size[ZONE_NORMAL] = 0; - zones_holes[ZONE_NORMAL] = 0; - - zones_size[ZONE_HIGHMEM] = max_high_pfn - min_high_pfn; - zones_holes[ZONE_HIGHMEM] = 0; - - free_area_init_node(1, /* node-id */ - zones_size, /* num pages per zone */ - min_high_pfn, /* first pfn of node */ - zones_holes); /* holes */ + /* + * max_high_pfn should be ok here for both HIGHMEM and HIGHMEM+PAE. + * For HIGHMEM without PAE max_high_pfn should be less than + * min_low_pfn to guarantee that these two regions don't overlap. + * For PAE case highmem is greater than lowmem, so it is natural + * to use max_high_pfn. + * + * In both cases, holes should be handled by pfn_valid(). + */ + max_zone_pfn[ZONE_HIGHMEM] = max_high_pfn; high_memory = (void *)(min_high_pfn << PAGE_SHIFT); + + arch_pfn_offset = min(min_low_pfn, min_high_pfn); kmap_init(); + +#else /* CONFIG_HIGHMEM */ + /* pfn_valid() uses this when FLATMEM=y and HIGHMEM=n */ + max_mapnr = max_low_pfn - min_low_pfn; + +#endif /* CONFIG_HIGHMEM */ + + free_area_init(max_zone_pfn); +} + +static void __init highmem_init(void) +{ +#ifdef CONFIG_HIGHMEM + unsigned long tmp; + + memblock_phys_free(high_mem_start, high_mem_sz); + for (tmp = min_high_pfn; tmp < max_high_pfn; tmp++) + free_highmem_page(pfn_to_page(tmp)); #endif } @@ -192,14 +183,20 @@ void __init setup_arch_memory(void) */ void __init mem_init(void) { -#ifdef CONFIG_HIGHMEM - unsigned long tmp; + memblock_free_all(); + highmem_init(); - reset_all_zones_managed_pages(); - for (tmp = min_high_pfn; tmp < max_high_pfn; tmp++) - free_highmem_page(pfn_to_page(tmp)); -#endif + BUILD_BUG_ON((PTRS_PER_PGD * sizeof(pgd_t)) > PAGE_SIZE); + BUILD_BUG_ON((PTRS_PER_PUD * sizeof(pud_t)) > PAGE_SIZE); + BUILD_BUG_ON((PTRS_PER_PMD * sizeof(pmd_t)) > PAGE_SIZE); + BUILD_BUG_ON((PTRS_PER_PTE * sizeof(pte_t)) > PAGE_SIZE); +} - memblock_free_all(); - mem_init_print_info(NULL); +#ifdef CONFIG_HIGHMEM +int pfn_valid(unsigned long pfn) +{ + return (pfn >= min_high_pfn && pfn <= max_high_pfn) || + (pfn >= min_low_pfn && pfn <= max_low_pfn); } +EXPORT_SYMBOL(pfn_valid); +#endif diff --git a/arch/arc/mm/ioremap.c b/arch/arc/mm/ioremap.c index fac4adc90204..b07004d53267 100644 --- a/arch/arc/mm/ioremap.c +++ b/arch/arc/mm/ioremap.c @@ -8,7 +8,6 @@ #include <linux/module.h> #include <linux/io.h> #include <linux/mm.h> -#include <linux/slab.h> #include <linux/cache.h> static inline bool arc_uncached_addr_space(phys_addr_t paddr) @@ -25,13 +24,6 @@ static inline bool arc_uncached_addr_space(phys_addr_t paddr) void __iomem *ioremap(phys_addr_t paddr, unsigned long size) { - phys_addr_t end; - - /* Don't allow wraparound or zero size */ - end = paddr + size - 1; - if (!size || (end < paddr)) - return NULL; - /* * If the region is h/w uncached, MMU mapping can be elided as optim * The cast to u32 is fine as this region can only be inside 4GB @@ -39,7 +31,8 @@ void __iomem *ioremap(phys_addr_t paddr, unsigned long size) if (arc_uncached_addr_space(paddr)) return (void __iomem *)(u32)paddr; - return ioremap_prot(paddr, size, PAGE_KERNEL_NO_CACHE); + return ioremap_prot(paddr, size, + pgprot_val(pgprot_noncached(PAGE_KERNEL))); } EXPORT_SYMBOL(ioremap); @@ -50,54 +43,22 @@ EXPORT_SYMBOL(ioremap); * ARC hardware uncached region, this one still goes thru the MMU as caller * might need finer access control (R/W/X) */ -void __iomem *ioremap_prot(phys_addr_t paddr, unsigned long size, +void __iomem *ioremap_prot(phys_addr_t paddr, size_t size, unsigned long flags) { - unsigned long vaddr; - struct vm_struct *area; - phys_addr_t off, end; pgprot_t prot = __pgprot(flags); - /* Don't allow wraparound, zero size */ - end = paddr + size - 1; - if ((!size) || (end < paddr)) - return NULL; - - /* An early platform driver might end up here */ - if (!slab_is_available()) - return NULL; - /* force uncached */ - prot = pgprot_noncached(prot); - - /* Mappings have to be page-aligned */ - off = paddr & ~PAGE_MASK; - paddr &= PAGE_MASK; - size = PAGE_ALIGN(end + 1) - paddr; - - /* - * Ok, go for it.. - */ - area = get_vm_area(size, VM_IOREMAP); - if (!area) - return NULL; - area->phys_addr = paddr; - vaddr = (unsigned long)area->addr; - if (ioremap_page_range(vaddr, vaddr + size, paddr, prot)) { - vunmap((void __force *)vaddr); - return NULL; - } - return (void __iomem *)(off + (char __iomem *)vaddr); + return generic_ioremap_prot(paddr, size, pgprot_noncached(prot)); } EXPORT_SYMBOL(ioremap_prot); - -void iounmap(const void __iomem *addr) +void iounmap(volatile void __iomem *addr) { /* weird double cast to handle phys_addr_t > 32 bits */ if (arc_uncached_addr_space((phys_addr_t)(u32)addr)) return; - vfree((void *)(PAGE_MASK & (unsigned long __force)addr)); + generic_iounmap(addr); } EXPORT_SYMBOL(iounmap); diff --git a/arch/arc/mm/mmap.c b/arch/arc/mm/mmap.c index 722d26b94307..3c1c7ae73292 100644 --- a/arch/arc/mm/mmap.c +++ b/arch/arc/mm/mmap.c @@ -14,10 +14,6 @@ #include <asm/cacheflush.h> -#define COLOUR_ALIGN(addr, pgoff) \ - ((((addr) + SHMLBA - 1) & ~(SHMLBA - 1)) + \ - (((pgoff) << PAGE_SHIFT) & (SHMLBA - 1))) - /* * Ensure that shared mappings are correctly aligned to * avoid aliasing issues with VIPT caches. @@ -31,21 +27,13 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr, { struct mm_struct *mm = current->mm; struct vm_area_struct *vma; - int do_align = 0; - int aliasing = cache_is_vipt_aliasing(); struct vm_unmapped_area_info info; /* - * We only need to do colour alignment if D cache aliases. - */ - if (aliasing) - do_align = filp || (flags & MAP_SHARED); - - /* * We enforce the MAP_FIXED case. */ if (flags & MAP_FIXED) { - if (aliasing && flags & MAP_SHARED && + if (flags & MAP_SHARED && (addr - (pgoff << PAGE_SHIFT)) & (SHMLBA - 1)) return -EINVAL; return addr; @@ -55,10 +43,7 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr, return -ENOMEM; if (addr) { - if (do_align) - addr = COLOUR_ALIGN(addr, pgoff); - else - addr = PAGE_ALIGN(addr); + addr = PAGE_ALIGN(addr); vma = find_vma(mm, addr); if (TASK_SIZE - len >= addr && @@ -70,7 +55,27 @@ arch_get_unmapped_area(struct file *filp, unsigned long addr, info.length = len; info.low_limit = mm->mmap_base; info.high_limit = TASK_SIZE; - info.align_mask = do_align ? (PAGE_MASK & (SHMLBA - 1)) : 0; + info.align_mask = 0; info.align_offset = pgoff << PAGE_SHIFT; return vm_unmapped_area(&info); } + +static const pgprot_t protection_map[16] = { + [VM_NONE] = PAGE_U_NONE, + [VM_READ] = PAGE_U_R, + [VM_WRITE] = PAGE_U_R, + [VM_WRITE | VM_READ] = PAGE_U_R, + [VM_EXEC] = PAGE_U_X_R, + [VM_EXEC | VM_READ] = PAGE_U_X_R, + [VM_EXEC | VM_WRITE] = PAGE_U_X_R, + [VM_EXEC | VM_WRITE | VM_READ] = PAGE_U_X_R, + [VM_SHARED] = PAGE_U_NONE, + [VM_SHARED | VM_READ] = PAGE_U_R, + [VM_SHARED | VM_WRITE] = PAGE_U_W_R, + [VM_SHARED | VM_WRITE | VM_READ] = PAGE_U_W_R, + [VM_SHARED | VM_EXEC] = PAGE_U_X_R, + [VM_SHARED | VM_EXEC | VM_READ] = PAGE_U_X_R, + [VM_SHARED | VM_EXEC | VM_WRITE] = PAGE_U_X_W_R, + [VM_SHARED | VM_EXEC | VM_WRITE | VM_READ] = PAGE_U_X_W_R +}; +DECLARE_VM_GET_PAGE_PROT diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c index c340acd989a0..ad702b49aeb3 100644 --- a/arch/arc/mm/tlb.c +++ b/arch/arc/mm/tlb.c @@ -1,51 +1,9 @@ // SPDX-License-Identifier: GPL-2.0-only /* - * TLB Management (flush/create/diagnostics) for ARC700 + * TLB Management (flush/create/diagnostics) for MMUv3 and MMUv4 * * Copyright (C) 2004, 2007-2010, 2011-2012 Synopsys, Inc. (www.synopsys.com) * - * vineetg: Aug 2011 - * -Reintroduce duplicate PD fixup - some customer chips still have the issue - * - * vineetg: May 2011 - * -No need to flush_cache_page( ) for each call to update_mmu_cache() - * some of the LMBench tests improved amazingly - * = page-fault thrice as fast (75 usec to 28 usec) - * = mmap twice as fast (9.6 msec to 4.6 msec), - * = fork (5.3 msec to 3.7 msec) - * - * vineetg: April 2011 : - * -MMU v3: PD{0,1} bits layout changed: They don't overlap anymore, - * helps