diff options
Diffstat (limited to 'arch/x86/mm')
| -rw-r--r-- | arch/x86/mm/dump_pagetables.c | 27 | ||||
| -rw-r--r-- | arch/x86/mm/fault.c | 2 | ||||
| -rw-r--r-- | arch/x86/mm/init.c | 60 | ||||
| -rw-r--r-- | arch/x86/mm/init_64.c | 14 | ||||
| -rw-r--r-- | arch/x86/mm/kmmio.c | 25 | ||||
| -rw-r--r-- | arch/x86/mm/mmap.c | 21 | ||||
| -rw-r--r-- | arch/x86/mm/numa_emulation.c | 107 | ||||
| -rw-r--r-- | arch/x86/mm/pageattr.c | 27 | ||||
| -rw-r--r-- | arch/x86/mm/pgtable.c | 169 | ||||
| -rw-r--r-- | arch/x86/mm/pti.c | 262 | ||||
| -rw-r--r-- | arch/x86/mm/tlb.c | 224 |
11 files changed, 730 insertions, 208 deletions
diff --git a/arch/x86/mm/dump_pagetables.c b/arch/x86/mm/dump_pagetables.c index 2f3c9196b834..a12afff146d1 100644 --- a/arch/x86/mm/dump_pagetables.c +++ b/arch/x86/mm/dump_pagetables.c @@ -111,6 +111,8 @@ static struct addr_marker address_markers[] = { [END_OF_SPACE_NR] = { -1, NULL } }; +#define INIT_PGD ((pgd_t *) &init_top_pgt) + #else /* CONFIG_X86_64 */ enum address_markers_idx { @@ -121,6 +123,9 @@ enum address_markers_idx { #ifdef CONFIG_HIGHMEM PKMAP_BASE_NR, #endif +#ifdef CONFIG_MODIFY_LDT_SYSCALL + LDT_NR, +#endif CPU_ENTRY_AREA_NR, FIXADDR_START_NR, END_OF_SPACE_NR, @@ -134,11 +139,16 @@ static struct addr_marker address_markers[] = { #ifdef CONFIG_HIGHMEM [PKMAP_BASE_NR] = { 0UL, "Persistent kmap() Area" }, #endif +#ifdef CONFIG_MODIFY_LDT_SYSCALL + [LDT_NR] = { 0UL, "LDT remap" }, +#endif [CPU_ENTRY_AREA_NR] = { 0UL, "CPU entry area" }, [FIXADDR_START_NR] = { 0UL, "Fixmap area" }, [END_OF_SPACE_NR] = { -1, NULL } }; +#define INIT_PGD (swapper_pg_dir) + #endif /* !CONFIG_X86_64 */ /* Multipliers for offsets within the PTEs */ @@ -496,11 +506,7 @@ static inline bool is_hypervisor_range(int idx) static void ptdump_walk_pgd_level_core(struct seq_file *m, pgd_t *pgd, bool checkwx, bool dmesg) { -#ifdef CONFIG_X86_64 - pgd_t *start = (pgd_t *) &init_top_pgt; -#else - pgd_t *start = swapper_pg_dir; -#endif + pgd_t *start = INIT_PGD; pgprotval_t prot, eff; int i; struct pg_state st = {}; @@ -563,12 +569,13 @@ void ptdump_walk_pgd_level_debugfs(struct seq_file *m, pgd_t *pgd, bool user) } EXPORT_SYMBOL_GPL(ptdump_walk_pgd_level_debugfs); -static void ptdump_walk_user_pgd_level_checkwx(void) +void ptdump_walk_user_pgd_level_checkwx(void) { #ifdef CONFIG_PAGE_TABLE_ISOLATION - pgd_t *pgd = (pgd_t *) &init_top_pgt; + pgd_t *pgd = INIT_PGD; - if (!static_cpu_has(X86_FEATURE_PTI)) + if (!(__supported_pte_mask & _PAGE_NX) || + !static_cpu_has(X86_FEATURE_PTI)) return; pr_info("x86/mm: Checking user space page tables\n"); @@ -580,7 +587,6 @@ static void ptdump_walk_user_pgd_level_checkwx(void) void ptdump_walk_pgd_level_checkwx(void) { ptdump_walk_pgd_level_core(NULL, NULL, true, false); - ptdump_walk_user_pgd_level_checkwx(); } static int __init pt_dump_init(void) @@ -609,6 +615,9 @@ static int __init pt_dump_init(void) # endif address_markers[FIXADDR_START_NR].start_address = FIXADDR_START; address_markers[CPU_ENTRY_AREA_NR].start_address = CPU_ENTRY_AREA_BASE; +# ifdef CONFIG_MODIFY_LDT_SYSCALL + address_markers[LDT_NR].start_address = LDT_BASE_ADDR; +# endif #endif return 0; } diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c index 2aafa6ab6103..db1c042e9853 100644 --- a/arch/x86/mm/fault.c +++ b/arch/x86/mm/fault.c @@ -317,8 +317,6 @@ static noinline int vmalloc_fault(unsigned long address) if (!(address >= VMALLOC_START && address < VMALLOC_END)) return -1; - WARN_ON_ONCE(in_nmi()); - /* * Synchronize this task's top level page-table * with the 'reference' page table. diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c index cee58a972cb2..156ed8154af8 100644 --- a/arch/x86/mm/init.c +++ b/arch/x86/mm/init.c @@ -4,6 +4,8 @@ #include <linux/swap.h> #include <linux/memblock.h> #include <linux/bootmem.h> /* for max_low_pfn */ +#include <linux/swapfile.h> +#include <linux/swapops.h> #include <asm/set_memory.h> #include <asm/e820/api.h> @@ -773,13 +775,44 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end) } } +/* + * begin/end can be in the direct map or the "high kernel mapping" + * used for the kernel image only. free_init_pages() will do the + * right thing for either kind of address. + */ +void free_kernel_image_pages(void *begin, void *end) +{ + unsigned long begin_ul = (unsigned long)begin; + unsigned long end_ul = (unsigned long)end; + unsigned long len_pages = (end_ul - begin_ul) >> PAGE_SHIFT; + + + free_init_pages("unused kernel image", begin_ul, end_ul); + + /* + * PTI maps some of the kernel into userspace. For performance, + * this includes some kernel areas that do not contain secrets. + * Those areas might be adjacent to the parts of the kernel image + * being freed, which may contain secrets. Remove the "high kernel + * image mapping" for these freed areas, ensuring they are not even + * potentially vulnerable to Meltdown regardless of the specific + * optimizations PTI is currently using. + * + * The "noalias" prevents unmapping the direct map alias which is + * needed to access the freed pages. + * + * This is only valid for 64bit kernels. 32bit has only one mapping + * which can't be treated in this way for obvious reasons. + */ + if (IS_ENABLED(CONFIG_X86_64) && cpu_feature_enabled(X86_FEATURE_PTI)) + set_memory_np_noalias(begin_ul, len_pages); +} + void __ref free_initmem(void) { e820__reallocate_tables(); - free_init_pages("unused kernel", - (unsigned long)(&__init_begin), - (unsigned long)(&__init_end)); + free_kernel_image_pages(&__init_begin, &__init_end); } #ifdef CONFIG_BLK_DEV_INITRD @@ -880,3 +913,24 @@ void update_cache_mode_entry(unsigned entry, enum page_cache_mode cache) __cachemode2pte_tbl[cache] = __cm_idx2pte(entry); __pte2cachemode_tbl[entry] = cache; } + +unsigned long max_swapfile_size(void) +{ + unsigned long pages; + + pages = generic_max_swapfile_size(); + + if (boot_cpu_has_bug(X86_BUG_L1TF)) { + /* Limit the swap file size to MAX_PA/2 for L1TF workaround */ + unsigned long l1tf_limit = l1tf_pfn_limit() + 1; + /* + * We