diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2017-09-04 12:21:28 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2017-09-04 12:21:28 -0700 |
commit | b1b6f83ac938d176742c85757960dec2cf10e468 (patch) | |
tree | f99e605318232a9327500896b9187b5ec9cad0c1 /arch/x86/mm | |
parent | 5f82e71a001d14824a7728ad9e49f6aea420f161 (diff) | |
parent | 9e52fc2b50de3a1c08b44f94c610fbe998c0031a (diff) | |
download | linux-stable-b1b6f83ac938d176742c85757960dec2cf10e468.tar.gz linux-stable-b1b6f83ac938d176742c85757960dec2cf10e468.tar.bz2 linux-stable-b1b6f83ac938d176742c85757960dec2cf10e468.zip |
Merge branch 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 mm changes from Ingo Molnar:
"PCID support, 5-level paging support, Secure Memory Encryption support
The main changes in this cycle are support for three new, complex
hardware features of x86 CPUs:
- Add 5-level paging support, which is a new hardware feature on
upcoming Intel CPUs allowing up to 128 PB of virtual address space
and 4 PB of physical RAM space - a 512-fold increase over the old
limits. (Supercomputers of the future forecasting hurricanes on an
ever warming planet can certainly make good use of more RAM.)
Many of the necessary changes went upstream in previous cycles,
v4.14 is the first kernel that can enable 5-level paging.
This feature is activated via CONFIG_X86_5LEVEL=y - disabled by
default.
(By Kirill A. Shutemov)
- Add 'encrypted memory' support, which is a new hardware feature on
upcoming AMD CPUs ('Secure Memory Encryption', SME) allowing system
RAM to be encrypted and decrypted (mostly) transparently by the
CPU, with a little help from the kernel to transition to/from
encrypted RAM. Such RAM should be more secure against various
attacks like RAM access via the memory bus and should make the
radio signature of memory bus traffic harder to intercept (and
decrypt) as well.
This feature is activated via CONFIG_AMD_MEM_ENCRYPT=y - disabled
by default.
(By Tom Lendacky)
- Enable PCID optimized TLB flushing on newer Intel CPUs: PCID is a
hardware feature that attaches an address space tag to TLB entries
and thus allows to skip TLB flushing in many cases, even if we
switch mm's.
(By Andy Lutomirski)
All three of these features were in the works for a long time, and
it's coincidence of the three independent development paths that they
are all enabled in v4.14 at once"
* 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (65 commits)
x86/mm: Enable RCU based page table freeing (CONFIG_HAVE_RCU_TABLE_FREE=y)
x86/mm: Use pr_cont() in dump_pagetable()
x86/mm: Fix SME encryption stack ptr handling
kvm/x86: Avoid clearing the C-bit in rsvd_bits()
x86/CPU: Align CR3 defines
x86/mm, mm/hwpoison: Clear PRESENT bit for kernel 1:1 mappings of poison pages
acpi, x86/mm: Remove encryption mask from ACPI page protection type
x86/mm, kexec: Fix memory corruption with SME on successive kexecs
x86/mm/pkeys: Fix typo in Documentation/x86/protection-keys.txt
x86/mm/dump_pagetables: Speed up page tables dump for CONFIG_KASAN=y
x86/mm: Implement PCID based optimization: try to preserve old TLB entries using PCID
x86: Enable 5-level paging support via CONFIG_X86_5LEVEL=y
x86/mm: Allow userspace have mappings above 47-bit
x86/mm: Prepare to expose larger address space to userspace
x86/mpx: Do not allow MPX if we have mappings above 47-bit
x86/mm: Rename tasksize_32bit/64bit to task_size_32bit/64bit()
x86/xen: Redefine XEN_ELFNOTE_INIT_P2M using PUD_SIZE * PTRS_PER_PUD
x86/mm/dump_pagetables: Fix printout of p4d level
x86/mm/dump_pagetables: Generalize address normalization
x86/boot: Fix memremap() related build failure
...
Diffstat (limited to 'arch/x86/mm')
-rw-r--r-- | arch/x86/mm/Makefile | 2 | ||||
-rw-r--r-- | arch/x86/mm/dump_pagetables.c | 93 | ||||
-rw-r--r-- | arch/x86/mm/fault.c | 26 | ||||
-rw-r--r-- | arch/x86/mm/hugetlbpage.c | 27 | ||||
-rw-r--r-- | arch/x86/mm/ident_map.c | 12 | ||||
-rw-r--r-- | arch/x86/mm/init.c | 2 | ||||
-rw-r--r-- | arch/x86/mm/ioremap.c | 287 | ||||
-rw-r--r-- | arch/x86/mm/kasan_init_64.c | 6 | ||||
-rw-r--r-- | arch/x86/mm/mem_encrypt.c | 593 | ||||
-rw-r--r-- | arch/x86/mm/mem_encrypt_boot.S | 149 | ||||
-rw-r--r-- | arch/x86/mm/mmap.c | 12 | ||||
-rw-r--r-- | arch/x86/mm/mpx.c | 33 | ||||
-rw-r--r-- | arch/x86/mm/pageattr.c | 67 | ||||
-rw-r--r-- | arch/x86/mm/pat.c | 9 | ||||
-rw-r--r-- | arch/x86/mm/pgtable.c | 8 | ||||
-rw-r--r-- | arch/x86/mm/tlb.c | 331 |
16 files changed, 1481 insertions, 176 deletions
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile index 0fbdcb64f9f8..72bf8c01c6e3 100644 --- a/arch/x86/mm/Makefile +++ b/arch/x86/mm/Makefile @@ -39,3 +39,5 @@ obj-$(CONFIG_X86_INTEL_MPX) += mpx.o obj-$(CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS) += pkeys.o obj-$(CONFIG_RANDOMIZE_MEMORY) += kaslr.o +obj-$(CONFIG_AMD_MEM_ENCRYPT) += mem_encrypt.o +obj-$(CONFIG_AMD_MEM_ENCRYPT) += mem_encrypt_boot.o diff --git a/arch/x86/mm/dump_pagetables.c b/arch/x86/mm/dump_pagetables.c index 0470826d2bdc..5e3ac6fe6c9e 100644 --- a/arch/x86/mm/dump_pagetables.c +++ b/arch/x86/mm/dump_pagetables.c @@ -13,12 +13,12 @@ */ #include <linux/debugfs.h> +#include <linux/kasan.h> #include <linux/mm.h> #include <linux/init.h> #include <linux/sched.h> #include <linux/seq_file.h> -#include <asm/kasan.h> #include <asm/pgtable.h> /* @@ -138,7 +138,7 @@ static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg) { pgprotval_t pr = pgprot_val(prot); static const char * const level_name[] = - { "cr3", "pgd", "pud", "pmd", "pte" }; + { "cr3", "pgd", "p4d", "pud", "pmd", "pte" }; if (!pgprot_val(prot)) { /* Not present */ @@ -162,12 +162,12 @@ static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg) pt_dump_cont_printf(m, dmsg, " "); /* Bit 7 has a different meaning on level 3 vs 4 */ - if (level <= 3 && pr & _PAGE_PSE) + if (level <= 4 && pr & _PAGE_PSE) pt_dump_cont_printf(m, dmsg, "PSE "); else pt_dump_cont_printf(m, dmsg, " "); - if ((level == 4 && pr & _PAGE_PAT) || - ((level == 3 || level == 2) && pr & _PAGE_PAT_LARGE)) + if ((level == 5 && pr & _PAGE_PAT) || + ((level == 4 || level == 3) && pr & _PAGE_PAT_LARGE)) pt_dump_cont_printf(m, dmsg, "PAT "); else pt_dump_cont_printf(m, dmsg, " "); @@ -188,11 +188,12 @@ static void printk_prot(struct seq_file *m, pgprot_t prot, int level, bool dmsg) */ static unsigned long normalize_addr(unsigned long u) { -#ifdef CONFIG_X86_64 - return (signed long)(u << 16) >> 16; -#else - return u; -#endif + int shift; + if (!IS_ENABLED(CONFIG_X86_64)) + return u; + + shift = 64 - (__VIRTUAL_MASK_SHIFT + 1); + return (signed long)(u << shift) >> shift; } /* @@ -297,32 +298,62 @@ static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr, for (i = 0; i < PTRS_PER_PTE; i++) { prot = pte_flags(*start); st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT); - note_page(m, st, __pgprot(prot), 4); + note_page(m, st, __pgprot(prot), 5); start++; } } +#ifdef CONFIG_KASAN + +/* + * This is an optimization for KASAN=y case. Since all kasan page tables + * eventually point to the kasan_zero_page we could call note_page() + * right away without walking through lower level page tables. This saves + * us dozens of seconds (minutes for 5-level config) while checking for + * W+X mapping or reading kernel_page_tables debugfs file. + */ +static inline bool kasan_page_table(struct seq_file *m, struct pg_state *st, + void *pt) +{ + if (__pa(pt) == __pa(kasan_zero_pmd) || +#ifdef CONFIG_X86_5LEVEL + __pa(pt) == __pa(kasan_zero_p4d) || +#endif + __pa(pt) == __pa(kasan_zero_pud)) { + pgprotval_t prot = pte_flags(kasan_zero_pte[0]); + note_page(m, st, __pgprot(prot), 5); + return true; + } + return false; +} +#else +static inline bool kasan_page_table(struct seq_file *m, struct pg_state *st, + void *pt) +{ + return false; +} +#endif #if PTRS_PER_PMD > 1 static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr, unsigned long P) { int i; - pmd_t *start; + pmd_t *start, *pmd_start; pgprotval_t prot; - start = (pmd_t *)pud_page_vaddr(addr); + pmd_start = start = (pmd_t *)pud_page_vaddr(addr); for (i = 0; i < PTRS_PER_PMD; i++) { st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT); if (!pmd_none(*start)) { if (pmd_large(*start) || !pmd_present(*start)) { prot = pmd_flags(*start); - note_page(m, st, __pgprot(prot), 3); - } else { + note_page(m, st, __pgprot(prot), 4); + } else if (!kasan_page_table(m, st, pmd_start)) { walk_pte_level(m, st, *start, P + i * PMD_LEVEL_MULT); } } else - note_page(m, st, __pgprot(0), 3); + note_page(m, st, __pgprot(0), 4); start++; } } @@ -335,39 +366,27 @@ static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr, #if PTRS_PER_PUD > 1 -/* - * This is an optimization for CONFIG_DEBUG_WX=y + CONFIG_KASAN=y - * KASAN fills page tables with the same values. Since there is no - * point in checking page table more than once we just skip repeated - * entries. This saves us dozens of seconds during boot. - */ -static