avoid a shift when preparing PD0 from PTE - * - * vineetg: April 2011 : Preparing for MMU V3 - * -MMU v2/v3 BCRs decoded differently - * -Remove TLB_SIZE hardcoding as it's variable now: 256 or 512 - * -tlb_entry_erase( ) can be void - * -local_flush_tlb_range( ): - * = need not "ceil" @end - * = walks MMU only if range spans < 32 entries, as opposed to 256 - * - * Vineetg: Sept 10th 2008 - * -Changes related to MMU v2 (Rel 4.8) - * - * Vineetg: Aug 29th 2008 - * -In TLB Flush operations (Metal Fix MMU) there is a explict command to - * flush Micro-TLBS. If TLB Index Reg is invalid prior to TLBIVUTLB cmd, - * it fails. Thus need to load it with ANY valid value before invoking - * TLBIVUTLB cmd - * - * Vineetg: Aug 21th 2008: - * -Reduced the duration of IRQ lockouts in TLB Flush routines - * -Multiple copies of TLB erase code seperated into a "single" function - * -In TLB Flush routines, interrupt disabling moved UP to retrieve ASID - * in interrupt-safe region. - * - * Vineetg: April 23rd Bug #93131 - * Problem: tlb_flush_kernel_range() doesn't do anything if the range to - * flush is more than the size of TLB itself. - * - * Rahul Trivedi : Codito Technologies 2004 */ #include <linux/module.h> @@ -57,51 +15,12 @@ #include <asm/mmu_context.h> #include <asm/mmu.h> -/* Need for ARC MMU v2 - * - * ARC700 MMU-v1 had a Joint-TLB for Code and Data and is 2 way set-assoc. - * For a memcpy operation with 3 players (src/dst/code) such that all 3 pages - * map into same set, there would be contention for the 2 ways causing severe - * Thrashing. - * - * Although J-TLB is 2 way set assoc, ARC700 caches J-TLB into uTLBS which has - * much higher associativity. u-D-TLB is 8 ways, u-I-TLB is 4 ways. - * Given this, the thrasing problem should never happen because once the 3 - * J-TLB entries are created (even though 3rd will knock out one of the prev - * two), the u-D-TLB and u-I-TLB will have what is required to accomplish memcpy - * - * Yet we still see the Thrashing because a J-TLB Write cause flush of u-TLBs. - * This is a simple design for keeping them in sync. So what do we do? - * The solution which James came up was pretty neat. It utilised the assoc - * of uTLBs by not invalidating always but only when absolutely necessary. - * - * - Existing TLB commands work as before - * - New command (TLBWriteNI) for TLB write without clearing uTLBs - * - New command (TLBIVUTLB) to invalidate uTLBs. - * - * The uTLBs need only be invalidated when pages are being removed from the - * OS page table. If a 'victim' TLB entry is being overwritten in the main TLB - * as a result of a miss, the removed entry is still allowed to exist in the - * uTLBs as it is still valid and present in the OS page table. This allows the - * full associativity of the uTLBs to hide the limited associativity of the main - * TLB. - * - * During a miss handler, the new "TLBWriteNI" command is used to load - * entries without clearing the uTLBs. - * - * When the OS page table is updated, TLB entries that may be associated with a - * removed page are removed (flushed) from the TLB using TLBWrite. In this - * circumstance, the uTLBs must also be cleared. This is done by using the - * existing TLBWrite command. An explicit IVUTLB is also required for those - * corner cases when TLBWrite was not executed at all because the corresp - * J-TLB entry got evicted/replaced. - */ - - /* A copy of the ASID from the PID reg is kept in asid_cache */ DEFINE_PER_CPU(unsigned int, asid_cache) = MM_CTXT_FIRST_CYCLE; -static int __read_mostly pae_exists; +static struct cpuinfo_arc_mmu { + unsigned int ver, pg_sz_k, s_pg_sz_m, pae, sets, ways; +} mmuinfo; /* * Utility Routine to erase a J-TLB entry @@ -120,32 +39,10 @@ static inline void __tlb_entry_erase(void) static void utlb_invalidate(void) { -#if (CONFIG_ARC_MMU_VER >= 2) - -#if (CONFIG_ARC_MMU_VER == 2) - /* MMU v2 introduced the uTLB Flush command. - * There was however an obscure