encode swap offsets also with 3 bits below those for pfn + * which makes the usable limit higher. + */ +#if CONFIG_PGTABLE_LEVELS > 2 + l1tf_limit <<= PAGE_SHIFT - SWP_OFFSET_FIRST_BIT; +#endif + pages = min_t(unsigned long, l1tf_limit, pages); + } + return pages; +} diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index a688617c727e..dd519f372169 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c @@ -1283,20 +1283,10 @@ void mark_rodata_ro(void) set_memory_ro(start, (end-start) >> PAGE_SHIFT); #endif - free_init_pages("unused kernel", - (unsigned long) __va(__pa_symbol(text_end)), - (unsigned long) __va(__pa_symbol(rodata_start))); - free_init_pages("unused kernel", - (unsigned long) __va(__pa_symbol(rodata_end)), - (unsigned long) __va(__pa_symbol(_sdata))); + free_kernel_image_pages((void *)text_end, (void *)rodata_start); + free_kernel_image_pages((void *)rodata_end, (void *)_sdata); debug_checkwx(); - - /* - * Do this after all of the manipulation of the - * kernel text page tables are complete. - */ - pti_clone_kernel_text(); } int kern_addr_valid(unsigned long addr) diff --git a/arch/x86/mm/kmmio.c b/arch/x86/mm/kmmio.c index 7c8686709636..79eb55ce69a9 100644 --- a/arch/x86/mm/kmmio.c +++ b/arch/x86/mm/kmmio.c @@ -126,24 +126,29 @@ static struct kmmio_fault_page *get_kmmio_fault_page(unsigned long addr) static void clear_pmd_presence(pmd_t *pmd, bool clear, pmdval_t *old) { + pmd_t new_pmd; pmdval_t v = pmd_val(*pmd); if (clear) { - *old = v & _PAGE_PRESENT; - v &= ~_PAGE_PRESENT; - } else /* presume this has been called with clear==true previously */ - v |= *old; - set_pmd(pmd, __pmd(v)); + *old = v; + new_pmd = pmd_mknotpresent(*pmd); + } else { + /* Presume this has been called with clear==true previously */ + new_pmd = __pmd(*old); + } + set_pmd(pmd, new_pmd); } static void clear_pte_presence(pte_t *pte, bool clear, pteval_t *old) { pteval_t v = pte_val(*pte); if (clear) { - *old = v & _PAGE_PRESENT; - v &= ~_PAGE_PRESENT; - } else /* presume this has been called with clear==true previously */ - v |= *old; - set_pte_atomic(pte, __pte(v)); + *old = v; + /* Nothing should care about address */ + pte_clear(&init_mm, 0, pte); + } else { + /* Presume this has been called with clear==true previously */ + set_pte_atomic(pte, __pte(*old)); + } } static int clear_page_presence(struct kmmio_fault_page *f, bool clear) diff --git a/arch/x86/mm/mmap.c b/arch/x86/mm/mmap.c index 48c591251600..f40ab8185d94 100644 --- a/arch/x86/mm/mmap.c +++ b/arch/x86/mm/mmap.c @@ -240,3 +240,24 @@ int valid_mmap_phys_addr_range(unsigned long pfn, size_t count) return phys_addr_valid(addr + count - 1); } + +/* + * Only allow root to set high MMIO mappings to PROT_NONE. + * This prevents an unpriv. user to set them to PROT_NONE and invert + * them, then pointing to valid memory for L1TF speculation. + * + * Note: for locked down kernels may want to disable the root override. + */ +bool pfn_modify_allowed(unsigned long pfn, pgprot_t prot) +{ + if (!boot_cpu_has_bug(X86_BUG_L1TF)) + return true; + if (!__pte_needs_invert(pgprot_val(prot))) + return true; + /* If it's real memory always allow */ + if (pfn_valid(pfn)) + return true; + if (pfn > l1tf_pfn_limit() && !capable(CAP_SYS_ADMIN)) + return false; + return true; +} diff --git a/arch/x86/mm/numa_emulation.c b/arch/x86/mm/numa_emulation.c index 34a2a3bfde9c..b54d52a2d00a 100644 --- a/arch/x86/mm/numa_emulation.c +++ b/arch/x86/mm/numa_emulation.c @@ -61,7 +61,7 @@ static int __init emu_setup_memblk(struct numa_meminfo *ei, eb->nid = nid; if (emu_nid_to_phys[nid] == NUMA_NO_NODE) - emu_nid_to_phys[nid] = nid; + emu_nid_to_phys[nid] = pb->nid; pb->start += size; if (pb->start >= pb->end) { @@ -198,40 +198,73 @@ static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size) return end; } +static u64 uniform_size(u64 max_addr, u64 base, u64 hole, int nr_nodes) +{ + unsigned long max_pfn = PHYS_PFN(max_addr); + unsigned long base_pfn = PHYS_PFN(base); + unsigned long hole_pfns = PHYS_PFN(hole); + + return PFN_PHYS((max_pfn - base_pfn - hole_pfns) / nr_nodes); +} + /* * Sets up fake nodes of `size' interleaved over physical nodes ranging from * `addr' to `max_addr'. * * Returns zero on success or negative on error. */ -static int __init split_nodes_size_interleave(struct numa_meminfo *ei, +static int __init split_nodes_size_interleave_uniform(struct numa_meminfo *ei, struct numa_meminfo *pi, - u64 addr, u64 max_addr, u64 size) + u64 addr, u64 max_addr, u64 size, + int nr_nodes, struct numa_memblk *pblk, + int nid) { nodemask_t physnode_mask = numa_nodes_parsed; + int i, ret, uniform = 0; u64 min_size; - int nid = 0; - int i, ret; - if (!size) + if ((!size && !nr_nodes) || (nr_nodes && !pblk)) return -1; + /* - * The limit on emulated nodes is MAX_NUMNODES, so the size per node is - * increased accordingly if the requested size is too small. This - * creates a uniform distribution of node sizes across the entire - * machine (but not necessarily over physical nodes). + * In the 'uniform' case split the passed in physical node by + * nr_nodes, in the non-uniform case, ignore the passed in + * physical block and try to create nodes of at least size + * @size. + * + * In the uniform case, split the nodes strictly by physical + * capacity, i.e. ignore holes. In the non-uniform case account + * for holes and treat @size as a minimum floor. */ - min_size = (max_addr - addr - mem_hole_size(addr, max_addr)) / MAX_NUMNODES; - min_size = max(min_size, FAKE_NODE_MIN_SIZE); - if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size) - min_size = (min_size + FAKE_NODE_MIN_SIZE) & - FAKE_NODE_MIN_HASH_MASK; + if (!nr_nodes) + nr_nodes = MAX_NUMNODES; + else { + nodes_clear(physnode_mask); + node_set(pblk->nid, physnode_mask); + uniform = 1; + } + + if (uniform) { + min_size = uniform_size(max_addr, addr, 0, nr_nodes); + size = min_size; + } else { + /* + * The limit on emulated nodes is MAX_NUMNODES, so the + * size per node is increased accordingly if the + * requested size is too