diff options
Diffstat (limited to 'arch/x86/kernel/alternative.c')
| -rw-r--r-- | arch/x86/kernel/alternative.c | 1797 |
1 files changed, 1262 insertions, 535 deletions
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c index a5ead6a6d233..28518371d8bf 100644 --- a/arch/x86/kernel/alternative.c +++ b/arch/x86/kernel/alternative.c @@ -1,35 +1,18 @@ // SPDX-License-Identifier: GPL-2.0-only #define pr_fmt(fmt) "SMP alternatives: " fmt -#include <linux/module.h> -#include <linux/sched.h> +#include <linux/mmu_context.h> #include <linux/perf_event.h> -#include <linux/mutex.h> -#include <linux/list.h> -#include <linux/stringify.h> -#include <linux/highmem.h> -#include <linux/mm.h> #include <linux/vmalloc.h> #include <linux/memory.h> -#include <linux/stop_machine.h> -#include <linux/slab.h> -#include <linux/kdebug.h> -#include <linux/kprobes.h> -#include <linux/mmu_context.h> -#include <linux/bsearch.h> -#include <linux/sync_core.h> +#include <linux/execmem.h> + #include <asm/text-patching.h> -#include <asm/alternative.h> -#include <asm/sections.h> -#include <asm/mce.h> -#include <asm/nmi.h> -#include <asm/cacheflush.h> -#include <asm/tlbflush.h> #include <asm/insn.h> -#include <asm/io.h> -#include <asm/fixmap.h> -#include <asm/paravirt.h> -#include <asm/asm-prototypes.h> +#include <asm/insn-eval.h> +#include <asm/ibt.h> +#include <asm/set_memory.h> +#include <asm/nmi.h> int __read_mostly alternatives_patched; @@ -44,7 +27,7 @@ EXPORT_SYMBOL_GPL(alternatives_patched); #define DA_ENDBR 0x08 #define DA_SMP 0x10 -static unsigned int __initdata_or_module debug_alternative; +static unsigned int debug_alternative; static int __init debug_alt(char *str) { @@ -123,6 +106,213 @@ const unsigned char * const x86_nops[ASM_NOP_MAX+1] = #endif }; +#ifdef CONFIG_FINEIBT +static bool cfi_paranoid __ro_after_init; +#endif + +#ifdef CONFIG_MITIGATION_ITS + +#ifdef CONFIG_MODULES +static struct module *its_mod; +#endif +static void *its_page; +static unsigned int its_offset; +struct its_array its_pages; + +static void *__its_alloc(struct its_array *pages) +{ + void *page __free(execmem) = execmem_alloc_rw(EXECMEM_MODULE_TEXT, PAGE_SIZE); + if (!page) + return NULL; + + void *tmp = krealloc(pages->pages, (pages->num+1) * sizeof(void *), + GFP_KERNEL); + if (!tmp) + return NULL; + + pages->pages = tmp; + pages->pages[pages->num++] = page; + + return no_free_ptr(page); +} + +/* Initialize a thunk with the "jmp *reg; int3" instructions. */ +static void *its_init_thunk(void *thunk, int reg) +{ + u8 *bytes = thunk; + int offset = 0; + int i = 0; + +#ifdef CONFIG_FINEIBT + if (cfi_paranoid) { + /* + * When ITS uses indirect branch thunk the fineibt_paranoid + * caller sequence doesn't fit in the caller site. So put the + * remaining part of the sequence (UDB + JNE) into the ITS + * thunk. + */ + bytes[i++] = 0xd6; /* UDB */ + bytes[i++] = 0x75; /* JNE */ + bytes[i++] = 0xfd; + + offset = 1; + } +#endif + + if (reg >= 8) { + bytes[i++] = 0x41; /* REX.B prefix */ + reg -= 8; + } + bytes[i++] = 0xff; + bytes[i++] = 0xe0 + reg; /* JMP *reg */ + bytes[i++] = 0xcc; + + return thunk + offset; +} + +static void its_pages_protect(struct its_array *pages) +{ + for (int i = 0; i < pages->num; i++) { + void *page = pages->pages[i]; + execmem_restore_rox(page, PAGE_SIZE); + } +} + +static void its_fini_core(void) +{ + if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) + its_pages_protect(&its_pages); + kfree(its_pages.pages); +} + +#ifdef CONFIG_MODULES +void its_init_mod(struct module *mod) +{ + if (!cpu_feature_enabled(X86_FEATURE_INDIRECT_THUNK_ITS)) + return; + + mutex_lock(&text_mutex); + its_mod = mod; + its_page = NULL; +} + +void its_fini_mod(struct module *mod) +{ + if (!cpu_feature_enabled(X86_FEATURE_INDIRECT_THUNK_ITS)) + return; + + WARN_ON_ONCE(its_mod != mod); + + its_mod = NULL; + its_page = NULL; + mutex_unlock(&text_mutex); + + if (IS_ENABLED(CONFIG_STRICT_MODULE_RWX)) + its_pages_protect(&mod->arch.its_pages); +} + +void its_free_mod(struct module *mod) +{ + if (!cpu_feature_enabled(X86_FEATURE_INDIRECT_THUNK_ITS)) + return; + + for (int i = 0; i < mod->arch.its_pages.num; i++) { + void *page = mod->arch.its_pages.pages[i]; + execmem_free(page); + } + kfree(mod->arch.its_pages.pages); +} +#endif /* CONFIG_MODULES */ + +static void *its_alloc(void) +{ + struct its_array *pages = &its_pages; + void *page; + +#ifdef CONFIG_MODULES + if (its_mod) + pages = &its_mod->arch.its_pages; +#endif + + page = __its_alloc(pages); + if (!page) + return NULL; + + if (pages == &its_pages) + set_memory_x((unsigned long)page, 1); + + return page; +} + +static void *its_allocate_thunk(int reg) +{ + int size = 3 + (reg / 8); + void *thunk; + +#ifdef CONFIG_FINEIBT + /* + * The ITS thunk contains an indirect jump and an int3 instruction so + * its size is 3 or 4 bytes depending on the register used. If CFI + * paranoid is used then 3 extra bytes are added in the ITS thunk to + * complete the fineibt_paranoid caller sequence. + */ + if (cfi_paranoid) + size += 3; +#endif + + if (!its_page || (its_offset + size - 1) >= PAGE_SIZE) { + its_page = its_alloc(); + if (!its_page) { + pr_err("ITS page allocation failed\n"); + return NULL; + } + memset(its_page, INT3_INSN_OPCODE, PAGE_SIZE); + its_offset = 32; + } + + /* + * If the indirect branch instruction will be in the lower half + * of a cacheline, then update the offset to reach the upper half. + */ + if ((its_offset + size - 1) % 64 < 32) + its_offset = ((its_offset - 1) | 0x3F) + 33; + + thunk = its_page + its_offset; + its_offset += size; + + return its_init_thunk(thunk, reg); +} + +u8 *its_static_thunk(int reg) +{ + u8 *thunk = __x86_indirect_its_thunk_array[reg]; + +#ifdef CONFIG_FINEIBT + /* Paranoid thunk starts 2 bytes before */ + if (cfi_paranoid) + return thunk - 2; +#endif + return thunk; +} + +#else +static inline void its_fini_core(void) {} +#endif /* CONFIG_MITIGATION_ITS */ + +/* + * Nomenclature for variable names to simplify and clarify this code and ease + * any potential staring at it: + * + * @instr: source address of the original instructions in the kernel text as + * generated by the compiler. + * + * @buf: temporary buffer on which the patching operates. This buffer is + * eventually text-poked into the kernel image. + * + * @replacement/@repl: pointer to the opcodes which are replacing @instr, located + * in the .altinstr_replacement section. + */ + /* * Fill the buffer with a single effective instruction of size @len. * @@ -132,67 +322,40 @@ const unsigned char * const x86_nops[ASM_NOP_MAX+1] = * each single-byte NOPs). If @len to fill out is > ASM_NOP_MAX, pad with INT3 and * *jump* over instead of executing long and daft NOPs. */ -static void __init_or_module add_nop(u8 *instr, unsigned int len) +static void add_nop(u8 *buf, unsigned int len) { - u8 *target = instr + len; + u8 *target = buf + len; if (!len) return; if (len <= ASM_NOP_MAX) { - memcpy(instr, x86_nops[len], len); + memcpy(buf, x86_nops[len], len); return; } if (len < 128) { - __text_gen_insn(instr, JMP8_INSN_OPCODE, instr, target, JMP8_INSN_SIZE); - instr += JMP8_INSN_SIZE; + __text_gen_insn(buf, JMP8_INSN_OPCODE, buf, target, JMP8_INSN_SIZE); + buf += JMP8_INSN_SIZE; } else { - __text_gen_insn(instr, JMP32_INSN_OPCODE, instr, target, JMP32_INSN_SIZE); - instr += JMP32_INSN_SIZE; + __text_gen_insn(buf, JMP32_INSN_OPCODE, buf, target, JMP32_INSN_SIZE); + buf += JMP32_INSN_SIZE; } - for (;instr < target; instr++) - *instr = INT3_INSN_OPCODE; -} - -extern s32 __retpoline_sites[], __retpoline_sites_end[]; -extern s32 __return_sites[], __return_sites_end[]; -extern s32 __cfi_sites[], __cfi_sites_end[]; -extern s32 __ibt_endbr_seal[], __ibt_endbr_seal_end[]; -extern struct alt_instr __alt_instructions[], __alt_instructions_end[]; -extern s32 __smp_locks[], __smp_locks_end[]; -void text_poke_early(void *addr, const void *opcode, size_t len); - -/* - * Matches NOP and NOPL, not any of the other possible NOPs. - */ -static bool insn_is_nop(struct insn *insn) -{ - /* Anything NOP, but no REP NOP */ - if (insn->opcode.bytes[0] == 0x90 && - (!insn->prefixes.nbytes || insn->prefixes.bytes[0] != 0xF3)) - return true; - - /* NOPL */ - if (insn->opcode.bytes[0] == 0x0F && insn->opcode.bytes[1] == 0x1F) - return true; - - /* TODO: more nops */ - - return false; + for (;buf < target; buf++) + *buf = INT3_INSN_OPCODE; } /* * Find the offset of the first non-NOP instruction starting at @offset * but no further than @len. */ -static int skip_nops(u8 *instr, int offset, int len) +static int skip_nops(u8 *buf, int offset, int len) { struct insn insn; for (; offset < len; offset += insn.length) { - if (insn_decode_kernel(&insn, &instr[offset])) + if (insn_decode_kernel(&insn, &buf[offset])) break; if (!insn_is_nop(&insn)) @@ -203,55 +366,31 @@ static int skip_nops(u8 *instr, int offset, int len) } /* - * Optimize a sequence of NOPs, possibly preceded by an unconditional jump - * to the end of the NOP sequence into a single NOP. - */ -static bool __init_or_module -__optimize_nops(u8 *instr, size_t len, struct insn *insn, int *next, int *prev, int *target) -{ - int i = *next - insn->length; - - switch (insn->opcode.bytes[0]) { - case JMP8_INSN_OPCODE: - case JMP32_INSN_OPCODE: - *prev = i; - *target = *next + insn->immediate.value; - return false; - } - - if (insn_is_nop(insn)) { - int nop = i; - - *next = skip_nops(instr, *next, len); - if (*target && *next == *target) - nop = *prev; - - add_nop(instr + nop, *next - nop); - DUMP_BYTES(ALT, instr, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, *next); - return true; - } - - *target = 0; - return false; -} - -/* * "noinline" to cause control flow change and thus invalidate I$ and * cause refetch after modification. */ -static void __init_or_module noinline optimize_nops(u8 *instr, size_t len) +static void noinline optimize_nops(const u8 * const instr, u8 *buf, size_t len) { - int prev, target = 0; - for (int next, i = 0; i < len; i = next) { struct insn insn; - if (insn_decode_kernel(&insn, &instr[i])) + if (insn_decode_kernel(&insn, &buf[i])) return; next = i + insn.length; - __optimize_nops(instr, len, &insn, &next, &prev, &target); + if (insn_is_nop(&insn)) { + int nop = i; + + /* Has the NOP already been optimized? */ + if (i + insn.length == len) + return; + + next = skip_nops(buf, next, len); + + add_nop(buf + nop, next - nop); + DUMP_BYTES(ALT, buf, len, "%px: [%d:%d) optimized NOPs: ", instr, nop, next); + } } } @@ -325,12 +464,9 @@ bool need_reloc(unsigned long offset, u8 *src, size_t src_len) return (target < src || target > src + src_len); } -static void __init_or_module