hardware bug, where uTLB flush would - * fail when a prior probe for J-TLB (both totally unrelated) would - * return lkup err - because the entry didn't exist in MMU. - * The Workround was to set Index reg with some valid value, prior to - * flush. This was fixed in MMU v3 - */ - unsigned int idx; - - /* make sure INDEX Reg is valid */ - idx = read_aux_reg(ARC_REG_TLBINDEX); - - /* If not write some dummy val */ - if (unlikely(idx & TLB_LKUP_ERR)) - write_aux_reg(ARC_REG_TLBINDEX, 0xa); -#endif - write_aux_reg(ARC_REG_TLBCOMMAND, TLBIVUTLB); -#endif - } -#if (CONFIG_ARC_MMU_VER < 4) +#ifdef CONFIG_ARC_MMU_V3 static inline unsigned int tlb_entry_lkup(unsigned long vaddr_n_asid) { @@ -176,7 +73,7 @@ static void tlb_entry_erase(unsigned int vaddr_n_asid) } } -static void tlb_entry_insert(unsigned int pd0, pte_t pd1) +static void tlb_entry_insert(unsigned int pd0, phys_addr_t pd1) { unsigned int idx; @@ -206,7 +103,7 @@ static void tlb_entry_insert(unsigned int pd0, pte_t pd1) write_aux_reg(ARC_REG_TLBCOMMAND, TLBWrite); } -#else /* CONFIG_ARC_MMU_VER >= 4) */ +#else /* MMUv4 */ static void tlb_entry_erase(unsigned int vaddr_n_asid) { @@ -214,13 +111,16 @@ static void tlb_entry_erase(unsigned int vaddr_n_asid) write_aux_reg(ARC_REG_TLBCOMMAND, TLBDeleteEntry); } -static void tlb_entry_insert(unsigned int pd0, pte_t pd1) +static void tlb_entry_insert(unsigned int pd0, phys_addr_t pd1) { write_aux_reg(ARC_REG_TLBPD0, pd0); - write_aux_reg(ARC_REG_TLBPD1, pd1); - if (is_pae40_enabled()) + if (!is_pae40_enabled()) { + write_aux_reg(ARC_REG_TLBPD1, pd1); + } else { + write_aux_reg(ARC_REG_TLBPD1, pd1 & 0xFFFFFFFF); write_aux_reg(ARC_REG_TLBPD1HI, (u64)pd1 >> 32); + } write_aux_reg(ARC_REG_TLBCOMMAND, TLBInsertEntry); } @@ -233,7 +133,7 @@ static void tlb_entry_insert(unsigned int pd0, pte_t pd1) noinline void local_flush_tlb_all(void) { - struct cpuinfo_arc_mmu *mmu = &cpuinfo_arc700[smp_processor_id()].mmu; + struct cpuinfo_arc_mmu *mmu = &mmuinfo; unsigned long flags; unsigned int entry; int num_tlb = mmu->sets * mmu->ways; @@ -272,7 +172,7 @@ noinline void local_flush_tlb_all(void) } /* - * Flush the entrie MM for userland. The fastest way is to move to Next ASID + * Flush the entire MM for userland. The fastest way is to move to Next ASID */ noinline void local_flush_tlb_mm(struct mm_struct *mm) { @@ -303,7 +203,7 @@ noinline void local_flush_tlb_mm(struct mm_struct *mm) * Difference between this and Kernel Range Flush is * -Here the fastest way (if range is too large) is to move to next ASID * without doing any explicit Shootdown - * -In case of kernel Flush, entry has to be shot down explictly + * -In case of kernel Flush, entry has to be shot down explicitly */ void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end) @@ -491,12 +391,12 @@ void flush_tlb_kernel_range(unsigned long start, unsigned long end) /* * Routine to create a TLB entry */ -void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep) +static void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep) { unsigned long flags; unsigned int asid_or_sasid, rwx; unsigned long pd0; - pte_t pd1; + phys_addr_t pd1; /* * create_tlb() assumes that current->mm == vma->mm, since @@ -505,7 +405,6 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep) * * Removing the assumption involves * -Using vma->mm->context{ASID,SASID}, as opposed to MMU reg. - * -Fix the TLB paranoid debug code to not trigger false negatives. * -More importantly it makes this handler inconsistent with fast-path * TLB Refill handler which always deals with "current" * @@ -528,8 +427,6 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep) local_irq_save(flags); - tlb_paranoid_check(asid_mm(vma->vm_mm, smp_processor_id()), vaddr); - vaddr &= PAGE_MASK; /* update this PTE credentials */ @@ -572,39 +469,37 @@ void create_tlb(struct vm_area_struct *vma, unsigned long vaddr, pte_t *ptep) * Note that flush (when done) involves both WBACK - so