small. This creates a uniform + * distribution of node sizes across the entire machine + * (but not necessarily over physical nodes). + */ + min_size = uniform_size(max_addr, addr, + mem_hole_size(addr, max_addr), nr_nodes); + } + min_size = ALIGN(max(min_size, FAKE_NODE_MIN_SIZE), FAKE_NODE_MIN_SIZE); if (size < min_size) { pr_err("Fake node size %LuMB too small, increasing to %LuMB\n", size >> 20, min_size >> 20); size = min_size; } - size &= FAKE_NODE_MIN_HASH_MASK; + size = ALIGN_DOWN(size, FAKE_NODE_MIN_SIZE); /* * Fill physical nodes with fake nodes of size until there is no memory @@ -248,10 +281,14 @@ static int __init split_nodes_size_interleave(struct numa_meminfo *ei, node_clear(i, physnode_mask); continue; } + start = pi->blk[phys_blk].start; limit = pi->blk[phys_blk].end; - end = find_end_of_node(start, limit, size); + if (uniform) + end = start + size; + else + end = find_end_of_node(start, limit, size); /* * If there won't be at least FAKE_NODE_MIN_SIZE of * non-reserved memory in ZONE_DMA32 for the next node, @@ -266,7 +303,8 @@ static int __init split_nodes_size_interleave(struct numa_meminfo *ei, * next node, this one must extend to the end of the * physical node. */ - if (limit - end - mem_hole_size(end, limit) < size) + if ((limit - end - mem_hole_size(end, limit) < size) + && !uniform) end = limit; ret = emu_setup_memblk(ei, pi, nid++ % MAX_NUMNODES, @@ -276,7 +314,15 @@ static int __init split_nodes_size_interleave(struct numa_meminfo *ei, return ret; } } - return 0; + return nid; +} + +static int __init split_nodes_size_interleave(struct numa_meminfo *ei, + struct numa_meminfo *pi, + u64 addr, u64 max_addr, u64 size) +{ + return split_nodes_size_interleave_uniform(ei, pi, addr, max_addr, size, + 0, NULL, NUMA_NO_NODE); } int __init setup_emu2phys_nid(int *dfl_phys_nid) @@ -346,7 +392,28 @@ void __init numa_emulation(struct numa_meminfo *numa_meminfo, int numa_dist_cnt) * the fixed node size. Otherwise, if it is just a single number N, * split the system RAM into N fake nodes. */ - if (strchr(emu_cmdline, 'M') || strchr(emu_cmdline, 'G')) { + if (strchr(emu_cmdline, 'U')) { + nodemask_t physnode_mask = numa_nodes_parsed; + unsigned long n; + int nid = 0; + + n = simple_strtoul(emu_cmdline, &emu_cmdline, 0); + ret = -1; + for_each_node_mask(i, physnode_mask) { + ret = split_nodes_size_interleave_uniform(&ei, &pi, + pi.blk[i].start, pi.blk[i].end, 0, + n, &pi.blk[i], nid); + if (ret < 0) + break; + if (ret < n) { + pr_info("%s: phys: %d only got %d of %ld nodes, failing\n", + __func__, i, ret, n); + ret = -1; + break; + } + nid = ret; + } + } else if (strchr(emu_cmdline, 'M') || strchr(emu_cmdline, 'G')) { u64 size; size = memparse(emu_cmdline, &emu_cmdline); diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c index 3bded76e8d5c..8d6c34fe49be 100644 --- a/arch/x86/mm/pageattr.c +++ b/arch/x86/mm/pageattr.c @@ -53,6 +53,7 @@ static DEFINE_SPINLOCK(cpa_lock); #define CPA_FLUSHTLB 1 #define CPA_ARRAY 2 #define CPA_PAGES_ARRAY 4 +#define CPA_NO_CHECK_ALIAS 8 /* Do not search for aliases */ #ifdef CONFIG_PROC_FS static unsigned long direct_pages_count[PG_LEVEL_NUM]; @@ -1014,8 +1015,8 @@ static long populate_pmd(struct cpa_data *cpa, pmd = pmd_offset(pud, start); - set_pmd(pmd, __pmd(cpa->pfn << PAGE_SHIFT | _PAGE_PSE | - massage_pgprot(pmd_pgprot))); + set_pmd(pmd, pmd_mkhuge(pfn_pmd(cpa->pfn, + canon_pgprot(pmd_pgprot)))); start += PMD_SIZE; cpa->pfn += PMD_SIZE >> PAGE_SHIFT; @@ -1087,8 +1088,8 @@ static int populate_pud(struct cpa_data *cpa, unsigned long start, p4d_t *p4d, * Map everything starting from the Gb boundary, possibly with 1G pages */ while (boot_cpu_has(X86_FEATURE_GBPAGES) && end - start >= PUD_SIZE) { - set_pud(pud, __pud(cpa->pfn << PAGE_SHIFT | _PAGE_PSE | - massage_pgprot(pud_pgprot))); + set_pud(pud, pud_mkhuge(pfn_pud(cpa->pfn, + canon_pgprot(pud_pgprot)))); start += PUD_SIZE; cpa->pfn += PUD_SIZE >> PAGE_SHIFT; @@ -1486,6 +1487,9 @@ static int change_page_attr_set_clr(unsigned long *addr, int numpages, /* No alias checking for _NX bit modifications */ checkalias = (pgprot_val(mask_set) | pgprot_val(mask_clr)) != _PAGE_NX; + /* Has caller explicitly disabled alias checking? */ + if (in_flag & CPA_NO_CHECK_ALIAS) + checkalias = 0; ret = __change_page_attr_set_clr(&cpa, checkalias); @@ -1772,6 +1776,15 @@ int set_memory_np(unsigned long addr, int numpages) return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_PRESENT), 0); } +int set_memory_np_noalias(unsigned long addr, int numpages) +{ + int cpa_flags = CPA_NO_CHECK_ALIAS; + + return change_page_attr_set_clr(&addr, numpages, __pgprot(0), + __pgprot(_PAGE_PRESENT), 0, + cpa_flags, NULL); +} + int set_memory_4k(unsigned long addr, int numpages) { return change_page_attr_set_clr(&addr, numpages, __pgprot(0), @@ -1784,6 +1797,12 @@ int set_memory_nonglobal(unsigned long addr, int numpages) __pgprot(_PAGE_GLOBAL), 0); } +int set_memory_global(unsigned long addr, int numpages) +{ + return change_page_attr_set(&addr, numpages, + __pgprot(_PAGE_GLOBAL), 0); +} + static int __set_memory_enc_dec(unsigned long addr, int numpages, bool enc) { struct cpa_data cpa; diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c index 47b5951e592b..3ef095c70ae3 100644 --- a/arch/x86/mm/pgtable.c +++ b/arch/x86/mm/pgtable.c @@ -182,6 +182,14 @@ static void pgd_dtor(pgd_t *pgd) */ #define PREALLOCATED_PMDS UNSHARED_PTRS_PER_PGD +/* + * We allocate separate PMDs for the kernel part of the user page-table + * when PTI is enabled. We need them to map the per-process LDT into the + * user-space page-table. + */ +#define PREALLOCATED_USER_PMDS (static_cpu_has(X86_FEATURE_PTI) ? \ + KERNEL_PGD_PTRS : 0) + void pud_populate(struct mm_struct *mm, pud_t *pudp, pmd_t *pmd) { paravirt_alloc_pmd(mm, __pa(pmd) >> PAGE_SHIFT); @@ -202,14 +210,14 @@ void pud_populate(struct mm_struct *mm, pud_t *pudp, pmd_t *pmd) /* No need to prepopulate any pagetable entries in non-PAE modes. */ #define PREALLOCATED_PMDS 0 - +#define PREALLOCATED_USER_PMDS 0 #endif /* CONFIG_X86_PAE */ -static void free_pmds(struct mm_struct *mm, pmd_t *pmds[]) +static void free_pmds(struct mm_struct *mm, pmd_t *pmds[], int count) { int i; - for(i = 0; i < PREALLOCATED_PMDS; i++) + for (i = 0; i < count; i++) if (pmds[i]) { pgtable_pmd_page_dtor(virt_to_page(pmds[i])); free_page((unsigned long)pmds[i]); @@ -217,7 +225,7 @@ static void free_pmds(struct mm_struct *mm, pmd_t *pmds[]) } } -static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[]) +static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[], int count) { int i; bool failed = false; @@ -226,7 +234,7 @@ static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[]) if (mm == &init_mm) gfp &= ~__GFP_ACCOUNT; - for(i = 0; i < PREALLOCATED_PMDS; i++) { + for (i = 0; i < count; i++) { pmd_t *pmd = (pmd_t *)__get_free_page(gfp); if (!pmd) failed = true; @@ -241,7 +249,7 @@ static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[]) } if (failed) { - free_pmds(mm, pmds); + free_pmds(mm, pmds, count); return -ENOMEM; } @@ -254,23 +262,38 @@ static int preallocate_pmds(struct mm_struct *mm, pmd_t *pmds[]) * preallocate which never got a corresponding vma will need to be * freed manually. */ +static void mop_up_one_pmd(struct mm_struct *mm, pgd_t *pgdp) +{ + pgd_t pgd = *pgdp; + + if (pgd_val(pgd) != 0) { + pmd_t *pmd = (pmd_t *)pgd_page_vaddr(pgd); + + *pgdp = native_make_pgd(0); + + paravirt_release_pmd(pgd_val(pgd) >> PAGE_SHIFT); + pmd_free(mm, pmd); + mm_dec_nr_pmds(mm); + } +} + static void pgd_mop_up_pmds(struct mm_struct *mm, pgd_t *pgdp) { int i; - for(i = 0; i < PREALLOCATED_PMDS; i++) { - pgd_t pgd = pgdp[i]; + for (i = 0; i < PREALLOCATED_PMDS; i++) + mop_up_one_pmd(mm, &pgdp[i]); - if (pgd_val(pgd) != 0) { - pmd_t *pmd = (pmd_t *)pgd_page_vaddr(pgd); +#ifdef CONFIG_PAGE_TABLE_ISOLATION - pgdp[i] = native_make_pgd(0); + if (!static_cpu_has(X86_FEATURE_PTI)) + return; - paravirt_release_pmd(pgd_val(pgd) >> PAGE_SHIFT); - pmd_free(mm, pmd); - mm_dec_nr_pmds(mm); - } - } + pgdp = kernel_to_user_pgdp(pgdp); + + for (i = 0; i < PREALLOCATED_USER_PMDS; i++) + mop_up_one_pmd(mm, &pgdp[i + KERNEL_PGD_BOUNDARY]); +#endif } static void pgd_prepopulate_pmd(struct mm_struct *mm, pgd_t *pgd, pmd_t *pmds[]) @@ -296,6 +319,38 @@ static void pgd_prepopulate_pmd(struct mm_struct *mm, pgd_t *pgd, pmd_t *pmds[]) } } +#ifdef CONFIG_PAGE_TABLE_ISOLATION +static void pgd_prepopulate_user_pmd(struct mm_struct *mm, + pgd_t *k_pgd, pmd_t *pmds[]) +{ + pgd_t *s_pgd = kernel_to_user_pgdp(swapper_pg_dir); + pgd_t *u_pgd = kernel_to_user_pgdp(k_pgd); + p4d_t *u_p4d; + pud_t *u_pud; + int i; + + u_p4d = p4d_offset(u_pgd, 0); + u_pud = pud_offset(u_p4d, 0); + + s_pgd += KERNEL_PGD_BOUNDARY; + u_pud += KERNEL_PGD_BOUNDARY; + + for (i = 0; i < PREALLOCATED_USER_PMDS; i++, u_pud++, s_pgd++) { + pmd_t *pmd = pmds[i]; + + memcpy(pmd, (pmd_t *)pgd_page_vaddr(*s_pgd), + sizeof(pmd_t) * PTRS_PER_PMD); + + pud_populate(mm, u_pud, pmd); + } + +} +#else +static void pgd_prepopulate_user_pmd(struct mm_struct *mm, + pgd_t *k_pgd, pmd_t *pmds[]) +{ +} +#endif /* * Xen paravirt assumes pgd table should be in one page. 64 bit kernel also * assumes that pgd should be in one page. @@ -329,9 +384,6 @@ static int __init pgd_cache_init(void) */ pgd_cache = kmem_cache_create("pgd_cache", PGD_SIZE, PGD_ALIGN, SLAB_PANIC, NULL); - if (!pgd_cache) - return -ENOMEM; - return 0; } core_initcall(pgd_cache_init); @@ -343,7 +395,8 @@ static inline pgd_t *_pgd_alloc(void) * We allocate one page for pgd. */ if (!SHARED_KERNEL_PMD) - return (pgd_t *)__get_free_page(PGALLOC_GFP); + return (pgd_t *)__get_free_pages(PGALLOC_GFP, + PGD_ALLOCATION_ORDER); /* * Now PAE kernel is not running as a Xen domain. We can allocate @@ -355,7 +408,7 @@ static inline pgd_t *_pgd_alloc(void) static inline void _pgd_free(pgd_t *pgd) { if (!SHARED_KERNEL_PMD) - free_page((unsigned long)pgd); + free_pages((unsigned long)pgd, PGD_ALLOCATION_ORDER); else kmem_cache_free(pgd_cache, pgd); } @@ -375,6 +428,7 @@ static inline void _pgd_free(pgd_t *pgd) pgd_t *pgd_alloc(struct mm_struct *mm) { pgd_t *pgd; + pmd_t *u_pmds[PREALLOCATED_USER_PMDS]; pmd_t *pmds[PREALLOCATED_PMDS]; pgd = _pgd_alloc(); @@ -384,12 +438,15 @@ pgd_t *pgd_alloc(struct mm_struct *mm) mm->pgd = pgd; - if (preallocate_pmds(mm, pmds) != 0) + if (preallocate_pmds(mm, pmds, PREALLOCATED_PMDS) != 0) goto out_free_pgd; - if (paravirt_pgd_alloc(mm) != 0) + if (preallocate_pmds(mm, u_pmds, PREALLOCATED_USER_PMDS) != 0) goto out_free_pmds; + if (paravirt_pgd_alloc(mm) != 0) + goto out_free_user_pmds; + /* * Make sure that pre-populating the pmds is atomic with * respect to anything walking the pgd_list, so that they @@ -399,13 +456,16 @@ pgd_t *pgd_alloc(struct mm_struct *mm) pgd_ctor(mm, pgd); pgd_prepopulate_pmd(mm, pgd, pmds); + pgd_prepopulate_user_pmd(mm, pgd, u_pmds); spin_unlock(&pgd_lock); return pgd; +out_free_user_pmds: + free_pmds(mm, u_pmds, PREALLOCATED_USER_PMDS); out_free_pmds: - free_pmds(mm, pmds); + free_pmds(mm, pmds, PREALLOCATED_PMDS); out_free_pgd: _pgd_free(pgd); out: @@ -719,28 +779,50 @@ int pmd_clear_huge(pmd_t *pmd) return 0; } +#ifdef CONFIG_X86_64 /** * pud_free_pmd_page - Clear pud entry and free pmd page. * @pud: Pointer to a PUD. + * @addr: Virtual address associated with pud. * - * Context: The pud range has been unmaped and TLB purged. + * Context: The pud range has been unmapped and TLB purged. * Return: 1 if clearing the entry succeeded. 