bool pud_already_checked(pud_t *prev_pud, pud_t *pud, bool checkwx) -{ - return checkwx && prev_pud && (pud_val(*prev_pud) == pud_val(*pud)); -} - static void walk_pud_level(struct seq_file *m, struct pg_state *st, p4d_t addr, unsigned long P) { int i; - pud_t *start; + pud_t *start, *pud_start; pgprotval_t prot; pud_t *prev_pud = NULL; - start = (pud_t *)p4d_page_vaddr(addr); + pud_start = start = (pud_t *)p4d_page_vaddr(addr); for (i = 0; i < PTRS_PER_PUD; i++) { st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT); - if (!pud_none(*start) && - !pud_already_checked(prev_pud, start, st->check_wx)) { + if (!pud_none(*start)) { if (pud_large(*start) || !pud_present(*start)) { prot = pud_flags(*start); - note_page(m, st, __pgprot(prot), 2); - } else { + note_page(m, st, __pgprot(prot), 3); + } else if (!kasan_page_table(m, st, pud_start)) { walk_pmd_level(m, st, *start, P + i * PUD_LEVEL_MULT); } } else - note_page(m, st, __pgprot(0), 2); + note_page(m, st, __pgprot(0), 3); prev_pud = start; start++; @@ -385,10 +404,10 @@ static void walk_pud_level(struct seq_file *m, struct pg_state *st, p4d_t addr, static void walk_p4d_level(struct seq_file *m, struct pg_state *st, pgd_t addr, unsigned long P) { int i; - p4d_t *start; + p4d_t *start, *p4d_start; pgprotval_t prot; - start = (p4d_t *)pgd_page_vaddr(addr); + p4d_start = start = (p4d_t *)pgd_page_vaddr(addr); for (i = 0; i < PTRS_PER_P4D; i++) { st->current_address = normalize_addr(P + i * P4D_LEVEL_MULT); @@ -396,7 +415,7 @@ static void walk_p4d_level(struct seq_file *m, struct pg_state *st, pgd_t addr, if (p4d_large(*start) || !p4d_present(*start)) { prot = p4d_flags(*start); note_page(m, st, __pgprot(prot), 2); - } else { + } else if (!kasan_page_table(m, st, p4d_start)) { walk_pud_level(m, st, *start, P + i * P4D_LEVEL_MULT); } diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c index 2a1fa10c6a98..0cdf14cf3270 100644 --- a/arch/x86/mm/fault.c +++ b/arch/x86/mm/fault.c @@ -396,14 +396,18 @@ static void dump_pagetable(unsigned long address) pte_t *pte; #ifdef CONFIG_X86_PAE - printk("*pdpt = %016Lx ", pgd_val(*pgd)); + pr_info("*pdpt = %016Lx ", pgd_val(*pgd)); if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd)) goto out; +#define pr_pde pr_cont +#else +#define pr_pde pr_info #endif p4d = p4d_offset(pgd, address); pud = pud_offset(p4d, address); pmd = pmd_offset(pud, address); - printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd)); + pr_pde("*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd)); +#undef pr_pde /* * We must not directly access the pte in the highpte @@ -415,9 +419,9 @@ static void dump_pagetable(unsigned long address) goto out; pte = pte_offset_kernel(pmd, address); - printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte)); + pr_cont("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte)); out: - printk("\n"); + pr_cont("\n"); } #else /* CONFIG_X86_64: */ @@ -565,7 +569,7 @@ static void dump_pagetable(unsigned long address) if (bad_address(pgd)) goto bad; - printk("PGD %lx ", pgd_val(*pgd)); + pr_info("PGD %lx ", pgd_val(*pgd)); if (!pgd_present(*pgd)) goto out; @@ -574,7 +578,7 @@ static void dump_pagetable(unsigned long address) if (bad_address(p4d)) goto bad; - printk("P4D %lx ", p4d_val(*p4d)); + pr_cont("P4D %lx ", p4d_val(*p4d)); if (!p4d_present(*p4d) || p4d_large(*p4d)) goto out; @@ -582,7 +586,7 @@ static void dump_pagetable(unsigned long address) if (bad_address(pud)) goto bad; - printk("PUD %lx ", pud_val(*pud)); + pr_cont("PUD %lx ", pud_val(*pud)); if (!pud_present(*pud) || pud_large(*pud)) goto out; @@ -590,7 +594,7 @@ static void dump_pagetable(unsigned long address) if (bad_address(pmd)) goto bad; - printk("PMD %lx ", pmd_val(*pmd)); + pr_cont("PMD %lx ", pmd_val(*pmd)); if (!pmd_present(*pmd) || pmd_large(*pmd)) goto out; @@ -598,12 +602,12 @@ static void dump_pagetable(unsigned long address) if (bad_address(pte)) goto bad; - printk("PTE %lx", pte_val(*pte)); + pr_cont("PTE %lx", pte_val(*pte)); out: - printk("\n"); + pr_cont("\n"); return; bad: - printk("BAD\n"); + pr_info("BAD\n"); } #endif /* CONFIG_X86_64 */ diff --git a/arch/x86/mm/hugetlbpage.c b/arch/x86/mm/hugetlbpage.c index 2824607df108..6d06cf33e3de 100644 --- a/arch/x86/mm/hugetlbpage.c +++ b/arch/x86/mm/hugetlbpage.c @@ -18,6 +18,7 @@ #include <asm/tlbflush.h> #include <asm/pgalloc.h> #include <asm/elf.h> +#include <asm/mpx.h> #if 0 /* This is just for testing */ struct page * @@ -85,25 +86,38 @@ static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *file, info.flags = 0; info.length = len; info.low_limit = get_mmap_base(1); + + /* + * If hint address is above DEFAULT_MAP_WINDOW, look for unmapped area + * in the full address space. + */ info.high_limit = in_compat_syscall() ? - tasksize_32bit() : tasksize_64bit(); + task_size_32bit() : task_size_64bit(addr > DEFAULT_MAP_WINDOW); + info.align_mask = PAGE_MASK & ~huge_page_mask(h); info.align_offset = 0; return vm_unmapped_area(&info); } static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, - unsigned long addr0, unsigned long len, + unsigned long addr, unsigned long len, unsigned long pgoff, unsigned long flags) { struct hstate *h = hstate_file(file); struct vm_unmapped_area_info info; - unsigned long addr; info.flags = VM_UNMAPPED_AREA_TOPDOWN; info.length = len; info.low_limit = PAGE_SIZE; info.high_limit = get_mmap_base(0); + + /* + * If hint address is above DEFAULT_MAP_WINDOW, look for unmapped area + * in the full address space. + */ + if (addr > DEFAULT_MAP_WINDOW && !in_compat_syscall()) + info.high_limit += TASK_SIZE_MAX - DEFAULT_MAP_WINDOW; + info.align_mask = PAGE_MASK & ~huge_page_mask(h); info.align_offset = 0; addr = vm_unmapped_area(&info); @@ -118,7 +132,7 @@ static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, VM_BUG_ON(addr != -ENOMEM); info.flags = 0; info.low_limit = TASK_UNMAPPED_BASE; - info.high_limit = TASK_SIZE; + info.high_limit = TASK_SIZE_LOW; addr = vm_unmapped_area(&info); } @@ -135,6 +149,11 @@ hugetlb_get_unmapped_area(struct file *file, unsigned long addr, if (len & ~huge_page_mask(h)) return -EINVAL; + + addr = mpx_unmapped_area_check(addr, len, flags); + if (IS_ERR_VALUE(addr)) + return addr; + if (len > TASK_SIZE) return -ENOMEM; diff --git a/arch/x86/mm/ident_map.c b/arch/x86/mm/ident_map.c index adab1595f4bd..31cea988fa36 100644 --- a/arch/x86/mm/ident_map.c +++ b/arch/x86/mm/ident_map.c @@ -51,7 +51,7 @@ static int ident_pud_init(struct x86_mapping_info *info, pud_t *pud_page, if (!pmd) return -ENOMEM; ident_pmd_init(info, pmd, addr, next); - set_pud(pud, __pud(__pa(pmd) | _KERNPG_TABLE)); + set_pud(pud, __pud(__pa(pmd) | info->kernpg_flag)); } return 0; @@ -79,7 +79,7 @@ static int ident_p4d_init(struct x86_mapping_info *info, p4d_t *p4d_page, if (!pud) return -ENOMEM; ident_pud_init(info, pud, addr, next); - set_p4d(p4d, __p4d(__pa(pud) | _KERNPG_TABLE)); + set_p4d(p4d, __p4d(__pa(pud) | info->kernpg_flag)); } return 0; @@ -93,6 +93,10 @@ int kernel_ident_mapping_init(struct x86_mapping_info *info, pgd_t *pgd_page, unsigned long next; int result; + /* Set the default pagetable flags if not supplied */ + if (!info->kernpg_flag) + info->kernpg_flag = _KERNPG_TABLE; + for (; addr < end; addr = next) { pgd_t *pgd = pgd_page + pgd_index(addr); p4d_t *p4d; @@ -116,14 +120,14 @@ int kernel_ident_mapping_init(struct x86_mapping_info *info, pgd_t *pgd_page, if (result) return result; if (IS_ENABLED(CONFIG_X86_5LEVEL)) { - set_pgd(pgd, __pgd(__pa(p4d) | _KERNPG_TABLE)); + set_pgd(pgd, __pgd(__pa(p4d) | info->kernpg_flag)); } else { /* * With p4d folded, pgd is equal to p4d. * The pgd entry has to point to the pud page table in this case. */ pud_t *pud = pud_offset(p4d, 0); - set_pgd(pgd, __pgd(__pa(pud) | _KERNPG_TABLE)); + set_pgd(pgd, __pgd(__pa(pud) | info->kernpg_flag)); } } diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c index bf3f1065d6ad..7777ccc0e9f9 100644 --- a/arch/x86/mm/init.c +++ b/arch/x86/mm/init.c @@ -815,7 +815,7 @@ void __init zone_sizes_init(void) DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = { .loaded_mm = &init_mm, - .state = 0, + .next_asid = 1, .cr4 = ~0UL, /* fail hard if we screw up cr4 shadow initialization */ }; EXPORT_SYMBOL_GPL(cpu_tlbstate); diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c index 4c1b5fd0c7ad..34f0e1847dd6 100644 --- a/arch/x86/mm/ioremap.c +++ b/arch/x86/mm/ioremap.c @@ -13,6 +13,8 @@ #include <linux/slab.h> #include <linux/vmalloc.h> #include <linux/mmiotrace.h> +#include <linux/mem_encrypt.h> +#include <linux/efi.h> #include <asm/set_memory.h> #include <asm/e820/api.h> @@ -21,6 +23,7 @@ #include <asm/tlbflush.h> #include <asm/pgalloc.h> #include <asm/pat.h> +#include <asm/setup.h> #include "physaddr.h" @@ -106,12 +109,6 @@ static void __iomem *__ioremap_caller(resource_size_t phys_addr, } /* - * Don't remap the low PCI/ISA area, it's always mapped.. - */ - if (is_ISA_range(phys_addr, last_addr)) - return (__force void __iomem *)phys_to_virt(phys_addr); - - /* * Don't allow anybody to remap normal RAM that we're using.. */ pfn = phys_addr >> PAGE_SHIFT; @@ -340,13 +337,17 @@ void iounmap(volatile void __iomem *addr) return; /* - * __ioremap special-cases the PCI/ISA range by not instantiating a - * vm_area and by simply returning an address into the kernel mapping - * of ISA space. So handle that here. + * The PCI/ISA range special-casing was removed from __ioremap() + * so this check, in theory, can be removed. However, there are + * cases where iounmap() is called for addresses not obtained via + * ioremap() (vga16fb for example). Add a warning so that these + * cases can be caught and fixed. */ if ((void __force *)addr >= phys_to_virt(ISA_START_ADDRESS) && - (void __force *)addr < phys_to_virt(ISA_END_ADDRESS)) + (void __force *)addr < phys_to_virt(ISA_END_ADDRESS)) { + WARN(1, "iounmap() called for ISA range not obtained using ioremap()\n"); return; + } addr = (volatile void __iomem *) (PAGE_MASK & (unsigned long __force)addr); @@ -399,12 +400,10 @@ void *xlate_dev_mem_ptr(phys_addr_t phys) unsigned long offset = phys & ~PAGE_MASK; void *vaddr; - /* If page is RAM, we can use __va. Otherwise ioremap and unmap. */ - if (page_is_ram(start >> PAGE_SHIFT)) - return __va(phys); + /* memremap() maps if RAM, otherwise falls back to ioremap() */ + vaddr = memremap(start, PAGE_SIZE, MEMREMAP_WB); - vaddr = ioremap_cache(start, PAGE_SIZE); - /* Only add the offset on success and return NULL if the ioremap() failed: */ + /* Only add the offset on success and return NULL if memremap() failed */ if (vaddr) vaddr += offset; @@ -413,11 +412,263 @@ void *xlate_dev_mem_ptr(phys_addr_t phys) void unxlate_dev_mem_ptr(phys_addr_t phys, void *addr) { - if (page_is_ram(phys >> PAGE_SHIFT)) - return; + memunmap((void *)((unsigned long)addr & PAGE_MASK)); +} + +/* + * Examine the physical address to determine if it is an area of memory + * that should be mapped decrypted. If the memory is not part of the + * kernel usable area it was accessed and created decrypted, so these + * areas should be mapped decrypted. And since the encryption key can + * change across reboots, persistent memory should also be mapped + * decrypted. + */ +static bool memremap_should_map_decrypted(resource_size_t phys_addr, + unsigned long size) +{ + int is_pmem; + + /* + * Check if the address is part of a persistent memory region. + * This check covers areas added by E820, EFI and ACPI. + */ + is_pmem = region_intersects(phys_addr, size, IORESOURCE_MEM, + IORES_DESC_PERSISTENT_MEMORY); + if (is_pmem != REGION_DISJOINT) + return true; + + /* + * Check if the non-volatile attribute is set for an EFI + * reserved area. + */ + if (efi_enabled(EFI_BOOT)) { + switch (efi_mem_type(phys_addr)) { + case EFI_RESERVED_TYPE: + if (efi_mem_attributes(phys_addr) & EFI_MEMORY_NV) + return true; + break; + default: + break; + } + } + + /* Check if the address is outside kernel usable area */ + switch (e820__get_entry_type(phys_addr, phys_addr + size - 1)) { + case E820_TYPE_RESERVED: + case E820_TYPE_ACPI: + case E820_TYPE_NVS: + case E820_TYPE_UNUSABLE: + case E820_TYPE_PRAM: + return true; + default: + break; + } + + return false; +} + +/* + * Examine the physical address to determine if it is EFI data. Check + * it against the boot params structure and EFI tables and memory types. + */ +static bool memremap_is_efi_data(resource_size_t phys_addr, + unsigned long size) +{ + u64 paddr; + + /* Check if the address is part of EFI boot/runtime data */ + if (!efi_enabled(EFI_BOOT)) + return false; + + paddr = boot_params.efi_info.efi_memmap_hi; + paddr <<= 32; + paddr |= boot_params.efi_info.efi_memmap; + if (phys_addr == paddr) + return true; + + paddr = boot_params.efi_info.efi_systab_hi; + paddr <<= 32; + paddr |= boot_params.efi_info.efi_systab; + if (phys_addr == paddr) + return true; + + if (efi_is_table_address(phys_addr)) + return true; + + switch (efi_mem_type(phys_addr)) { + case EFI_BOOT_SERVICES_DATA: + case EFI_RUNTIME_SERVICES_DATA: + return true; + default: + break; + } + + return false; +} + +/* + * Examine the physical address to determine if it is boot data by checking + * it against the boot params setup_data chain. + */ +static bool memremap_is_setup_data(resource_size_t phys_addr, + unsigned long size) +{ + struct setup_data *data; + u64 paddr, paddr_next; + + paddr = boot_params.hdr.setup_data; + while (paddr) { + unsigned int len; + + if (phys_addr == paddr) + return true; + + data = memremap(paddr, sizeof(*data), + MEMREMAP_WB | MEMREMAP_DEC); + + paddr_next = data->next; + len = data->len; + + memunmap(data); + + if ((phys_addr > paddr) && (phys_addr < (paddr + len))) + return true; + + paddr = paddr_next; + } + + return false; +} + +/* + * Examine the physical address to determine if it is boot data by checking + * it against the boot params setup_data chain (early boot version). + */ +static bool __init early_memremap_is_setup_data(resource_size_t phys_addr, + unsigned long size) +{ + struct setup_data *data; + u64 paddr, paddr_next; + + paddr = boot_params.hdr.setup_data; + while (paddr) { + unsigned int len; + + if (phys_addr == paddr) + return true; + + data = early_memremap_decrypted(paddr, sizeof(*data)); + + paddr_next = data->next; + len = data->len; + + early_memunmap(data, sizeof(*data)); + + if ((phys_addr > paddr) && (phys_addr < (paddr + len))) + return true; + + paddr = paddr_next; + } + + return false; +} + +/* + * Architecture function to determine if RAM remap is allowed. By default, a + * RAM remap will map the data as encrypted. Determine if a RAM remap should + * not be done so that the data will be mapped decrypted. + */ +bool arch_memremap_can_ram_remap(resource_size_t phys_addr, unsigned long size, + unsigned long flags) +{ + if (!sme_active()) + return true; + + if (flags & MEMREMAP_ENC) + return true; + + if (flags & MEMREMAP_DEC) + return false; + + if (memremap_is_setup_data(phys_addr, size) || + memremap_is_efi_data(phys_addr, size) || + memremap_should_map_decrypted(phys_addr, size)) + return false; + + return true; +} + +/* + * Architecture override of __weak function to adjust the protection attributes + * used when remapping memory. By default, early_memremap() will map the data + * as encrypted. Determine if an encrypted mapping should not be done and set + * the appropriate protection attributes. + */ +pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr, + unsigned long size, + pgprot_t prot) +{ + if (!sme_active()) + return prot; + + if (early_memremap_is_setup_data(phys_addr, size) || + memremap_is_efi_data(phys_addr, size) || + memremap_should_map_decrypted(phys_addr, size)) + prot = pgprot_decrypted(prot); + else + prot = pgprot_encrypted(prot); + + return prot; +} + +bool phys_mem_access_encrypted(unsigned long phys_addr, unsigned long size) +{ + return arch_memremap_can_ram_remap(phys_addr, size, 0); +} + +#ifdef CONFIG_ARCH_USE_MEMREMAP_PROT +/* Remap memory with encryption */ +void __init *early_memremap_encrypted(resource_size_t phys_addr, + unsigned long size) +{ + return early_memremap_prot(phys_addr, size, __PAGE_KERNEL_ENC); +} + +/* + * Remap memory with encryption and write-protected - cannot be called + * before pat_init() is called + */ +void __init *early_memremap_encrypted_wp(resource_size_t phys_addr, + unsigned long size) +{ + /* Be sure the write-protect PAT entry is set for write-protect */ + if (__pte2cachemode_tbl[_PAGE_CACHE_MODE_WP] != _PAGE_CACHE_MODE_WP) + return NULL; + + return early_memremap_prot(phys_addr, size, __PAGE_KERNEL_ENC_WP); +} + +/* Remap memory without encryption */ +void __init *early_memremap_decrypted(resource_size_t phys_addr, + unsigned long size) +{ + return early_memremap_prot(phys_addr, size, __PAGE_KERNEL_NOENC); +} + +/* + * Remap memory without encryption and write-protected - cannot be called + * before pat_init() is called + */ +void __init *early_memremap_decrypted_wp(resource_size_t phys_addr, + unsigned long size) +{ + /* Be sure the write-protect PAT entry is set for write-protect */ + if (__pte2cachemode_tbl[_PAGE_CACHE_MODE_WP] != _PAGE_CACHE_MODE_WP) + return NULL; - iounmap((void __iomem *)((unsigned long)addr & PAGE_MASK)); + return early_memremap_prot(phys_addr, size, __PAGE_KERNEL_NOENC_WP); } +#endif /* CONFIG_ARCH_USE_MEMREMAP_PROT */ static pte_t bm_pte[PAGE_SIZE/sizeof(pte_t)] __page_aligned_bss; diff --git a/arch/x86/mm/kasan_init_64.c b/arch/x86/mm/kasan_init_64.c index 02c9d7553409..bc84b73684b7 100644 --- a/arch/x86/mm/kasan_init_64.c +++ b/arch/x86/mm/kasan_init_64.c @@ -11,8 +11,8 @@ #include <asm/e820/types.h> #include <asm/tlbflush.h> #include <asm/sections.h> +#include <asm/pgtable.h> -extern pgd_t early_top_pgt[PTRS_PER_PGD]; extern struct range pfn_mapped[E820_MAX_ENTRIES]; static int __init map_range(struct range *range) @@ -87,7 +87,7 @@ static