physical page is * in sync as well as INV - so any non-congruent aliases don't remain */ -void update_mmu_cache(struct vm_area_struct *vma, unsigned long vaddr_unaligned, - pte_t *ptep) +void update_mmu_cache_range(struct vm_fault *vmf, struct vm_area_struct *vma, + unsigned long vaddr_unaligned, pte_t *ptep, unsigned int nr) { unsigned long vaddr = vaddr_unaligned & PAGE_MASK; - phys_addr_t paddr = pte_val(*ptep) & PAGE_MASK; + phys_addr_t paddr = pte_val(*ptep) & PAGE_MASK_PHYS; struct page *page = pfn_to_page(pte_pfn(*ptep)); create_tlb(vma, vaddr, ptep); - if (page == ZERO_PAGE(0)) { + if (page == ZERO_PAGE(0)) return; - } /* - * Exec page : Independent of aliasing/page-color considerations, - * since icache doesn't snoop dcache on ARC, any dirty - * K-mapping of a code page needs to be wback+inv so that - * icache fetch by userspace sees code correctly. - * !EXEC page: If K-mapping is NOT congruent to U-mapping, flush it - * so userspace sees the right data. - * (Avoids the flush for Non-exec + congruent mapping case) + * For executable pages, since icache doesn't snoop dcache, any + * dirty K-mapping of a code page needs to be wback+inv so that + * icache fetch by userspace sees code correctly. */ - if ((vma->vm_flags & VM_EXEC) || - addr_not_cache_congruent(paddr, vaddr)) { - - int dirty = !test_and_set_bit(PG_dc_clean, &page->flags); + if (vma->vm_flags & VM_EXEC) { + struct folio *folio = page_folio(page); + int dirty = !test_and_set_bit(PG_dc_clean, &folio->flags); if (dirty) { + unsigned long offset = offset_in_folio(folio, paddr); + nr = folio_nr_pages(folio); + paddr -= offset; + vaddr -= offset; /* wback + inv dcache lines (K-mapping) */ - __flush_dcache_page(paddr, paddr); + __flush_dcache_pages(paddr, paddr, nr); /* invalidate any existing icache lines (U-mapping) */ if (vma->vm_flags & VM_EXEC) - __inv_icache_page(paddr, vaddr); + __inv_icache_pages(paddr, vaddr, nr); } } } @@ -620,7 +515,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long vaddr_unaligned, * Super Page size is configurable in hardware (4K to 16M), but fixed once * RTL builds. * - * The exact THP size a Linx configuration will support is a function of: + * The exact THP size a Linux configuration will support is a function of: * - MMU page size (typical 8K, RTL fixed) * - software page walker address split between PGD:PTE:PFN (typical * 11:8:13, but can be changed with 1 line) @@ -636,44 +531,7 @@ void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long addr, pmd_t *pmd) { pte_t pte = __pte(pmd_val(*pmd)); - update_mmu_cache(vma, addr, &pte); -} - -void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp, - pgtable_t pgtable) -{ - struct list_head *lh = (struct list_head *) pgtable; - - assert_spin_locked(&mm->page_table_lock); - - /* FIFO */ - if (!pmd_huge_pte(mm, pmdp)) - INIT_LIST_HEAD(lh); - else - list_add(lh, (struct list_head *) pmd_huge_pte(mm, pmdp)); - pmd_huge_pte(mm, pmdp) = pgtable; -} - -pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp) -{ - struct list_head *lh; - pgtable_t pgtable; - - assert_spin_locked(&mm->page_table_lock); - - pgtable = pmd_huge_pte(mm, pmdp); - lh = (struct list_head *) pgtable; - if (list_empty(lh)) - pmd_huge_pte(mm, pmdp) = NULL; - else { - pmd_huge_pte(mm, pmdp) = (pgtable_t) lh->next; - list_del(lh); - } - - pte_val(pgtable[0]) = 0; - pte_val(pgtable[1]) = 0; - - return pgtable; + update_mmu_cache_range(NULL, vma, addr, &pte, HPAGE_PMD_NR); } void local_flush_pmd_tlb_range(struct vm_area_struct *vma, unsigned long start, @@ -698,139 +556,92 @@ void local_flush_pmd_tlb_range(struct vm_area_struct *vma, unsigned long start, #endif -/* Read the Cache Build Confuration Registers, Decode them and save into +/* Read the Cache Build Configuration Registers, Decode them and save into * the cpuinfo structure for later use. * No Validation is done here, simply read/convert the BCRs */ -void read_decode_mmu_bcr(void) +int arc_mmu_mumbojumbo(int