0 otherwise. + * + * NOTE: Callers must allow a single page allocation. */ -int pud_free_pmd_page(pud_t *pud) +int pud_free_pmd_page(pud_t *pud, unsigned long addr) { - pmd_t *pmd; + pmd_t *pmd, *pmd_sv; + pte_t *pte; int i; if (pud_none(*pud)) return 1; pmd = (pmd_t *)pud_page_vaddr(*pud); + pmd_sv = (pmd_t *)__get_free_page(GFP_KERNEL); + if (!pmd_sv) + return 0; - for (i = 0; i < PTRS_PER_PMD; i++) - if (!pmd_free_pte_page(&pmd[i])) - return 0; + for (i = 0; i < PTRS_PER_PMD; i++) { + pmd_sv[i] = pmd[i]; + if (!pmd_none(pmd[i])) + pmd_clear(&pmd[i]); + } pud_clear(pud); + + /* INVLPG to clear all paging-structure caches */ + flush_tlb_kernel_range(addr, addr + PAGE_SIZE-1); + + for (i = 0; i < PTRS_PER_PMD; i++) { + if (!pmd_none(pmd_sv[i])) { + pte = (pte_t *)pmd_page_vaddr(pmd_sv[i]); + free_page((unsigned long)pte); + } + } + + free_page((unsigned long)pmd_sv); free_page((unsigned long)pmd); return 1; @@ -749,11 +831,12 @@ int pud_free_pmd_page(pud_t *pud) /** * pmd_free_pte_page - Clear pmd entry and free pte page. * @pmd: Pointer to a PMD. + * @addr: Virtual address associated with pmd. * - * Context: The pmd range has been unmaped and TLB purged. + * Context: The pmd range has been unmapped and TLB purged. * Return: 1 if clearing the entry succeeded. 0 otherwise. */ -int pmd_free_pte_page(pmd_t *pmd) +int pmd_free_pte_page(pmd_t *pmd, unsigned long addr) { pte_t *pte; @@ -762,8 +845,30 @@ int pmd_free_pte_page(pmd_t *pmd) pte = (pte_t *)pmd_page_vaddr(*pmd); pmd_clear(pmd); + + /* INVLPG to clear all paging-structure caches */ + flush_tlb_kernel_range(addr, addr + PAGE_SIZE-1); + free_page((unsigned long)pte); return 1; } + +#else /* !CONFIG_X86_64 */ + +int pud_free_pmd_page(pud_t *pud, unsigned long addr) +{ + return pud_none(*pud); +} + +/* + * Disable free page handling on x86-PAE. This assures that ioremap() + * does not update sync'd pmd entries. See vmalloc_sync_one(). + */ +int pmd_free_pte_page(pmd_t *pmd, unsigned long addr) +{ + return pmd_none(*pmd); +} + +#endif /* CONFIG_X86_64 */ #endif /* CONFIG_HAVE_ARCH_HUGE_VMAP */ diff --git a/arch/x86/mm/pti.c b/arch/x86/mm/pti.c index 4d418e705878..31341ae7309f 100644 --- a/arch/x86/mm/pti.c +++ b/arch/x86/mm/pti.c @@ -45,6 +45,7 @@ #include <asm/pgalloc.h> #include <asm/tlbflush.h> #include <asm/desc.h> +#include <asm/sections.h> #undef pr_fmt #define pr_fmt(fmt) "Kernel/User page tables isolation: " fmt @@ -54,6 +55,16 @@ #define __GFP_NOTRACK 0 #endif +/* + * Define the page-table levels we clone for user-space on 32 + * and 64 bit. + */ +#ifdef CONFIG_X86_64 +#define PTI_LEVEL_KERNEL_IMAGE PTI_CLONE_PMD +#else +#define PTI_LEVEL_KERNEL_IMAGE PTI_CLONE_PTE +#endif + static void __init pti_print_if_insecure(const char *reason) { if (boot_cpu_has_bug(X86_BUG_CPU_MELTDOWN)) @@ -117,7 +128,7 @@ enable: setup_force_cpu_cap(X86_FEATURE_PTI); } -pgd_t __pti_set_user_pgd(pgd_t *pgdp, pgd_t pgd) +pgd_t __pti_set_user_pgtbl(pgd_t *pgdp, pgd_t pgd) { /* * Changes to the high (kernel) portion of the kernelmode page @@ -176,7 +187,7 @@ static p4d_t *pti_user_pagetable_walk_p4d(unsigned long address) if (pgd_none(*pgd)) { unsigned long new_p4d_page = __get_free_page(gfp); - if (!new_p4d_page) + if (WARN_ON_ONCE(!new_p4d_page)) return NULL; set_pgd(pgd, __pgd(_KERNPG_TABLE | __pa(new_p4d_page))); @@ -195,13 +206,17 @@ static p4d_t *pti_user_pagetable_walk_p4d(unsigned long address) static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address) { gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO); - p4d_t *p4d = pti_user_pagetable_walk_p4d(address); + p4d_t *p4d; pud_t *pud; + p4d = pti_user_pagetable_walk_p4d(address); + if (!p4d) + return NULL; + BUILD_BUG_ON(p4d_large(*p4d) != 0); if (p4d_none(*p4d)) { unsigned long new_pud_page = __get_free_page(gfp); - if (!new_pud_page) + if (WARN_ON_ONCE(!new_pud_page)) return NULL; set_p4d(p4d, __p4d(_KERNPG_TABLE | __pa(new_pud_page))); @@ -215,7 +230,7 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address) } if (pud_none(*pud)) { unsigned long new_pmd_page = __get_free_page(gfp); - if (!new_pmd_page) + if (WARN_ON_ONCE(!new_pmd_page)) return NULL; set_pud(pud, __pud(_KERNPG_TABLE | __pa(new_pmd_page))); @@ -224,7 +239,6 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address) return pmd_offset(pud, address); } -#ifdef CONFIG_X86_VSYSCALL_EMULATION /* * Walk the shadow copy of the page tables (optionally) trying to allocate * page table pages on the way down. Does not support large pages. @@ -237,9 +251,13 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address) static __init pte_t *pti_user_pagetable_walk_pte(unsigned long address) { gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO); - pmd_t *pmd = pti_user_pagetable_walk_pmd(address); + pmd_t *pmd; pte_t *pte; + pmd = pti_user_pagetable_walk_pmd(address); + if (!pmd) + return NULL; + /* We can't do anything sensible if we hit a large mapping. */ if (pmd_large(*pmd)) { WARN_ON(1); @@ -262,6 +280,7 @@ static __init pte_t *pti_user_pagetable_walk_pte(unsigned long address) return pte; } +#ifdef CONFIG_X86_VSYSCALL_EMULATION static void __init pti_setup_vsyscall(void) { pte_t *pte, *target_pte; @@ -282,8 +301,14 @@ static void __init pti_setup_vsyscall(void) static void __init pti_setup_vsyscall(void) { } #endif +enum pti_clone_level { + PTI_CLONE_PMD, + PTI_CLONE_PTE, +}; + static void -pti_clone_pmds(unsigned long start, unsigned long end, pmdval_t clear) +pti_clone_pgtable(unsigned long start, unsigned long end, + enum pti_clone_level level) { unsigned long addr; @@ -291,59 +316,105 @@ pti_clone_pmds(unsigned long start, unsigned long end, pmdval_t clear) * Clone the populated PMDs which cover start to end. These PMD areas * can have holes. */ - for (addr = start; addr < end; addr += PMD_SIZE) { + for (addr = start; addr < end;) { + pte_t *pte, *target_pte; pmd_t *pmd, *target_pmd; pgd_t *pgd; p4d_t *p4d; pud_t *pud; + /* Overflow check */ + if (addr < start) + break; + pgd = pgd_offset_k(addr); if (WARN_ON(pgd_none(*pgd))) return; p4d = p4d_offset(pgd, addr); if (WARN_ON(p4d_none(*p4d))) return; + pud = pud_offset(p4d, addr); - if (pud_none(*pud)) + if (pud_none(*pud)) { + addr += PUD_SIZE; continue; + } + pmd = pmd_offset(pud, addr); - if (pmd_none(*pmd)) + if (pmd_none(*pmd)) { + addr += PMD_SIZE; continue; + } - target_pmd = pti_user_pagetable_walk_pmd(addr); - if (WARN_ON(!target_pmd)) - return; - - /* - * Only clone present PMDs. This ensures only setting - * _PAGE_GLOBAL on present PMDs. This should only be - * called on well-known addresses anyway, so a non- - * present PMD would be a surprise. - */ - if (WARN_ON(!