struct notifier_block kasan_die_notifier = { void __init kasan_early_init(void) { int i; - pteval_t pte_val = __pa_nodebug(kasan_zero_page) | __PAGE_KERNEL; + pteval_t pte_val = __pa_nodebug(kasan_zero_page) | __PAGE_KERNEL | _PAGE_ENC; pmdval_t pmd_val = __pa_nodebug(kasan_zero_pte) | _KERNPG_TABLE; pudval_t pud_val = __pa_nodebug(kasan_zero_pmd) | _KERNPG_TABLE; p4dval_t p4d_val = __pa_nodebug(kasan_zero_pud) | _KERNPG_TABLE; @@ -153,7 +153,7 @@ void __init kasan_init(void) */ memset(kasan_zero_page, 0, PAGE_SIZE); for (i = 0; i < PTRS_PER_PTE; i++) { - pte_t pte = __pte(__pa(kasan_zero_page) | __PAGE_KERNEL_RO); + pte_t pte = __pte(__pa(kasan_zero_page) | __PAGE_KERNEL_RO | _PAGE_ENC); set_pte(&kasan_zero_pte[i], pte); } /* Flush TLBs again to be sure that write protection applied. */ diff --git a/arch/x86/mm/mem_encrypt.c b/arch/x86/mm/mem_encrypt.c new file mode 100644 index 000000000000..0fbd09269757 --- /dev/null +++ b/arch/x86/mm/mem_encrypt.c @@ -0,0 +1,593 @@ +/* + * AMD Memory Encryption Support + * + * Copyright (C) 2016 Advanced Micro Devices, Inc. + * + * Author: Tom Lendacky <thomas.lendacky@amd.com> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/linkage.h> +#include <linux/init.h> +#include <linux/mm.h> +#include <linux/dma-mapping.h> +#include <linux/swiotlb.h> +#include <linux/mem_encrypt.h> + +#include <asm/tlbflush.h> +#include <asm/fixmap.h> +#include <asm/setup.h> +#include <asm/bootparam.h> +#include <asm/set_memory.h> +#include <asm/cacheflush.h> +#include <asm/sections.h> +#include <asm/processor-flags.h> +#include <asm/msr.h> +#include <asm/cmdline.h> + +static char sme_cmdline_arg[] __initdata = "mem_encrypt"; +static char sme_cmdline_on[] __initdata = "on"; +static char sme_cmdline_off[] __initdata = "off"; + +/* + * Since SME related variables are set early in the boot process they must + * reside in the .data section so as not to be zeroed out when the .bss + * section is later cleared. + */ +unsigned long sme_me_mask __section(.data) = 0; +EXPORT_SYMBOL_GPL(sme_me_mask); + +/* Buffer used for early in-place encryption by BSP, no locking needed */ +static char sme_early_buffer[PAGE_SIZE] __aligned(PAGE_SIZE); + +/* + * This routine does not change the underlying encryption setting of the + * page(s) that map this memory. It assumes that eventually the memory is + * meant to be accessed as either encrypted or decrypted but the contents + * are currently not in the desired state. + * + * This routine follows the steps outlined in the AMD64 Architecture + * Programmer's Manual Volume 2, Section 7.10.8 Encrypt-in-Place. + */ +static void __init __sme_early_enc_dec(resource_size_t paddr, + unsigned long size, bool enc) +{ + void *src, *dst; + size_t len; + + if (!sme_me_mask) + return; + + local_flush_tlb(); + wbinvd(); + + /* + * There are limited number of early mapping slots, so map (at most) + * one page at time. + */ + while (size) { + len = min_t(size_t, sizeof(sme_early_buffer), size); + + /* + * Create mappings for the current and desired format of + * the memory. Use a write-protected mapping for the source. + */ + src = enc ? early_memremap_decrypted_wp(paddr, len) : + early_memremap_encrypted_wp(paddr, len); + + dst = enc ? early_memremap_encrypted(paddr, len) : + early_memremap_decrypted(paddr, len); + + /* + * If a mapping can't be obtained to perform the operation, + * then eventual access of that area in the desired mode + * will cause a crash. + */ + BUG_ON(!src || !dst); + + /* + * Use a temporary buffer, of cache-line multiple size, to + * avoid data corruption as documented in the APM. + */ + memcpy(sme_early_buffer, src, len); + memcpy(dst, sme_early_buffer, len); + + early_memunmap(dst, len); + early_memunmap(src, len); + + paddr += len; + size -= len; + } +} + +void __init sme_early_encrypt(resource_size_t paddr, unsigned long size) +{ + __sme_early_enc_dec(paddr, size, true); +} + +void __init sme_early_decrypt(resource_size_t paddr, unsigned long size) +{ + __sme_early_enc_dec(paddr, size, false); +} + +static void __init __sme_early_map_unmap_mem(void *vaddr, unsigned long size, + bool map) +{ + unsigned long paddr = (unsigned long)vaddr - __PAGE_OFFSET; + pmdval_t pmd_flags, pmd; + + /* Use early_pmd_flags but remove the encryption mask */ + pmd_flags = __sme_clr(early_pmd_flags); + + do { + pmd = map ? (paddr & PMD_MASK) + pmd_flags : 0; + __early_make_pgtable((unsigned long)vaddr, pmd); + + vaddr += PMD_SIZE; + paddr += PMD_SIZE; + size = (size <= PMD_SIZE) ? 0 : size - PMD_SIZE; + } while (size); + + __native_flush_tlb(); +} + +void __init sme_unmap_bootdata(char *real_mode_data) +{ + struct boot_params *boot_data; + unsigned long cmdline_paddr; + + if (!sme_active()) + return; + + /* Get the command line address before unmapping the real_mode_data */ + boot_data = (struct boot_params *)real_mode_data; + cmdline_paddr = boot_data->hdr.cmd_line_ptr | ((u64)boot_data->ext_cmd_line_ptr << 32); + + __sme_early_map_unmap_mem(real_mode_data, sizeof(boot_params), false); + + if (!cmdline_paddr) + return; + + __sme_early_map_unmap_mem(__va(cmdline_paddr), COMMAND_LINE_SIZE, false); +} + +void __init sme_map_bootdata(char *real_mode_data) +{ + struct boot_params *boot_data; + unsigned long cmdline_paddr; + + if (!sme_active()) + return; + + __sme_early_map_unmap_mem(real_mode_data, sizeof(boot_params), true); + + /* Get the command line address after mapping the real_mode_data */ + boot_data = (struct boot_params *)real_mode_data; + cmdline_paddr = boot_data->hdr.cmd_line_ptr | ((u64)boot_data->ext_cmd_line_ptr << 32); + + if (!cmdline_paddr) + return; + + __sme_early_map_unmap_mem(__va(cmdline_paddr), COMMAND_LINE_SIZE, true); +} + +void __init sme_early_init(void) +{ + unsigned int i; + + if (!sme_me_mask) + return; + + early_pmd_flags = __sme_set(early_pmd_flags); + + __supported_pte_mask = __sme_set(__supported_pte_mask); + + /* Update the protection map with memory encryption mask */ + for (i = 0; i < ARRAY_SIZE(protection_map); i++) + protection_map[i] = pgprot_encrypted(protection_map[i]); +} + +/* Architecture __weak replacement functions */ +void __init mem_encrypt_init(void) +{ + if (!sme_me_mask) + return; + + /* Call into SWIOTLB to update the SWIOTLB DMA buffers */ + swiotlb_update_mem_attributes(); + + pr_info("AMD Secure Memory Encryption (SME) active\n"); +} + +void swiotlb_set_mem_attributes(void *vaddr, unsigned long size) +{ + WARN(PAGE_ALIGN(size) != size, + "size is not page-aligned (%#lx)\n", size); + + /* Make the SWIOTLB buffer area decrypted */ + set_memory_decrypted((unsigned long)vaddr, size >> PAGE_SHIFT); +} + +static void __init sme_clear_pgd(pgd_t *pgd_base, unsigned long start, + unsigned long end) +{ + unsigned long pgd_start, pgd_end, pgd_size; + pgd_t *pgd_p; + + pgd_start = start & PGDIR_MASK; + pgd_end = end & PGDIR_MASK; + + pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1); + pgd_size *= sizeof(pgd_t); + + pgd_p = pgd_base + pgd_index(start); + + memset(pgd_p, 0, pgd_size); +} + +#define PGD_FLAGS _KERNPG_TABLE_NOENC +#define P4D_FLAGS _KERNPG_TABLE_NOENC +#define PUD_FLAGS _KERNPG_TABLE_NOENC +#define PMD_FLAGS (__PAGE_KERNEL_LARGE_EXEC & ~_PAGE_GLOBAL) + +static void __init *sme_populate_pgd(pgd_t *pgd_base, void *pgtable_area, + unsigned long vaddr, pmdval_t pmd_val) +{ + pgd_t *pgd_p; + p4d_t *p4d_p; + pud_t *pud_p; + pmd_t *pmd_p; + + pgd_p = pgd_base + pgd_index(vaddr); + if (native_pgd_val(*pgd_p)) { + if (IS_ENABLED(CONFIG_X86_5LEVEL)) + p4d_p = (p4d_t *)(native_pgd_val(*pgd_p) & ~PTE_FLAGS_MASK); + else + pud_p = (pud_t *)(native_pgd_val(*pgd_p) & ~PTE_FLAGS_MASK); + } else { + pgd_t pgd; + + if (IS_ENABLED(CONFIG_X86_5LEVEL)) { + p4d_p = pgtable_area; + memset(p4d_p, 0, sizeof(*p4d_p) * PTRS_PER_P4D); + pgtable_area += sizeof(*p4d_p) * PTRS_PER_P4D; + + pgd = native_make_pgd((pgdval_t)p4d_p + PGD_FLAGS); + } else { + pud_p = pgtable_area; + memset(pud_p, 0, sizeof(*pud_p) * PTRS_PER_PUD); + pgtable_area += sizeof(*pud_p) * PTRS_PER_PUD; + + pgd = native_make_pgd((pgdval_t)pud_p + PGD_FLAGS); + } + native_set_pgd(pgd_p, pgd); + } + + if (IS_ENABLED(CONFIG_X86_5LEVEL)) { + p4d_p += p4d_index(vaddr); + if (native_p4d_val(*p4d_p)) { + pud_p = (pud_t *)(native_p4d_val(*p4d_p) & ~PTE_FLAGS_MASK); + } else { + p4d_t p4d; + + pud_p = pgtable_area; + memset(pud_p, 0, sizeof(*pud_p) * PTRS_PER_PUD); + pgtable_area += sizeof(*pud_p) * PTRS_PER_PUD; + + p4d = native_make_p4d((pudval_t)pud_p + P4D_FLAGS); + native_set_p4d(p4d_p, p4d); + } + } + + pud_p += pud_index(vaddr); + if (native_pud_val(*pud_p)) { + if (native_pud_val(*pud_p) & _PAGE_PSE) + goto out; + + pmd_p = (pmd_t *)(native_pud_val(*pud_p) & ~PTE_FLAGS_MASK); + } else { + pud_t pud; + + pmd_p = pgtable_area; + memset(pmd_p, 0, sizeof(*pmd_p) * PTRS_PER_PMD); + pgtable_area += sizeof(*pmd_p) * PTRS_PER_PMD; + + pud = native_make_pud((pmdval_t)pmd_p + PUD_FLAGS); + native_set_pud(pud_p, pud); + } + + pmd_p += pmd_index(vaddr); + if (!native_pmd_val(*pmd_p) || !