c, char *buf, int len) { - struct cpuinfo_arc_mmu *mmu = &cpuinfo_arc700[smp_processor_id()].mmu; - unsigned int tmp; - struct bcr_mmu_1_2 { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int ver:8, ways:4, sets:4, u_itlb:8, u_dtlb:8; -#else - unsigned int u_dtlb:8, u_itlb:8, sets:4, ways:4, ver:8; -#endif - } *mmu2; - - struct bcr_mmu_3 { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int ver:8, ways:4, sets:4, res:3, sasid:1, pg_sz:4, - u_itlb:4, u_dtlb:4; -#else - unsigned int u_dtlb:4, u_itlb:4, pg_sz:4, sasid:1, res:3, sets:4, - ways:4, ver:8; -#endif - } *mmu3; - - struct bcr_mmu_4 { -#ifdef CONFIG_CPU_BIG_ENDIAN - unsigned int ver:8, sasid:1, sz1:4, sz0:4, res:2, pae:1, - n_ways:2, n_entry:2, n_super:2, u_itlb:3, u_dtlb:3; -#else - /* DTLB ITLB JES JE JA */ - unsigned int u_dtlb:3, u_itlb:3, n_super:2, n_entry:2, n_ways:2, - pae:1, res:2, sz0:4, sz1:4, sasid:1, ver:8; -#endif - } *mmu4; - - tmp = read_aux_reg(ARC_REG_MMU_BCR); - mmu->ver = (tmp >> 24); - - if (is_isa_arcompact()) { - if (mmu->ver <= 2) { - mmu2 = (struct bcr_mmu_1_2 *)&tmp; - mmu->pg_sz_k = TO_KB(0x2000); - mmu->sets = 1 << mmu2->sets; - mmu->ways = 1 << mmu2->ways; - mmu->u_dtlb = mmu2->u_dtlb; - mmu->u_itlb = mmu2->u_itlb; - } else { - mmu3 = (struct bcr_mmu_3 *)&tmp; - mmu->pg_sz_k = 1 << (mmu3->pg_sz - 1); - mmu->sets = 1 << mmu3->sets; - mmu->ways = 1 << mmu3->ways; - mmu->u_dtlb = mmu3->u_dtlb; - mmu->u_itlb = mmu3->u_itlb; - mmu->sasid = mmu3->sasid; - } + struct cpuinfo_arc_mmu *mmu = &mmuinfo; + unsigned int bcr, u_dtlb, u_itlb, sasid; + struct bcr_mmu_3 *mmu3; + struct bcr_mmu_4 *mmu4; + char super_pg[64] = ""; + int n = 0; + + bcr = read_aux_reg(ARC_REG_MMU_BCR); + mmu->ver = (bcr >> 24); + + if (is_isa_arcompact() && mmu->ver == 3) { + mmu3 = (struct bcr_mmu_3 *)&bcr; + mmu->pg_sz_k = 1 << (mmu3->pg_sz - 1); + mmu->sets = 1 << mmu3->sets; + mmu->ways = 1 << mmu3->ways; + u_dtlb = mmu3->u_dtlb; + u_itlb = mmu3->u_itlb; + sasid = mmu3->sasid; } else { - mmu4 = (struct bcr_mmu_4 *)&tmp; + mmu4 = (struct bcr_mmu_4 *)&bcr; mmu->pg_sz_k = 1 << (mmu4->sz0 - 1); mmu->s_pg_sz_m = 1 << (mmu4->sz1 - 11); mmu->sets = 64 << mmu4->n_entry; mmu->ways = mmu4->n_ways * 2; - mmu->u_dtlb = mmu4->u_dtlb * 4; - mmu->u_itlb = mmu4->u_itlb * 4; - mmu->sasid = mmu4->sasid; - pae_exists = mmu->pae = mmu4->pae; + u_dtlb = mmu4->u_dtlb * 4; + u_itlb = mmu4->u_itlb * 4; + sasid = mmu4->sasid; + mmu->pae = mmu4->pae; } -} - -char *arc_mmu_mumbojumbo(int cpu_id, char *buf, int len) -{ - int n = 0; - struct cpuinfo_arc_mmu *p_mmu = &cpuinfo_arc700[cpu_id].mmu; - char super_pg[64] = ""; - if (p_mmu->s_pg_sz_m) - scnprintf(super_pg, 64, "%dM Super Page %s", - p_mmu->s_pg_sz_m, - IS_USED_CFG(CONFIG_TRANSPARENT_HUGEPAGE)); + if (mmu->s_pg_sz_m) + scnprintf(super_pg, 64, "/%dM%s", + mmu->s_pg_sz_m, + IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) ? " (THP enabled)":""); n += scnprintf(buf + n, len - n, - "MMU [v%x]\t: %dk PAGE, %sJTLB %d (%dx%d), uDTLB %d, uITLB %d%s%s\n", - p_mmu->ver, p_mmu->pg_sz_k, super_pg, - p_mmu->sets * p_mmu->ways, p_mmu->sets, p_mmu->ways, - p_mmu->u_dtlb, p_mmu->u_itlb, - IS_AVAIL2(p_mmu->pae, ", PAE40 ", CONFIG_ARC_HAS_PAE40)); + "MMU [v%x]\t: %dk%s, swalk %d lvl, JTLB %dx%d, uDTLB %d, uITLB %d%s%s%s\n", + mmu->ver, mmu->pg_sz_k, super_pg, CONFIG_PGTABLE_LEVELS, + mmu->sets, mmu->ways, + u_dtlb, u_itlb, + IS_AVAIL1(sasid, ", SASID"), + IS_AVAIL2(mmu->pae, ", PAE40 ", CONFIG_ARC_HAS_PAE40)); - return buf; + return n; } int pae40_exist_but_not_enab(void) { - return pae_exists && !is_pae40_enabled(); + return mmuinfo.pae && !is_pae40_enabled(); } void arc_mmu_init(void) { - struct cpuinfo_arc_mmu *mmu = &cpuinfo_arc700[smp_processor_id()].mmu; - char str[256]; + struct cpuinfo_arc_mmu *mmu = &mmuinfo; int compat = 0; - pr_info("%s", arc_mmu_mumbojumbo(0, str, sizeof(str))); - /* - * Can't be done in processor.h due to header include depenedencies + * Can't be done in processor.h