(pmd_flags(*pmd) & _PAGE_PRESENT))) - return; - - /* - * Setting 'target_pmd' below creates a mapping in both - * the user and kernel page tables. It is effectively - * global, so set it as global in both copies. Note: - * the X86_FEATURE_PGE check is not _required_ because - * the CPU ignores _PAGE_GLOBAL when PGE is not - * supported. The check keeps consistentency with - * code that only set this bit when supported. - */ - if (boot_cpu_has(X86_FEATURE_PGE)) - *pmd = pmd_set_flags(*pmd, _PAGE_GLOBAL); - - /* - * Copy the PMD. That is, the kernelmode and usermode - * tables will share the last-level page tables of this - * address range - */ - *target_pmd = pmd_clear_flags(*pmd, clear); + if (pmd_large(*pmd) || level == PTI_CLONE_PMD) { + target_pmd = pti_user_pagetable_walk_pmd(addr); + if (WARN_ON(!target_pmd)) + return; + + /* + * Only clone present PMDs. This ensures only setting + * _PAGE_GLOBAL on present PMDs. This should only be + * called on well-known addresses anyway, so a non- + * present PMD would be a surprise. + */ + if (WARN_ON(!(pmd_flags(*pmd) & _PAGE_PRESENT))) + return; + + /* + * Setting 'target_pmd' below creates a mapping in both + * the user and kernel page tables. It is effectively + * global, so set it as global in both copies. Note: + * the X86_FEATURE_PGE check is not _required_ because + * the CPU ignores _PAGE_GLOBAL when PGE is not + * supported. The check keeps consistentency with + * code that only set this bit when supported. + */ + if (boot_cpu_has(X86_FEATURE_PGE)) + *pmd = pmd_set_flags(*pmd, _PAGE_GLOBAL); + + /* + * Copy the PMD. That is, the kernelmode and usermode + * tables will share the last-level page tables of this + * address range + */ + *target_pmd = *pmd; + + addr += PMD_SIZE; + + } else if (level == PTI_CLONE_PTE) { + + /* Walk the page-table down to the pte level */ + pte = pte_offset_kernel(pmd, addr); + if (pte_none(*pte)) { + addr += PAGE_SIZE; + continue; + } + + /* Only clone present PTEs */ + if (WARN_ON(!(pte_flags(*pte) & _PAGE_PRESENT))) + return; + + /* Allocate PTE in the user page-table */ + target_pte = pti_user_pagetable_walk_pte(addr); + if (WARN_ON(!target_pte)) + return; + + /* Set GLOBAL bit in both PTEs */ + if (boot_cpu_has(X86_FEATURE_PGE)) + *pte = pte_set_flags(*pte, _PAGE_GLOBAL); + + /* Clone the PTE */ + *target_pte = *pte; + + addr += PAGE_SIZE; + + } else { + BUG(); + } } } +#ifdef CONFIG_X86_64 /* * Clone a single p4d (i.e. a top-level entry on 4-level systems and a * next-level entry on 5-level systems. @@ -354,6 +425,9 @@ static void __init pti_clone_p4d(unsigned long addr) pgd_t *kernel_pgd; user_p4d = pti_user_pagetable_walk_p4d(addr); + if (!user_p4d) + return; + kernel_pgd = pgd_offset_k(addr); kernel_p4d = p4d_offset(kernel_pgd, addr); *user_p4d = *kernel_p4d; @@ -367,6 +441,25 @@ static void __init pti_clone_user_shared(void) pti_clone_p4d(CPU_ENTRY_AREA_BASE); } +#else /* CONFIG_X86_64 */ + +/* + * On 32 bit PAE systems with 1GB of Kernel address space there is only + * one pgd/p4d for the whole kernel. Cloning that would map the whole + * address space into the user page-tables, making PTI useless. So clone + * the page-table on the PMD level to prevent that. + */ +static void __init pti_clone_user_shared(void) +{ + unsigned long start, end; + + start = CPU_ENTRY_AREA_BASE; + end = start + (PAGE_SIZE * CPU_ENTRY_AREA_PAGES); + + pti_clone_pgtable(start, end, PTI_CLONE_PMD); +} +#endif /* CONFIG_X86_64 */ + /* * Clone the ESPFIX P4D into the user space visible page table */ @@ -380,11 +473,11 @@ static void __init pti_setup_espfix64(void) /* * Clone the populated PMDs of the entry and irqentry text and force it RO. */ -static void __init pti_clone_entry_text(void) +static void pti_clone_entry_text(void) { - pti_clone_pmds((unsigned long) __entry_text_start, - (unsigned long) __irqentry_text_end, - _PAGE_RW); + pti_clone_pgtable((unsigned long) __entry_text_start, + (unsigned long) __irqentry_text_end, + PTI_CLONE_PMD); } /* @@ -435,10 +528,17 @@ static inline bool pti_kernel_image_global_ok(void) } /* + * This is the only user for these and it is not arch-generic + * like the other set_memory.h functions. Just extern them. + */ +extern int set_memory_nonglobal(unsigned long addr, int numpages); +extern int set_memory_global(unsigned long addr, int numpages); + +/* * For some configurations, map all of kernel text into the user page * tables. This reduces TLB misses, especially on non-PCID systems. */ -void pti_clone_kernel_text(void) +static void pti_clone_kernel_text(void) { /* * rodata is part of the kernel image and is normally @@ -446,7 +546,8 @@ void pti_clone_kernel_text(void) * clone the areas past rodata, they might contain secrets. */ unsigned long start = PFN_ALIGN(_text); - unsigned long end = (unsigned long)__end_rodata_hpage_align; + unsigned long end_clone = (unsigned long)__end_rodata_aligned; + unsigned long end_global = PFN_ALIGN((unsigned long)__stop___ex_table); if (!pti_kernel_image_global_ok()) return; @@ -458,14 +559,18 @@ void pti_clone_kernel_text(void) * pti_set_kernel_image_nonglobal() did to clear the * global bit. */ - pti_clone_pmds(start, end, _PAGE_RW); + pti_clone_pgtable(start, end_clone, PTI_LEVEL_KERNEL_IMAGE); + + /* + * pti_clone_pgtable() will set the global bit in any PMDs + * that it clones, but we also need to get any PTEs in + * the last level for areas that are not huge-page-aligned. + */ + + /* Set the global bit for normal non-__init kernel text: */ + set_memory_global(start, (end_global - start) >> PAGE_SHIFT); } -/* - * This is the only user for it and it is not arch-generic like - * the other set_memory.h functions. Just extern it. - */ -extern int set_memory_nonglobal(unsigned long addr, int numpages); void pti_set_kernel_image_nonglobal(void) { /* @@ -477,9 +582,11 @@ void pti_set_kernel_image_nonglobal(void) unsigned long start = PFN_ALIGN(_text); unsigned long end = ALIGN((unsigned long)_end, PMD_PAGE_SIZE); - if (pti_kernel_image_global_ok()) - return; - + /* + * This clears _PAGE_GLOBAL from the entire kernel image. + * pti_clone_kernel_text() map put _PAGE_GLOBAL back for + * areas that are mapped to userspace. + */ set_memory_nonglobal(start, (end - start) >> PAGE_SHIFT); } @@ -493,6 +600,28 @@ void __init pti_init(void) pr_info("enabled\n"); +#ifdef CONFIG_X86_32 + /* + * We check for X86_FEATURE_PCID here. But the init-code will + * clear the feature flag on 32 bit because the feature is not + * supported on 32 bit anyway. To print the warning we need to + * check with cpuid directly again. + */ + if (cpuid_ecx(0x1) & BIT(17)) { + /* Use printk to work around pr_fmt() */ + printk(KERN_WARNING "\n"); + printk(KERN_WARNING "************************************************************\n"); + printk(KERN_WARNING "** WARNING! WARNING! WARNING! WARNING! WARNING! WARNING! **\n"); + printk(KERN_WARNING "** **\n"); + printk(KERN_WARNING "** You are using 32-bit PTI on a 64-bit PCID-capable CPU. **\n"); + printk(KERN_WARNING "** Your performance will increase dramatically if you **\n"); + printk(KERN_WARNING "** switch to a 64-bit kernel! **\n"); + printk(KERN_WARNING "** **\n"); + printk(KERN_WARNING "** WARNING! WARNING! WARNING! WARNING! WARNING! WARNING! **\n"); + printk(KERN_WARNING "************************************************************\n"); + } +#endif + pti_clone_user_shared(); /* Undo all global bits from the init pagetables in head_64.S: */ @@ -502,3 +631,22 @@ void __init pti_init(void) pti_setup_espfix64(); pti_setup_vsyscall(); } + +/* + * Finalize the kernel mappings in the userspace page-table. Some of the + * mappings for the kernel image might have changed since pti_init() + * cloned them. This is because parts of the kernel image have been + * mapped RO and/or NX. These changes need to be cloned again to the + * userspace page-table. + */ +void pti_finalize(void) +{ + /* + * We need to clone everything (again) that maps parts of the + * kernel image. + */ + pti_clone_entry_text(); + pti_clone_kernel_text(); + + debug_checkwx_user(); +} diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 6eb1f34c3c85..752dbf4e0e50 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -7,6 +7,7 @@ #include <linux/export.h> #include <linux/cpu.h> #include <linux/debugfs.h> +#include <linux/gfp.h> #include <asm/tlbflush.h> #include <asm/mmu_context.h> @@ -35,7 +36,7 @@ * necessary invalidation by clearing out the 'ctx_id' which * forces a TLB flush when the context is loaded. */ -void clear_asid_other(void) +static void clear_asid_other(void) { u16 asid; @@ -185,8 +186,11 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, { struct mm_struct *real_prev = this_cpu_read(cpu_tlbstate.loaded_mm); u16 prev_asid = this_cpu_read(cpu_tlbstate.loaded_mm_asid); + bool was_lazy = this_cpu_read(cpu_tlbstate.is_lazy); unsigned cpu = smp_processor_id(); u64 next_tlb_gen; + bool need_flush; + u16 new_asid; /* * NB: The scheduler will call us with prev == next when switching @@ -240,20 +244,41 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, next->context.ctx_id); /* - * We don't currently support having a real mm loaded without - * our cpu set in mm_cpumask(). We have all the bookkeeping - * in place to figure out whether we would need to flush - * if our cpu were cleared in mm_cpumask(), but we don't - * currently use it. + * Even in lazy TLB mode, the CPU should stay set in the + * mm_cpumask. The TLB shootdown code can figure out from + * from cpu_tlbstate.is_lazy whether or not to send an IPI. */ if (WARN_ON_ONCE(real_prev != &init_mm && !cpumask_test_cpu(cpu, mm_cpumask(next)))) cpumask_set_cpu(cpu, mm_cpumask(next)); - return; + /* + * If the CPU is not in lazy TLB mode, we are just switching + * from one thread in a process to another thread in the same + * process. No TLB flush required. + */ + if (!was_lazy) + return; + + /* + * Read the tlb_gen to check whether a flush is needed. + * If the TLB is up to date, just use it. + * The barrier synchronizes with the tlb_gen increment in + * the TLB shootdown code. + */ + smp_mb(); + next_tlb_gen = atomic64_read(&next->context.tlb_gen); + if (this_cpu_read(cpu_tlbstate.ctxs[prev_asid].tlb_gen) == + next_tlb_gen) + return; + + /* + * TLB contents went out of date while we were in lazy + * mode. Fall through to the TLB switching code below. + */ + new_asid = prev_asid; + need_flush = true; } else { - u16 new_asid; - bool need_flush; u64 last_ctx_id = this_cpu_read(cpu_tlbstate.last_ctx_id); /* @@ -285,53 +310,60 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, sync_current_stack_to_mm(next); } - /* Stop remote flushes for the previous mm */ - VM_WARN_ON_ONCE(!cpumask_test_cpu(cpu, mm_cpumask(real_prev)) && - real_prev != &init_mm); - cpumask_clear_cpu(cpu, mm_cpumask(real_prev)); + /* + * Stop remote flushes for the previous mm. + * Skip kernel threads; we never send init_mm TLB flushing IPIs, + * but the bitmap manipulation can cause cache line contention. + */ + if (real_prev != &init_mm) { + VM_WARN_ON_ONCE(!cpumask_test_cpu(cpu, + mm_cpumask(real_prev))); + cpumask_clear_cpu(cpu, mm_cpumask(real_prev)); + } /* * Start remote flushes and then read tlb_gen. */ - cpumask_set_cpu(cpu, mm_cpumask(next)); + if (next != &init_mm) + cpumask_set_cpu(cpu, mm_cpumask(next)); next_tlb_gen = atomic64_read(&next->context.tlb_gen); choose_new_asid(next, next_tlb_gen, &new_asid, &need_flush); + } - if (need_flush) { - this_cpu_write(cpu_tlbstate.ctxs[new_asid].ctx_id, next->context.ctx_id); - this_cpu_write(cpu_tlbstate.ctxs[new_asid].tlb_gen, next_tlb_gen); - load_new_mm_cr3(next->pgd, new_asid, true); - - /* - * NB: This gets called via leave_mm() in the idle path - * where RCU functions differently. Tracing normally - * uses RCU, so we need to use the _rcuidle variant. - * - * (There is no good reason for this. The idle code should - * be rearranged to call this before rcu_idle_enter().) - */ - trace_tlb_flush_rcuidle(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); - } else { - /* The new ASID is already up to date. */ - load_new_mm_cr3(next->pgd, new_asid, false); - - /* See above wrt _rcuidle. */ - trace_tlb_flush_rcuidle(TLB_FLUSH_ON_TASK_SWITCH, 0); - } + if (need_flush) { + this_cpu_write(cpu_tlbstate.ctxs[new_asid].ctx_id, next->context.ctx_id); + this_cpu_write(cpu_tlbstate.ctxs[new_asid].tlb_gen, next_tlb_gen); + load_new_mm_cr3(next->pgd, new_asid, true); /* - * Record last user mm's context id, so we can avoid - * flushing branch buffer with IBPB if we switch back - * to the same user. + * NB: This gets called via leave_mm() in the idle path + * where RCU functions differently. Tracing normally + * uses RCU, so we need to use the _rcuidle variant. + * + * (There is no good reason for this. The idle code should + * be rearranged to call this before rcu_idle_enter().) */ - if (next != &init_mm) - this_cpu_write(cpu_tlbstate.last_ctx_id, next->context.ctx_id); + trace_tlb_flush_rcuidle(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); + } else { + /* The new ASID is already up to date. */ + load_new_mm_cr3(next->pgd, new_asid, false); - this_cpu_write(cpu_tlbstate.loaded_mm, next); - this_cpu_write(cpu_tlbstate.loaded_mm_asid, new_asid); + /* See above wrt _rcuidle. */ + trace_tlb_flush_rcuidle(TLB_FLUSH_ON_TASK_SWITCH, 0); } + /* + * Record last user mm's context id, so we can avoid + * flushing branch buffer with IBPB if we switch back + * to the same user. + */ + if (next != &init_mm) + this_cpu_write(cpu_tlbstate.last_ctx_id, next->context.ctx_id); + + this_cpu_write(cpu_tlbstate.loaded_mm, next); + this_cpu_write(cpu_tlbstate.loaded_mm_asid, new_asid); + load_mm_cr4(next); switch_ldt(real_prev, next); } @@ -354,20 +386,7 @@ void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk) if (this_cpu_read(cpu_tlbstate.loaded_mm) == &init_mm) return; - if (tlb_defer_switch_to_init_mm()) { - /* - * There's a significant optimization that may be possible - * here. We have accurate enough TLB flush tracking that we - * don't need to maintain coherence of TLB per se when we're - * lazy. We do, however, need to maintain coherence of - * paging-structure caches. We could, in principle, leave our - * old mm loaded and only switch to init_mm when - * tlb_remove_page() happens. - */ - this_cpu_write(cpu_tlbstate.is_lazy, true); - } else { - switch_mm(NULL, &init_mm, NULL); - } + this_cpu_write(cpu_tlbstate.is_lazy, true); } /* @@ -454,6 +473,9 @@ static void flush_tlb_func_common(const struct flush_tlb_info *f, * paging-structure cache to avoid speculatively reading * garbage into our TLB. Since switching to init_mm is barely * slower than a minimal flush, just switch to init_mm. + * + * This should be rare, with native_flush_tlb_others skipping + * IPIs to lazy TLB mode CPUs. */ switch_mm_irqs_off(NULL, &init_mm, NULL); return; @@ -560,6 +582,9 @@ static void flush_tlb_func_remote(void *info) void native_flush_tlb_others(const struct cpumask *cpumask, const struct flush_tlb_info *info) { + cpumask_var_t lazymask; + unsigned int cpu; + count_vm_tlb_event(NR_TLB_REMOTE_FLUSH); if (info->end == TLB_FLUSH_ALL) trace_tlb_flush(TLB_REMOTE_SEND_IPI, TLB_FLUSH_ALL); @@ -583,8 +608,6 @@ void native_flush_tlb_others(const struct cpumask *cpumask, * that UV should be updated so that smp_call_function_many(), * etc, are optimal on UV. */ - unsigned int cpu; - cpu = smp_processor_id(); cpumask = uv_flush_tlb_others(cpumask, info); if (cpumask) @@ -592,8 +615,29 @@ void native_flush_tlb_others(const struct cpumask *cpumask, (void *)info, 1); return; } - smp_call_function_many(cpumask, flush_tlb_func_remote, + + /* + * A temporary cpumask is used in order to skip sending IPIs + * to CPUs in lazy TLB state, while keeping them in mm_cpumask(mm). + * If the allocation fails, simply IPI every CPU in mm_cpumask. + */ + if (!alloc_cpumask_var(&lazymask, GFP_ATOMIC)) { + smp_call_function_many(cpumask, flush_tlb_func_remote, (void *)info, 1); + return; + } + + cpumask_copy(lazymask, cpumask); + + for_each_cpu(cpu, lazymask) { + if (per_cpu(cpu_tlbstate.is_lazy, cpu)) + cpumask_clear_cpu(cpu, lazymask); + } + + smp_call_function_many(lazymask, flush_tlb_func_remote, + (void *)info, 1); + + free_cpumask_var(lazymask); } /* @@ -646,6 +690,68 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, put_cpu(); } +void tlb_flush_remove_tables_local(void *arg) +{ + struct mm_struct *mm = arg; + + if (this_cpu_read(cpu_tlbstate.loaded_mm) == mm && + this_cpu_read(cpu_tlbstate.is_lazy)) { + /* + * We're in lazy mode. We need to at least flush our + * paging-structure cache to avoid speculatively reading + * garbage into our TLB. Since switching to init_mm is barely + * slower than a minimal flush, just switch to init_mm. + */ + switch_mm_irqs_off(NULL, &init_mm, NULL); + } +} + +static void mm_fill_lazy_tlb_cpu_mask(struct mm_struct *mm, + struct cpumask *lazy_cpus) +{ + int cpu; + + for_each_cpu(cpu, mm_cpumask(mm)) { + if (!per_cpu(cpu_tlbstate.is_lazy, cpu)) + cpumask_set_cpu(cpu, lazy_cpus); + } +} + +void tlb_flush_remove_tables(struct mm_struct *mm) +{ + int cpu = get_cpu(); + cpumask_var_t lazy_cpus; + + if (cpumask_any_but(mm_cpumask(mm), cpu) >= nr_cpu_ids) { + put_cpu(); + return; + } + + if (!zalloc_cpumask_var(&lazy_cpus, GFP_ATOMIC)) { + /* + * If the cpumask allocation fails, do a brute force flush + * on all the CPUs that have this mm loaded. + */ + smp_call_function_many(mm_cpumask(mm), + tlb_flush_remove_tables_local, (void *)mm, 1); + put_cpu(); + return; + } + + /* + * CPUs with !is_lazy either received a TLB flush IPI while the user + * pages in this address range were unmapped, or have context switched + * and reloaded %CR3 since then. + * + * Shootdown IPIs at page table freeing time only need to be sent to + * CPUs that may have out of date TLB contents. + */ + mm_fill_lazy_tlb_cpu_mask(mm, lazy_cpus); + smp_call_function_many(lazy_cpus, + tlb_flush_remove_tables_local, (void *)mm, 1); + free_cpumask_var(lazy_cpus); + put_cpu(); +} static void do_flush_tlb_all(void *info) { |