(native_pmd_val(*pmd_p) & _PAGE_PSE)) + native_set_pmd(pmd_p, native_make_pmd(pmd_val)); + +out: + return pgtable_area; +} + +static unsigned long __init sme_pgtable_calc(unsigned long len) +{ + unsigned long p4d_size, pud_size, pmd_size; + unsigned long total; + + /* + * Perform a relatively simplistic calculation of the pagetable + * entries that are needed. That mappings will be covered by 2MB + * PMD entries so we can conservatively calculate the required + * number of P4D, PUD and PMD structures needed to perform the + * mappings. Incrementing the count for each covers the case where + * the addresses cross entries. + */ + if (IS_ENABLED(CONFIG_X86_5LEVEL)) { + p4d_size = (ALIGN(len, PGDIR_SIZE) / PGDIR_SIZE) + 1; + p4d_size *= sizeof(p4d_t) * PTRS_PER_P4D; + pud_size = (ALIGN(len, P4D_SIZE) / P4D_SIZE) + 1; + pud_size *= sizeof(pud_t) * PTRS_PER_PUD; + } else { + p4d_size = 0; + pud_size = (ALIGN(len, PGDIR_SIZE) / PGDIR_SIZE) + 1; + pud_size *= sizeof(pud_t) * PTRS_PER_PUD; + } + pmd_size = (ALIGN(len, PUD_SIZE) / PUD_SIZE) + 1; + pmd_size *= sizeof(pmd_t) * PTRS_PER_PMD; + + total = p4d_size + pud_size + pmd_size; + + /* + * Now calculate the added pagetable structures needed to populate + * the new pagetables. + */ + if (IS_ENABLED(CONFIG_X86_5LEVEL)) { + p4d_size = ALIGN(total, PGDIR_SIZE) / PGDIR_SIZE; + p4d_size *= sizeof(p4d_t) * PTRS_PER_P4D; + pud_size = ALIGN(total, P4D_SIZE) / P4D_SIZE; + pud_size *= sizeof(pud_t) * PTRS_PER_PUD; + } else { + p4d_size = 0; + pud_size = ALIGN(total, PGDIR_SIZE) / PGDIR_SIZE; + pud_size *= sizeof(pud_t) * PTRS_PER_PUD; + } + pmd_size = ALIGN(total, PUD_SIZE) / PUD_SIZE; + pmd_size *= sizeof(pmd_t) * PTRS_PER_PMD; + + total += p4d_size + pud_size + pmd_size; + + return total; +} + +void __init sme_encrypt_kernel(void) +{ + unsigned long workarea_start, workarea_end, workarea_len; + unsigned long execute_start, execute_end, execute_len; + unsigned long kernel_start, kernel_end, kernel_len; + unsigned long pgtable_area_len; + unsigned long paddr, pmd_flags; + unsigned long decrypted_base; + void *pgtable_area; + pgd_t *pgd; + + if (!sme_active()) + return; + + /* + * Prepare for encrypting the kernel by building new pagetables with + * the necessary attributes needed to encrypt the kernel in place. + * + * One range of virtual addresses will map the memory occupied + * by the kernel as encrypted. + * + * Another range of virtual addresses will map the memory occupied + * by the kernel as decrypted and write-protected. + * + * The use of write-protect attribute will prevent any of the + * memory from being cached. + */ + + /* Physical addresses gives us the identity mapped virtual addresses */ + kernel_start = __pa_symbol(_text); + kernel_end = ALIGN(__pa_symbol(_end), PMD_PAGE_SIZE); + kernel_len = kernel_end - kernel_start; + + /* Set the encryption workarea to be immediately after the kernel */ + workarea_start = kernel_end; + + /* + * Calculate required number of workarea bytes needed: + * executable encryption area size: + * stack page (PAGE_SIZE) + * encryption routine page (PAGE_SIZE) + * intermediate copy buffer (PMD_PAGE_SIZE) + * pagetable structures for the encryption of the kernel + * pagetable structures for workarea (in case not currently mapped) + */ + execute_start = workarea_start; + execute_end = execute_start + (PAGE_SIZE * 2) + PMD_PAGE_SIZE; + execute_len = execute_end - execute_start; + + /* + * One PGD for both encrypted and decrypted mappings and a set of + * PUDs and PMDs for each of the encrypted and decrypted mappings. + */ + pgtable_area_len = sizeof(pgd_t) * PTRS_PER_PGD; + pgtable_area_len += sme_pgtable_calc(execute_end - kernel_start) * 2; + + /* PUDs and PMDs needed in the current pagetables for the workarea */ + pgtable_area_len += sme_pgtable_calc(execute_len + pgtable_area_len); + + /* + * The total workarea includes the executable encryption area and + * the pagetable area. + */ + workarea_len = execute_len + pgtable_area_len; + workarea_end = workarea_start + workarea_len; + + /* + * Set the address to the start of where newly created pagetable + * structures (PGDs, PUDs and PMDs) will be allocated. New pagetable + * structures are created when the workarea is added to the current + * pagetables and when the new encrypted and decrypted kernel + * mappings are populated. + */ + pgtable_area = (void *)execute_end; + + /* + * Make sure the current pagetable structure has entries for + * addressing the workarea. + */ + pgd = (pgd_t *)native_read_cr3_pa(); + paddr = workarea_start; + while (paddr < workarea_end) { + pgtable_area = sme_populate_pgd(pgd, pgtable_area, + paddr, + paddr + PMD_FLAGS); + + paddr += PMD_PAGE_SIZE; + } + + /* Flush the TLB - no globals so cr3 is enough */ + native_write_cr3(__native_read_cr3()); + + /* + * A new pagetable structure is being built to allow for the kernel + * to be encrypted. It starts with an empty PGD that will then be + * populated with new PUDs and PMDs as the encrypted and decrypted + * kernel mappings are created. + */ + pgd = pgtable_area; + memset(pgd, 0, sizeof(*pgd) * PTRS_PER_PGD); + pgtable_area += sizeof(*pgd) * PTRS_PER_PGD; + + /* Add encrypted kernel (identity) mappings */ + pmd_flags = PMD_FLAGS | _PAGE_ENC; + paddr = kernel_start; + while (paddr < kernel_end) { + pgtable_area = sme_populate_pgd(pgd, pgtable_area, + paddr, + paddr + pmd_flags); + + paddr += PMD_PAGE_SIZE; + } + + /* + * A different PGD index/entry must be used to get different + * pagetable entries for the decrypted mapping. Choose the next + * PGD index and convert it to a virtual address to be used as + * the base of the mapping. + */ + decrypted_base = (pgd_index(workarea_end) + 1) & (PTRS_PER_PGD - 1); + decrypted_base <<= PGDIR_SHIFT; + + /* Add decrypted, write-protected kernel (non-identity) mappings */ + pmd_flags = (PMD_FLAGS & ~_PAGE_CACHE_MASK) | (_PAGE_PAT | _PAGE_PWT); + paddr = kernel_start; + while (paddr < kernel_end) { + pgtable_area = sme_populate_pgd(pgd, pgtable_area, + paddr + decrypted_base, + paddr + pmd_flags); + + paddr += PMD_PAGE_SIZE; + } + + /* Add decrypted workarea mappings to both kernel mappings */ + paddr = workarea_start; + while (paddr < workarea_end) { + pgtable_area = sme_populate_pgd(pgd, pgtable_area, + paddr, + paddr + PMD_FLAGS); + + pgtable_area = sme_populate_pgd(pgd, pgtable_area, + paddr + decrypted_base, + paddr + PMD_FLAGS); + + paddr += PMD_PAGE_SIZE; + } + + /* Perform the encryption */ + sme_encrypt_execute(kernel_start, kernel_start + decrypted_base, + kernel_len, workarea_start, (unsigned long)pgd); + + /* + * At this point we are running encrypted. Remove the mappings for + * the decrypted areas - all that is needed for this is to remove + * the PGD entry/entries. + */ + sme_clear_pgd(pgd, kernel_start + decrypted_base, + kernel_end + decrypted_base); + + sme_clear_pgd(pgd, workarea_start + decrypted_base, + workarea_end + decrypted_base); + + /* Flush the TLB - no globals so cr3 is enough */ + native_write_cr3(__native_read_cr3()); +} + +void __init __nostackprotector sme_enable(struct boot_params *bp) +{ + const char *cmdline_ptr, *cmdline_arg, *cmdline_on, *cmdline_off; + unsigned int eax, ebx, ecx, edx; + bool active_by_default; + unsigned long me_mask; + char buffer[16]; + u64 msr; + + /* Check for the SME support leaf */ + eax = 0x80000000; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + if (eax < 0x8000001f) + return; + + /* + * Check for the SME feature: + * CPUID Fn8000_001F[EAX] - Bit 0 + * Secure Memory Encryption support + * CPUID Fn8000_001F[EBX] - Bits 5:0 + * Pagetable bit position used to indicate encryption + */ + eax = 0x8000001f; + ecx = 0; + native_cpuid(&eax, &ebx, &ecx, &edx); + if (!(eax & 1)) + return; + + me_mask = 1UL << (ebx & 0x3f); + + /* Check if SME is enabled */ + msr = __rdmsr(MSR_K8_SYSCFG); + if (!