noinline -apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) +static void __apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl, size_t repl_len) { - int prev, target = 0; - - for (int next, i = 0; i < len; i = next) { + for (int next, i = 0; i < instrlen; i = next) { struct insn insn; if (WARN_ON_ONCE(insn_decode_kernel(&insn, &buf[i]))) @@ -338,9 +474,6 @@ apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) next = i + insn.length; - if (__optimize_nops(buf, len, &insn, &next, &prev, &target)) - continue; - switch (insn.opcode.bytes[0]) { case 0x0f: if (insn.opcode.bytes[1] < 0x80 || @@ -352,10 +485,10 @@ apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) case JMP8_INSN_OPCODE: case JMP32_INSN_OPCODE: case CALL_INSN_OPCODE: - if (need_reloc(next + insn.immediate.value, src, src_len)) { + if (need_reloc(next + insn.immediate.value, repl, repl_len)) { apply_reloc(insn.immediate.nbytes, buf + i + insn_offset_immediate(&insn), - src - dest); + repl - instr); } /* @@ -363,7 +496,7 @@ apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) */ if (insn.opcode.bytes[0] == JMP32_INSN_OPCODE) { s32 imm = insn.immediate.value; - imm += src - dest; + imm += repl - instr; imm += JMP32_INSN_SIZE - JMP8_INSN_SIZE; if ((imm >> 31) == (imm >> 7)) { buf[i+0] = JMP8_INSN_OPCODE; @@ -376,15 +509,83 @@ apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) } if (insn_rip_relative(&insn)) { - if (need_reloc(next + insn.displacement.value, src, src_len)) { + if (need_reloc(next + insn.displacement.value, repl, repl_len)) { apply_reloc(insn.displacement.nbytes, buf + i + insn_offset_displacement(&insn), - src - dest); + repl - instr); } } } } +void text_poke_apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u8 *repl, size_t repl_len) +{ + __apply_relocation(buf, instr, instrlen, repl, repl_len); + optimize_nops(instr, buf, instrlen); +} + +/* Low-level backend functions usable from alternative code replacements. */ +DEFINE_ASM_FUNC(nop_func, "", .entry.text); +EXPORT_SYMBOL_GPL(nop_func); + +noinstr void BUG_func(void) +{ + BUG(); +} +EXPORT_SYMBOL(BUG_func); + +#define CALL_RIP_REL_OPCODE 0xff +#define CALL_RIP_REL_MODRM 0x15 + +/* + * Rewrite the "call BUG_func" replacement to point to the target of the + * indirect pv_ops call "call *disp(%ip)". + */ +static unsigned int alt_replace_call(u8 *instr, u8 *insn_buff, struct alt_instr *a) +{ + void *target, *bug = &BUG_func; + s32 disp; + + if (a->replacementlen != 5 || insn_buff[0] != CALL_INSN_OPCODE) { + pr_err("ALT_FLAG_DIRECT_CALL set for a non-call replacement instruction\n"); + BUG(); + } + + if (a->instrlen != 6 || + instr[0] != CALL_RIP_REL_OPCODE || + instr[1] != CALL_RIP_REL_MODRM) { + pr_err("ALT_FLAG_DIRECT_CALL set for unrecognized indirect call\n"); + BUG(); + } + + /* Skip CALL_RIP_REL_OPCODE and CALL_RIP_REL_MODRM */ + disp = *(s32 *)(instr + 2); +#ifdef CONFIG_X86_64 + /* ff 15 00 00 00 00 call *0x0(%rip) */ + /* target address is stored at "next instruction + disp". */ + target = *(void **)(instr + a->instrlen + disp); +#else + /* ff 15 00 00 00 00 call *0x0 */ + /* target address is stored at disp. */ + target = *(void **)disp; +#endif + if (!target) + target = bug; + + /* (BUG_func - .) + (target - BUG_func) := target - . */ + *(s32 *)(insn_buff + 1) += target - bug; + + if (target == &nop_func) + return 0; + + return 5; +} + +static inline u8 * instr_va(struct alt_instr *i) +{ + return (u8 *)&i->instr_offset + i->instr_offset; +} + /* * Replace instructions with better alternatives for this CPU