due to header include dependencies */ BUILD_BUG_ON(!IS_ALIGNED((CONFIG_ARC_KVADDR_SIZE << 20), PMD_SIZE)); /* * stack top size sanity check, - * Can't be done in processor.h due to header include depenedencies + * Can't be done in processor.h due to header include dependencies */ BUILD_BUG_ON(!IS_ALIGNED(STACK_TOP, PMD_SIZE)); /* * Ensure that MMU features assumed by kernel exist in hardware. - * For older ARC700 cpus, it has to be exact match, since the MMU - * revisions were not backwards compatible (MMUv3 TLB layout changed - * so even if kernel for v2 didn't use any new cmds of v3, it would - * still not work. - * For HS cpus, MMUv4 was baseline and v5 is backwards compatible - * (will run older software). + * - For older ARC700 cpus, only v3 supported + * - For HS cpus, v4 was baseline and v5 is backwards compatible + * (will run older software). */ - if (is_isa_arcompact() && mmu->ver == CONFIG_ARC_MMU_VER) + if (is_isa_arcompact() && mmu->ver == 3) compat = 1; - else if (is_isa_arcv2() && mmu->ver >= CONFIG_ARC_MMU_VER) + else if (is_isa_arcv2() && mmu->ver >= 4) compat = 1; - if (!compat) { - panic("MMU ver %d doesn't match kernel built for %d...\n", - mmu->ver, CONFIG_ARC_MMU_VER); - } + if (!compat) + panic("MMU ver %d doesn't match kernel built for\n", mmu->ver); if (mmu->pg_sz_k != TO_KB(PAGE_SIZE)) panic("MMU pg size != PAGE_SIZE (%luk)\n", TO_KB(PAGE_SIZE)); @@ -843,14 +654,11 @@ void arc_mmu_init(void) if (IS_ENABLED(CONFIG_ARC_HAS_PAE40) && !mmu->pae) panic("Hardware doesn't support PAE40\n"); - /* Enable the MMU */ - write_aux_reg(ARC_REG_PID, MMU_ENABLE); + /* Enable the MMU with ASID 0 */ + mmu_setup_asid(NULL, 0); - /* In smp we use this reg for interrupt 1 scratch */ -#ifdef ARC_USE_SCRATCH_REG - /* swapper_pg_dir is the pgd for the kernel, used by vmalloc */ - write_aux_reg(ARC_REG_SCRATCH_DATA0, swapper_pg_dir); -#endif + /* cache the pgd pointer in MMU SCRATCH reg (ARCv2 only) */ + mmu_setup_pgd(NULL, swapper_pg_dir); if (pae40_exist_but_not_enab()) write_aux_reg(ARC_REG_TLBPD1HI, 0); @@ -881,12 +689,12 @@ void arc_mmu_init(void) * the duplicate one. * -Knob to be verbose abt it.(TODO: hook them up to debugfs) */ -volatile int dup_pd_silent; /* Be slient abt it or complain (default) */ +volatile int dup_pd_silent; /* Be silent abt it or complain (default) */ void do_tlb_overlap_fault(unsigned long cause, unsigned long address, struct pt_regs *regs) { - struct cpuinfo_arc_mmu *mmu = &cpuinfo_arc700[smp_processor_id()].mmu; + struct cpuinfo_arc_mmu *mmu = &mmuinfo; unsigned long flags; int set, n_ways = mmu->ways; @@ -945,40 +753,3 @@ void do_tlb_overlap_fault(unsigned long cause, unsigned long address, local_irq_restore(flags); } - -/*********************************************************************** - * Diagnostic Routines - * -Called from Low Level TLB Hanlders if things don;t look good - **********************************************************************/ - -#ifdef CONFIG_ARC_DBG_TLB_PARANOIA - -/* - * Low Level ASM TLB handler calls this if it finds that HW and SW ASIDS - * don't match - */ -void print_asid_mismatch(int mm_asid, int mmu_asid, int is_fast_path) -{ - pr_emerg("ASID Mismatch in %s Path Handler: sw-pid=0x%x hw-pid=0x%x\n", - is_fast_path ? "Fast" : "Slow", mm_asid, mmu_asid); - - __asm__ __volatile__("flag 1"); -} - -void tlb_paranoid_check(unsigned int mm_asid, unsigned long addr) -{ - unsigned int mmu_asid; - - mmu_asid = read_aux_reg(ARC_REG_PID) & 0xff; - - /* - * At the time of a TLB miss/installation - * - HW version needs to match SW version - * - SW needs to have a valid ASID - */ - if (addr < 0x70000000 && - ((mm_asid == MM_CTXT_NO_ASID) || - (mmu_asid != (mm_asid & MM_CTXT_ASID_MASK)))) - print_asid_mismatch(mm_asid, mmu_asid, 0); -} -#endif diff --git a/arch/arc/mm/tlbex.S b/arch/arc/mm/tlbex.S index 2efaf6ca0c06..e054780a8fe0 100644 --- a/arch/arc/mm/tlbex.S +++ b/arch/arc/mm/tlbex.S @@ -33,13 +33,12 @@ */ #include <linux/linkage.h> +#include <linux/pgtable.h> #include <asm/entry.h> #include <asm/mmu.h> -#include <asm/pgtable.h> #include <asm/arcregs.h> #include <asm/cache.h> #include <asm/processor.h> -#include <asm/tlb-mmu1.h> #ifdef CONFIG_ISA_ARCOMPACT ;----------------------------------------------------------------- @@ -94,11 +93,6 @@ ex_saved_reg1: st_s r1, [r0, 4] st_s r2, [r0, 8] st_s r3, [r0, 12] - - ; VERIFY if the ASID in MMU-PID Reg is same as - ; one in Linux data structures - - tlb_paranoid_check_asm .endm .macro TLBMISS_RESTORE_REGS @@ -148,53 +142,16 @@ ex_saved_reg1: #endif ;============================================================================ -; Troubleshooting Stuff +;TLB Miss handling Code ;============================================================================ -; Linux keeps ASID (Address Space ID) in task->active_mm->context.asid -; When Creating TLB Entries, instead of doing 3 dependent loads from memory, -; we use the MMU PID Reg to get current ASID. -; In bizzare scenrios SW and HW ASID can get out-of-sync which is trouble. -; So we try to detect this in TLB Mis shandler - -.macro tlb_paranoid_check_asm - -#ifdef CONFIG_ARC_DBG_TLB_PARANOIA - - GET_CURR_TASK_ON_CPU r3 - ld r0, [r3, TASK_ACT_MM] - ld r0, [r0, MM_CTXT+MM_CTXT_ASID] - breq r0, 0, 55f ; Error if no ASID allocated - - lr r1, [ARC_REG_PID] - and r1, r1, 0xFF - - and r2, r0, 0xFF ; MMU PID bits only for comparison - breq r1, r2, 5f - -55: - ; Error if H/w and S/w ASID don't match, but NOT if in kernel mode - lr r2, [erstatus] - bbit0 r2, STATUS_U_BIT, 5f - - ; We sure are in troubled waters, Flag the error, but to do so - ; need to switch to kernel mode stack to call error routine - GET_TSK_STACK_BASE r3, sp - - ; Call printk to shoutout aloud - mov r2, 1 - j print_asid_mismatch - -5: ; ASIDs match so proceed normally - nop - +#ifndef PMD_SHIFT +#define PMD_SHIFT PUD_SHIFT #endif -.endm - -;============================================================================ -;TLB Miss handling Code -;============================================================================ +#ifndef PUD_SHIFT +#define PUD_SHIFT PGDIR_SHIFT +#endif ;----------------------------------------------------------------------------- ; This macro does the page-table lookup for the faulting address. @@ -203,7 +160,7 @@ ex_saved_reg1: lr r2, [efa] -#ifdef ARC_USE_SCRATCH_REG +#ifdef CONFIG_ISA_ARCV2 lr r1, [ARC_REG_SCRATCH_DATA0] ; current pgd #else GET_CURR_TASK_ON_CPU r1 @@ -216,6 +173,24 @@ ex_saved_reg1: tst r3, r3 bz do_slow_path_pf ; if no Page Table, do page fault +#if CONFIG_PGTABLE_LEVELS > 3 + lsr r0, r2, PUD_SHIFT ; Bits for indexing into PUD + and r0, r0, (PTRS_PER_PUD - 1) + ld.as r1, [r3, r0] ; PMD entry + tst r1, r1 + bz do_slow_path_pf + mov r3, r1 +#endif + +#if CONFIG_PGTABLE_LEVELS > 2 + lsr r0, r2, PMD_SHIFT ; Bits for indexing into PMD + and r0, r0, (PTRS_PER_PMD - 1) + ld.as r1, [r3, r0] ; PMD entry + tst r1, r1 + bz do_slow_path_pf + mov r3, r1 +#endif + #ifdef CONFIG_TRANSPARENT_HUGEPAGE and.f 0, r3, _PAGE_HW_SZ ; Is this Huge PMD (thp) add2.nz r1, r1, r0 @@ -279,14 +254,7 @@ ex_saved_reg1: ; Commit the TLB entry into MMU .macro COMMIT_ENTRY_TO_MMU -#if (CONFIG_ARC_MMU_VER < 4) - -#ifdef CONFIG_EZNPS_MTM_EXT - /* verify if entry for this vaddr+ASID already exists */ - sr TLBProbe, [ARC_REG_TLBCOMMAND] - lr r0, [ARC_REG_TLBINDEX] - bbit0 r0, 31, 88f -#endif +#ifdef CONFIG_ARC_MMU_V3 /* Get free TLB slot: Set = computed from vaddr, way = random */ sr TLBGetIndex, [ARC_REG_TLBCOMMAND] @@ -382,13 +350,6 @@ ENTRY(EV_TLBMissD) CONV_PTE_TO_TLB -#if (CONFIG_ARC_MMU_VER == 1) - ; MMU with 2 way set assoc J-TLB, needs some help in pathetic case of - ; memcpy where 3 parties contend for 2 ways, ensuing a livelock. - ; But only for old MMU or one with Metal Fix - TLB_WRITE_HEURISTICS -#endif - COMMIT_ENTRY_TO_MMU TLBMISS_RESTORE_REGS EV_TLBMissD_fast_ret: ; additional label for VDK OS-kit instrumentation |