(msr & MSR_K8_SYSCFG_MEM_ENCRYPT)) + return; + + /* + * Fixups have not been applied to phys_base yet and we're running + * identity mapped, so we must obtain the address to the SME command + * line argument data using rip-relative addressing. + */ + asm ("lea sme_cmdline_arg(%%rip), %0" + : "=r" (cmdline_arg) + : "p" (sme_cmdline_arg)); + asm ("lea sme_cmdline_on(%%rip), %0" + : "=r" (cmdline_on) + : "p" (sme_cmdline_on)); + asm ("lea sme_cmdline_off(%%rip), %0" + : "=r" (cmdline_off) + : "p" (sme_cmdline_off)); + + if (IS_ENABLED(CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT)) + active_by_default = true; + else + active_by_default = false; + + cmdline_ptr = (const char *)((u64)bp->hdr.cmd_line_ptr | + ((u64)bp->ext_cmd_line_ptr << 32)); + + cmdline_find_option(cmdline_ptr, cmdline_arg, buffer, sizeof(buffer)); + + if (!strncmp(buffer, cmdline_on, sizeof(buffer))) + sme_me_mask = me_mask; + else if (!strncmp(buffer, cmdline_off, sizeof(buffer))) + sme_me_mask = 0; + else + sme_me_mask = active_by_default ? me_mask : 0; +} diff --git a/arch/x86/mm/mem_encrypt_boot.S b/arch/x86/mm/mem_encrypt_boot.S new file mode 100644 index 000000000000..730e6d541df1 --- /dev/null +++ b/arch/x86/mm/mem_encrypt_boot.S @@ -0,0 +1,149 @@ +/* + * AMD Memory Encryption Support + * + * Copyright (C) 2016 Advanced Micro Devices, Inc. + * + * Author: Tom Lendacky <thomas.lendacky@amd.com> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/linkage.h> +#include <asm/pgtable.h> +#include <asm/page.h> +#include <asm/processor-flags.h> +#include <asm/msr-index.h> + + .text + .code64 +ENTRY(sme_encrypt_execute) + + /* + * Entry parameters: + * RDI - virtual address for the encrypted kernel mapping + * RSI - virtual address for the decrypted kernel mapping + * RDX - length of kernel + * RCX - virtual address of the encryption workarea, including: + * - stack page (PAGE_SIZE) + * - encryption routine page (PAGE_SIZE) + * - intermediate copy buffer (PMD_PAGE_SIZE) + * R8 - physcial address of the pagetables to use for encryption + */ + + push %rbp + movq %rsp, %rbp /* RBP now has original stack pointer */ + + /* Set up a one page stack in the non-encrypted memory area */ + movq %rcx, %rax /* Workarea stack page */ + leaq PAGE_SIZE(%rax), %rsp /* Set new stack pointer */ + addq $PAGE_SIZE, %rax /* Workarea encryption routine */ + + push %r12 + movq %rdi, %r10 /* Encrypted kernel */ + movq %rsi, %r11 /* Decrypted kernel */ + movq %rdx, %r12 /* Kernel length */ + + /* Copy encryption routine into the workarea */ + movq %rax, %rdi /* Workarea encryption routine */ + leaq __enc_copy(%rip), %rsi /* Encryption routine */ + movq $(.L__enc_copy_end - __enc_copy), %rcx /* Encryption routine length */ + rep movsb + + /* Setup registers for call */ + movq %r10, %rdi /* Encrypted kernel */ + movq %r11, %rsi /* Decrypted kernel */ + movq %r8, %rdx /* Pagetables used for encryption */ + movq %r12, %rcx /* Kernel length */ + movq %rax, %r8 /* Workarea encryption routine */ + addq $PAGE_SIZE, %r8 /* Workarea intermediate copy buffer */ + + call *%rax /* Call the encryption routine */ + + pop %r12 + + movq %rbp, %rsp /* Restore original stack pointer */ + pop %rbp + + ret +ENDPROC(sme_encrypt_execute) + +ENTRY(__enc_copy) +/* + * Routine used to encrypt kernel. + * This routine must be run outside of the kernel proper since + * the kernel will be encrypted during the process. So this + * routine is defined here and then copied to an area outside + * of the kernel where it will remain and run decrypted + * during execution. + * + * On entry the registers must be: + * RDI - virtual address for the encrypted kernel mapping + * RSI - virtual address for the decrypted kernel mapping + * RDX - address of the pagetables to use for encryption + * RCX - length of kernel + * R8 - intermediate copy buffer + * + * RAX - points to this routine + * + * The kernel will be encrypted by copying from the non-encrypted + * kernel space to an intermediate buffer and then copying from the + * intermediate buffer back to the encrypted kernel space. The physical + * addresses of the two kernel space mappings are the same which + * results in the kernel being encrypted "in place". + */ + /* Enable the new page tables */ + mov %rdx, %cr3 + + /* Flush any global TLBs */ + mov %cr4, %rdx + andq $~X86_CR4_PGE, %rdx + mov %rdx, %cr4 + orq $X86_CR4_PGE, %rdx + mov %rdx, %cr4 + + /* Set the PAT register PA5 entry to write-protect */ + push %rcx + movl $MSR_IA32_CR_PAT, %ecx + rdmsr + push %rdx /* Save original PAT value */ + andl $0xffff00ff, %edx /* Clear PA5 */ + orl $0x00000500, %edx /* Set PA5 to WP */ + wrmsr + pop %rdx /* RDX contains original PAT value */ + pop %rcx + + movq %rcx, %r9 /* Save kernel length */ + movq %rdi, %r10 /* Save encrypted kernel address */ + movq %rsi, %r11 /* Save decrypted kernel address */ + + wbinvd /* Invalidate any cache entries */ + + /* Copy/encrypt 2MB at a time */ +1: + movq %r11, %rsi /* Source - decrypted kernel */ + movq %r8, %rdi /* Dest - intermediate copy buffer */ + movq $PMD_PAGE_SIZE, %rcx /* 2MB length */ + rep movsb + + movq %r8, %rsi /* Source - intermediate copy buffer */ + movq %r10, %rdi /* Dest - encrypted kernel */ + movq $PMD_PAGE_SIZE, %rcx /* 2MB length */ + rep movsb + + addq $PMD_PAGE_SIZE, %r11 + addq $PMD_PAGE_SIZE, %r10 + subq $PMD_PAGE_SIZE, %r9 /* Kernel length decrement */ + jnz 1b /* Kernel length not zero? */ + + /* Restore PAT register */ + push %rdx /* Save original PAT value */ + movl $MSR_IA32_CR_PAT, %ecx + rdmsr + pop %rdx /* Restore original PAT value */ + wrmsr + + ret +.L__enc_copy_end: +ENDPROC(__enc_copy) diff --git a/arch/x86/mm/mmap.c b/arch/x86/mm/mmap.c index a88cfbfbd078..a99679826846 100644 --- a/arch/x86/mm/mmap.c +++ b/arch/x86/mm/mmap.c @@ -37,21 +37,21 @@ struct va_alignment __read_mostly va_align = { .flags = -1, }; -unsigned long tasksize_32bit(void) +unsigned long task_size_32bit(void) { return IA32_PAGE_OFFSET; } -unsigned long tasksize_64bit(void) +unsigned long task_size_64bit(int full_addr_space) { - return TASK_SIZE_MAX; + return full_addr_space ? TASK_SIZE_MAX : DEFAULT_MAP_WINDOW; } static unsigned long stack_maxrandom_size(unsigned long task_size) { unsigned long max = 0; if (current->flags & PF_RANDOMIZE) { - max = (-1UL) & __STACK_RND_MASK(task_size == tasksize_32bit()); + max = (-1UL) & __STACK_RND_MASK(task_size == task_size_32bit()); max <<= PAGE_SHIFT; } @@ -141,7 +141,7 @@ void arch_pick_mmap_layout(struct mm_struct *mm) mm->get_unmapped_area = arch_get_unmapped_area_topdown; arch_pick_mmap_base(&mm->mmap_base, &mm->mmap_legacy_base, - arch_rnd(mmap64_rnd_bits), tasksize_64bit()); + arch_rnd(mmap64_rnd_bits), task_size_64bit(0)); #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES /* @@ -151,7 +151,7 @@ void arch_pick_mmap_layout(struct mm_struct *mm) * mmap_base, the compat syscall uses mmap_compat_base. */ arch_pick_mmap_base(&mm->mmap_compat_base, &mm->mmap_compat_legacy_base, - arch_rnd(mmap32_rnd_bits), tasksize_32bit()); + arch_rnd(mmap32_rnd_bits), task_size_32bit()); #endif } diff --git a/arch/x86/mm/mpx.c b/arch/x86/mm/mpx.c index 1c34b767c84c..9ceaa955d2ba 100644 --- a/arch/x86/mm/mpx.c +++ b/arch/x86/mm/mpx.c @@ -355,10 +355,19 @@ int mpx_enable_management(void) */ bd_base = mpx_get_bounds_dir(); down_write(&mm->mmap_sem); + + /* MPX doesn't support addresses above 47 bits yet. */ + if (find_vma(mm, DEFAULT_MAP_WINDOW)) { + pr_warn_once("%s (%d): MPX cannot handle addresses " + "above 47-bits. Disabling.", + current->comm, current->pid); + ret = -ENXIO; + goto out; + } mm->context.bd_addr = bd_base; if (mm->context.bd_addr == MPX_INVALID_BOUNDS_DIR) ret = -ENXIO; - +out: up_write(&mm->mmap_sem); return ret; } @@ -1030,3 +1039,25 @@ void mpx_notify_unmap(struct mm_struct *mm, struct vm_area_struct *vma, if (ret) force_sig(SIGSEGV, current); } + +/* MPX cannot handle addresses above 47 bits yet. */ +unsigned long mpx_unmapped_area_check(unsigned long addr, unsigned long len, + unsigned long flags) +{ + if (!kernel_managing_mpx_tables(current->mm)) + return addr; + if (addr + len <= DEFAULT_MAP_WINDOW) + return addr; + if (flags & MAP_FIXED) + return -ENOMEM; + + /* + * Requested len is larger than the whole area we're allowed to map in. + * Resetting hinting address wouldn't do much good -- fail early. + */ + if (len > DEFAULT_MAP_WINDOW) + return -ENOMEM; + + /* Look for unmap area within DEFAULT_MAP_WINDOW */ + return 0; +} diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c index 757b0bcdf712..dfb7d657cf43 100644 --- a/arch/x86/mm/pageattr.c +++ b/arch/x86/mm/pageattr.c @@ -1775,6 +1775,70 @@ int set_memory_4k(unsigned long addr, int numpages) __pgprot(0), 1, 0, NULL); } +static int __set_memory_enc_dec(unsigned long addr, int numpages, bool enc) +{ + struct cpa_data cpa; + unsigned long start; + int ret; + + /* Nothing to do if the SME is not active */ + if (!sme_active()) + return 0; + + /* Should not be working on unaligned addresses */ + if (WARN_ONCE(addr & ~PAGE_MASK, "misaligned address: %#lx\n", addr)) + addr &= PAGE_MASK; + + start = addr; + + memset(&cpa, 0, sizeof(cpa)); + cpa.vaddr = &addr; + cpa.numpages = numpages; + cpa.mask_set = enc ? __pgprot(_PAGE_ENC) : __pgprot(0); + cpa.mask_clr = enc ? __pgprot(0) : __pgprot(_PAGE_ENC); + cpa.pgd = init_mm.pgd; + + /* Must avoid aliasing mappings in the highmem code */ + kmap_flush_unused(); + vm_unmap_aliases(); + + /* + * Before changing the encryption attribute, we need to flush caches. + */ + if (static_cpu_has(X86_FEATURE_CLFLUSH)) + cpa_flush_range(start, numpages, 1); + else + cpa_flush_all(1); + + ret = __change_page_attr_set_clr(&cpa, 1); + + /* + * After changing the encryption attribute, we need to flush TLBs + * again in case any speculative TLB caching occurred (but no need + * to flush caches again). We could just use cpa_flush_all(), but + * in case TLB flushing gets optimized in the cpa_flush_range() + * path use the same logic as above. + */ + if (static_cpu_has(X86_FEATURE_CLFLUSH)) + cpa_flush_range(start, numpages, 0); + else + cpa_flush_all(0); + + return ret; +} + +int set_memory_encrypted(unsigned long addr, int numpages) +{ + return __set_memory_enc_dec(addr, numpages, true); +} +EXPORT_SYMBOL_GPL(set_memory_encrypted); + +int set_memory_decrypted(unsigned long addr, int numpages) +{ + return __set_memory_enc_dec(addr, numpages, false); +} +EXPORT_SYMBOL_GPL(set_memory_decrypted); + int set_pages_uc(struct page *page, int numpages) { unsigned long addr = (unsigned long)page_address(page); @@ -2020,6 +2084,9 @@ int kernel_map_pages_in_pgd(pgd_t *pgd, u64 pfn, unsigned long address, if (!(page_flags & _PAGE_RW)) cpa.mask_clr = __pgprot(_PAGE_RW); + if (!(page_flags & _PAGE_ENC)) + cpa.mask_clr = pgprot_encrypted(cpa.mask_clr); + cpa.mask_set = __pgprot(_PAGE_PRESENT | page_flags); retval = __change_page_attr_set_clr(&cpa, 0); diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 45979502f64b..fe7d57a8fb60 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -293,7 +293,7 @@ void init_cache_modes(void) * pat_init - Initialize PAT MSR and PAT table * * This function initializes PAT MSR and PAT table with an OS-defined value - * to enable additional cache attributes, WC and WT. + * to enable additional cache attributes, WC, WT and WP. * * This function must be called on all CPUs using the specific sequence of * operations defined in Intel SDM. mtrr_rendezvous_handler() provides this @@ -352,7 +352,7 @@ void pat_init(void) * 010 2 UC-: _PAGE_CACHE_MODE_UC_MINUS * 011 3 UC : _PAGE_CACHE_MODE_UC * 100 4 WB : Reserved - * 101 5 WC : Reserved + * 101 5 WP : _PAGE_CACHE_MODE_WP * 110 6 UC-: Reserved * 111 7 WT : _PAGE_CACHE_MODE_WT * @@ -360,7 +360,7 @@ void pat_init(void) * corresponding types in the presence of PAT errata. */ pat = PAT(0, WB) | PAT(1, WC) | PAT(2, UC_MINUS) | PAT(3, UC) | - PAT(4, WB) | PAT(5, WC) | PAT(6, UC_MINUS) | PAT(7, WT); + PAT(4, WB) | PAT(5, WP) | PAT(6, UC_MINUS) | PAT(7, WT); } if (!boot_cpu_done) { @@ -744,6 +744,9 @@ EXPORT_SYMBOL(arch_io_free_memtype_wc); pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, unsigned long size, pgprot_t vma_prot) { + if (!phys_mem_access_encrypted(pfn << PAGE_SHIFT, size)) + vma_prot = pgprot_decrypted(vma_prot); + return vma_prot; } diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c index 508a708eb9a6..218834a3e9ad 100644 --- a/arch/x86/mm/pgtable.c +++ b/arch/x86/mm/pgtable.c @@ -56,7 +56,7 @@ void ___pte_free_tlb(struct mmu_gather *tlb, struct page *pte) { pgtable_page_dtor(pte); paravirt_release_pte(page_to_pfn(pte)); - tlb_remove_page(tlb, pte); + tlb_remove_table(tlb, pte); } #if CONFIG_PGTABLE_LEVELS > 2 @@ -72,21 +72,21 @@ void ___pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd) tlb->need_flush_all = 1; #endif pgtable_pmd_page_dtor(page); - tlb_remove_page(tlb, page); + tlb_remove_table(tlb, page); } #if CONFIG_PGTABLE_LEVELS > 3 void ___pud_free_tlb(struct mmu_gather *tlb, pud_t *pud) { paravirt_release_pud(__pa(pud) >> PAGE_SHIFT); - tlb_remove_page(tlb, virt_to_page(pud)); + tlb_remove_table(tlb, virt_to_page(pud)); } #if CONFIG_PGTABLE_LEVELS > 4 void ___p4d_free_tlb(struct mmu_gather *tlb, p4d_t *p4d) { paravirt_release_p4d(__pa(p4d) >> PAGE_SHIFT); - tlb_remove_page(tlb, virt_to_page(p4d)); + tlb_remove_table(tlb, virt_to_page(p4d)); } #endif /* CONFIG_PGTABLE_LEVELS > 4 */ #endif /* CONFIG_PGTABLE_LEVELS > 3 */ diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c index 014d07a80053..ce104b962a17 100644 --- a/arch/x86/mm/tlb.c +++ b/arch/x86/mm/tlb.c @@ -28,6 +28,42 @@ * Implement flush IPI by CALL_FUNCTION_VECTOR, Alex Shi */ +atomic64_t last_mm_ctx_id = ATOMIC64_INIT(1); + +static void choose_new_asid(struct mm_struct *next, u64 next_tlb_gen, + u16 *new_asid, bool *need_flush) +{ + u16 asid; + + if (!static_cpu_has(X86_FEATURE_PCID)) { + *new_asid = 0; + *need_flush = true; + return; + } + + for (asid = 0; asid < TLB_NR_DYN_ASIDS; asid++) { + if (this_cpu_read(cpu_tlbstate.ctxs[asid].ctx_id) != + next->context.ctx_id) + continue; + + *new_asid = asid; + *need_flush = (this_cpu_read(cpu_tlbstate.ctxs[asid].tlb_gen) < + next_tlb_gen); + return; + } + + /* + * We don't currently own an ASID slot on this CPU. + * Allocate a slot. + */ + *new_asid = this_cpu_add_return(cpu_tlbstate.next_asid, 1) - 1; + if (*new_asid >= TLB_NR_DYN_ASIDS) { + *new_asid = 0; + this_cpu_write(cpu_tlbstate.next_asid, 1); + } + *need_flush = true; +} + void leave_mm(int cpu) { struct mm_struct *loaded_mm = this_cpu_read(cpu_tlbstate.loaded_mm); @@ -43,12 +79,11 @@ void leave_mm(int cpu) if (loaded_mm == &init_mm) return; - if (this_cpu_read(cpu_tlbstate.state) == TLBSTATE_OK) - BUG(); + /* Warn if we're not lazy. */ + WARN_ON(cpumask_test_cpu(smp_processor_id(), mm_cpumask(loaded_mm))); switch_mm(NULL, &init_mm, NULL); } -EXPORT_SYMBOL_GPL(leave_mm); void switch_mm(struct mm_struct *prev, struct mm_struct *next, struct task_struct *tsk) @@ -63,115 +98,219 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next, void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next, struct task_struct *tsk) { - unsigned cpu = smp_processor_id(); struct mm_struct *real_prev = this_cpu_read(cpu_tlbstate.loaded_mm); + u16 prev_asid = this_cpu_read(cpu_tlbstate.loaded_mm_asid); + unsigned cpu = smp_processor_id(); + u64 next_tlb_gen; /* - * NB: The scheduler will call us with prev == next when - * switching from lazy TLB mode to normal mode if active_mm - * isn't changing. When this happens, there is no guarantee - * that CR3 (and hence cpu_tlbstate.loaded_mm) matches next. + * NB: The scheduler will call us with prev == next when switching + * from lazy TLB mode to normal mode if active_mm isn't changing. + * When this happens, we don't assume that CR3 (and hence + * cpu_tlbstate.loaded_mm) matches next. * * NB: leave_mm() calls us with prev == NULL and tsk == NULL. */ - this_cpu_write(cpu_tlbstate.state, TLBSTATE_OK); + /* We don't want flush_tlb_func_* to run concurrently with us. */ + if (IS_ENABLED(CONFIG_PROVE_LOCKING)) + WARN_ON_ONCE(!irqs_disabled()); + + /* + * Verify that CR3 is what we think it is. This will catch + * hypothetical buggy code that directly switches to swapper_pg_dir + * without going through leave_mm() / switch_mm_irqs_off() or that + * does something like write_cr3(read_cr3_pa()). + */ + VM_BUG_ON(__read_cr3() != (__sme_pa(real_prev->pgd) | prev_asid)); if (real_prev == next) { - /* - * There's nothing to do: we always keep the per-mm control - * regs in sync with cpu_tlbstate.loaded_mm. Just - * sanity-check mm_cpumask. - */ - if (WARN_ON_ONCE(!cpumask_test_cpu(cpu, mm_cpumask(next)))) - cpumask_set_cpu(cpu, mm_cpumask(next)); - return; - } + VM_BUG_ON(this_cpu_read(cpu_tlbstate.ctxs[prev_asid].ctx_id) != + next->context.ctx_id); + + if (cpumask_test_cpu(cpu, mm_cpumask(next))) { + /* + * There's nothing to do: we weren't lazy, and we + * aren't changing our mm. We don't need to flush + * anything, nor do we need to update CR3, CR4, or + * LDTR. + */ + return; + } + + /* Resume remote flushes and then read tlb_gen. */ + cpumask_set_cpu(cpu, mm_cpumask(next)); + next_tlb_gen = atomic64_read(&next->context.tlb_gen); + + if (this_cpu_read(cpu_tlbstate.ctxs[prev_asid].tlb_gen) < + next_tlb_gen) { + /* + * Ideally, we'd have a flush_tlb() variant that + * takes the known CR3 value as input. This would + * be faster on Xen PV and on hypothetical CPUs + * on which INVPCID is fast. + */ + this_cpu_write(cpu_tlbstate.ctxs[prev_asid].tlb_gen, + next_tlb_gen); + write_cr3(__sme_pa(next->pgd) | prev_asid); + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, + TLB_FLUSH_ALL); + } - if (IS_ENABLED(CONFIG_VMAP_STACK)) { /* - * If our current stack is in vmalloc space and isn't - * mapped in the new pgd, we'll double-fault. Forcibly - * map it. + * We just exited lazy mode, which means that CR4 and/or LDTR + * may be stale. (Changes to the required CR4 and LDTR states + * are not reflected in tlb_gen.) */ - unsigned int stack_pgd_index = pgd_index(current_stack_pointer()); - - pgd_t *pgd = next->pgd + stack_pgd_index; - - if (unlikely(pgd_none(*pgd))) - set_pgd(pgd, init_mm.pgd[stack_pgd_index]); - } + } else { + u16 new_asid; + bool need_flush; + + if (IS_ENABLED(CONFIG_VMAP_STACK)) { + /* + * If our current stack is in vmalloc space and isn't + * mapped in the new pgd, we'll double-fault. Forcibly + * map it. + */ + unsigned int index = pgd_index(current_stack_pointer()); + pgd_t *pgd = next->pgd + index; + + if (unlikely(pgd_none(*pgd))) + set_pgd(pgd, init_mm.pgd[index]); + } - this_cpu_write(cpu_tlbstate.loaded_mm, next); + /* Stop remote flushes for the previous mm */ + if (cpumask_test_cpu(cpu, mm_cpumask(real_prev))) + cpumask_clear_cpu(cpu, mm_cpumask(real_prev)); - WARN_ON_ONCE(cpumask_test_cpu(cpu, mm_cpumask(next))); - cpumask_set_cpu(cpu, mm_cpumask(next)); + VM_WARN_ON_ONCE(cpumask_test_cpu(cpu, mm_cpumask(next))); - /* - * Re-load page tables. - * - * This logic has an ordering constraint: - * - * CPU 0: Write to a PTE for 'next' - * CPU 0: load bit 1 in mm_cpumask. if nonzero, send IPI. - * CPU 1: set bit 1 in next's mm_cpumask - * CPU 1: load from the PTE that CPU 0 writes (implicit) - * - * We need to prevent an outcome in which CPU 1 observes - * the new PTE value and CPU 0 observes bit 1 clear in - * mm_cpumask. (If that occurs, then the IPI will never - * be sent, and CPU 0's TLB will contain a stale entry.) - * - * The bad outcome can occur if either CPU's load is - * reordered before that CPU's store, so both CPUs must - * execute full barriers to prevent this from happening. - * - * Thus, switch_mm needs a full barrier between the - * store to mm_cpumask and any operation that could load - * from next->pgd. TLB fills are special and can happen - * due to instruction fetches or for no reason at all, - * and neither LOCK nor MFENCE orders them. - * Fortunately, load_cr3() is serializing and gives the - * ordering guarantee we need. - */ - load_cr3(next->pgd); - - /* - * This gets called via leave_mm() in the idle path where RCU - * functions differently. Tracing normally uses RCU, so we have to - * call the tracepoint specially here. - */ - trace_tlb_flush_rcuidle(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); + /* + * Start remote flushes and then read tlb_gen. + */ + 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); + write_cr3(__sme_pa(next->pgd) | new_asid); + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, + TLB_FLUSH_ALL); + } else { + /* The new ASID is already up to date. */ + write_cr3(__sme_pa(next->pgd) | new_asid | CR3_NOFLUSH); + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, 0); + } - /* Stop flush ipis for the previous mm */ - WARN_ON_ONCE(!cpumask_test_cpu(cpu, mm_cpumask(real_prev)) && - real_prev != &init_mm); - cpumask_clear_cpu(cpu, mm_cpumask(real_prev)); + this_cpu_write(cpu_tlbstate.loaded_mm, next); + this_cpu_write(cpu_tlbstate.loaded_mm_asid, new_asid); + } - /* Load per-mm CR4 and LDTR state */ load_mm_cr4(next); switch_ldt(real_prev, next); } +/* + * flush_tlb_func_common()'s memory ordering requirement is that any + * TLB fills that happen after we flush the TLB are ordered after we + * read active_mm's tlb_gen. We don't need any explicit barriers + * because all x86 flush operations are serializing and the + * atomic64_read operation won't be reordered by the compiler. + */ static void flush_tlb_func_common(const struct flush_tlb_info *f, bool local, enum tlb_flush_reason reason) { + /* + * We have three different tlb_gen values in here. They are: + * + * - mm_tlb_gen: the latest generation. + * - local_tlb_gen: the generation that this CPU has already caught + * up to. + * - f->new_tlb_gen: the generation that the requester of the flush + * wants us to catch up to. + */ + struct mm_struct *loaded_mm = this_cpu_read(cpu_tlbstate.loaded_mm); + u32 loaded_mm_asid = this_cpu_read(cpu_tlbstate.loaded_mm_asid); + u64 mm_tlb_gen = atomic64_read(&loaded_mm->context.tlb_gen); + u64 local_tlb_gen = this_cpu_read(cpu_tlbstate.ctxs[loaded_mm_asid].tlb_gen); + /* This code cannot presently handle being reentered. */ VM_WARN_ON(!irqs_disabled()); - if (this_cpu_read(cpu_tlbstate.state) != TLBSTATE_OK) { - leave_mm(smp_processor_id()); + VM_WARN_ON(this_cpu_read(cpu_tlbstate.ctxs[loaded_mm_asid].ctx_id) != + loaded_mm->context.ctx_id); + + if (!cpumask_test_cpu(smp_processor_id(), mm_cpumask(loaded_mm))) { + /* + * We're in lazy mode -- don't flush. We can get here on + * remote flushes due to races and on local flushes if a + * kernel thread coincidentally flushes the mm it's lazily + * still using. + */ return; } - if (f->end == TLB_FLUSH_ALL) { - local_flush_tlb(); - if (local) - count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ALL); - trace_tlb_flush(reason, TLB_FLUSH_ALL); - } else { + if (unlikely(local_tlb_gen == mm_tlb_gen)) { + /* + * There's nothing to do: we're already up to date. This can + * happen if two concurrent flushes happen -- the first flush to + * be handled can catch us all the way up, leaving no work for + * the second flush. + */ + trace_tlb_flush(reason, 0); + return; + } + + WARN_ON_ONCE(local_tlb_gen > mm_tlb_gen); + WARN_ON_ONCE(f->new_tlb_gen > mm_tlb_gen); + + /* + * If we get to this point, we know that our TLB is out of date. + * This does not strictly imply that we need to flush (it's + * possible that f->new_tlb_gen <= local_tlb_gen), but we're + * going to need to flush in the very near future, so we might + * as well get it over with. + * + * The only question is whether to do a full or partial flush. + * + * We do a partial flush if requested and two extra conditions + * are met: + * + * 1. f->new_tlb_gen == local_tlb_gen + 1. We have an invariant that + * we've always done all needed flushes to catch up to + * local_tlb_gen. If, for example, local_tlb_gen == 2 and + * f->new_tlb_gen == 3, then we know that the flush needed to bring + * us up to date for tlb_gen 3 is the partial flush we're + * processing. + * + * As an example of why this check is needed, suppose that there + * are two concurrent flushes. The first is a full flush that + * changes context.tlb_gen from 1 to 2. The second is a partial + * flush that changes context.tlb_gen from 2 to 3. If they get + * processed on this CPU in reverse order, we'll see + * local_tlb_gen == 1, mm_tlb_gen == 3, and end != TLB_FLUSH_ALL. + * If we were to use __flush_tlb_single() and set local_tlb_gen to + * 3, we'd be break the invariant: we'd update local_tlb_gen above + * 1 without the full flush that's needed for tlb_gen 2. + * + * 2. f->new_tlb_gen == mm_tlb_gen. This is purely an optimiation. + * Partial TLB flushes are not all that much cheaper than full TLB + * flushes, so it seems unlikely that it would be a performance win + * to do a partial flush if that won't bring our TLB fully up to + * date. By doing a full flush instead, we can increase + * local_tlb_gen all the way to mm_tlb_gen and we can probably + * avoid another flush in the very near future. + */ + if (f->end != TLB_FLUSH_ALL && + f->new_tlb_gen == local_tlb_gen + 1 && + f->new_tlb_gen == mm_tlb_gen) { + /* Partial flush */ unsigned long addr; unsigned long nr_pages = (f->end - f->start) >> PAGE_SHIFT; + addr = f->start; while (addr < f->end) { __flush_tlb_single(addr); @@ -180,7 +319,16 @@ static void flush_tlb_func_common(const struct flush_tlb_info *f, if (local) count_vm_tlb_events(NR_TLB_LOCAL_FLUSH_ONE, nr_pages); trace_tlb_flush(reason, nr_pages); + } else { + /* Full flush. */ + local_flush_tlb(); + if (local) + count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ALL); + trace_tlb_flush(reason, TLB_FLUSH_ALL); } + + /* Both paths above update our state to mm_tlb_gen. */ + this_cpu_write(cpu_tlbstate.ctxs[loaded_mm_asid].tlb_gen, mm_tlb_gen); } static void flush_tlb_func_local(void *info, enum tlb_flush_reason reason) @@ -214,6 +362,21 @@ void native_flush_tlb_others(const struct cpumask *cpumask, (info->end - info->start) >> PAGE_SHIFT); if (is_uv_system()) { + /* + * This whole special case is confused. UV has a "Broadcast + * Assist Unit", which seems to be a fancy way to send IPIs. + * Back when x86 used an explicit TLB flush IPI, UV was + * optimized to use its own mechanism. These days, x86 uses + * smp_call_function_many(), but UV still uses a manual IPI, + * and that IPI's action is out of date -- it does a manual + * flush instead of calling flush_tlb_func_remote(). This + * means that the percpu tlb_gen variables won't be updated + * and we'll do pointless flushes on future context switches. + * + * Rather than hooking native_flush_tlb_others() here, I think + * that UV should be updated so that smp_call_function_many(), + * etc, are optimal on UV. + */ unsigned int cpu; cpu = smp_processor_id(); @@ -250,8 +413,8 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, cpu = get_cpu(); - /* Synchronize with switch_mm. */ - smp_mb(); + /* This is also a barrier that synchronizes with switch_mm(). */ + info.new_tlb_gen = inc_mm_tlb_gen(mm); /* Should we flush just the requested range? */ if ((end != TLB_FLUSH_ALL) && @@ -273,6 +436,7 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start, if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids) flush_tlb_others(mm_cpumask(mm), &info); + put_cpu(); } @@ -281,8 +445,6 @@ static void do_flush_tlb_all(void *info) { count_vm_tlb_event(NR_TLB_REMOTE_FLUSH_RECEIVED); __flush_tlb_all(); - if (this_cpu_read(cpu_tlbstate.state) == TLBSTATE_LAZY) - leave_mm(smp_processor_id()); } void flush_tlb_all(void) @@ -335,6 +497,7 @@ void arch_tlbbatch_flush(struct arch_tlbflush_unmap_batch *batch) if (cpumask_any_but(&batch->cpumask, cpu) < nr_cpu_ids) flush_tlb_others(&batch->cpumask, &info); + cpumask_clear(&batch->cpumask); put_cpu(); |