type. This runs * before SMP is initialized to avoid SMP problems with self modifying code. @@ -398,11 +599,22 @@ apply_relocation(u8 *buf, size_t len, u8 *dest, u8 *src, size_t src_len) void __init_or_module noinline apply_alternatives(struct alt_instr *start, struct alt_instr *end) { - struct alt_instr *a; - u8 *instr, *replacement; u8 insn_buff[MAX_PATCH_LEN]; + u8 *instr, *replacement; + struct alt_instr *a, *b; DPRINTK(ALT, "alt table %px, -> %px", start, end); + + /* + * KASAN_SHADOW_START is defined using + * cpu_feature_enabled(X86_FEATURE_LA57) and is therefore patched here. + * During the process, KASAN becomes confused seeing partial LA57 + * conversion and triggers a false-positive out-of-bound report. + * + * Disable KASAN until the patching is complete. + */ + kasan_disable_current(); + /* * The scan order should be from start to end. A later scanned * alternative code can overwrite previously scanned alternative code. @@ -413,9 +625,20 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start, * order. */ for (a = start; a < end; a++) { - int insn_buff_sz = 0; + unsigned int insn_buff_sz = 0; + + /* + * In case of nested ALTERNATIVE()s the outer alternative might + * add more padding. To ensure consistent patching find the max + * padding for all alt_instr entries for this site (nested + * alternatives result in consecutive entries). + */ + for (b = a+1; b < end && instr_va(b) == instr_va(a); b++) { + u8 len = max(a->instrlen, b->instrlen); + a->instrlen = b->instrlen = len; + } - instr = (u8 *)&a->instr_offset + a->instr_offset; + instr = instr_va(a); replacement = (u8 *)&a->repl_offset + a->repl_offset; BUG_ON(a->instrlen > sizeof(insn_buff)); BUG_ON(a->cpuid >= (NCAPINTS + NBUGINTS) * 32); @@ -427,24 +650,28 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start, * patch if feature is *NOT* present. */ if (!boot_cpu_has(a->cpuid) == !(a->flags & ALT_FLAG_NOT)) { - optimize_nops(instr, a->instrlen); + memcpy(insn_buff, instr, a->instrlen); + optimize_nops(instr, insn_buff, a->instrlen); + text_poke_early(instr, insn_buff, a->instrlen); continue; } - DPRINTK(ALT, "feat: %s%d*32+%d, old: (%pS (%px) len: %d), repl: (%px, len: %d)", - (a->flags & ALT_FLAG_NOT) ? "!" : "", + DPRINTK(ALT, "feat: %d*32+%d, old: (%pS (%px) len: %d), repl: (%px, len: %d) flags: 0x%x", a->cpuid >> 5, a->cpuid & 0x1f, instr, instr, a->instrlen, - replacement, a->replacementlen); + replacement, a->replacementlen, a->flags); memcpy(insn_buff, replacement, a->replacementlen); insn_buff_sz = a->replacementlen; + if (a->flags & ALT_FLAG_DIRECT_CALL) + insn_buff_sz = alt_replace_call(instr, insn_buff, a); + for (; insn_buff_sz < a->instrlen; insn_buff_sz++) insn_buff[insn_buff_sz] = 0x90; - apply_relocation(insn_buff, a->instrlen, instr, replacement, a->replacementlen); + text_poke_apply_relocation(insn_buff, instr, a->instrlen, replacement, a->replacementlen); DUMP_BYTES(ALT, instr, a->instrlen, "%px: old_insn: ", instr); DUMP_BYTES(ALT, replacement, a->replacementlen, "%px: rpl_insn: ", replacement); @@ -452,6 +679,8 @@ void __init_or_module noinline apply_alternatives(struct alt_instr *start, text_poke_early(instr, insn_buff, insn_buff_sz); } + + kasan_enable_current(); } static inline bool is_jcc32(struct insn *insn) @@ -460,23 +689,36 @@ static inline bool is_jcc32(struct insn *insn) return insn->opcode.bytes[0] == 0x0f && (insn->opcode.bytes[1] & 0xf0) == 0x80; } -#if defined(CONFIG_RETPOLINE) && defined(CONFIG_OBJTOOL) +#if defined(CONFIG_MITIGATION_RETPOLINE) && defined(CONFIG_OBJTOOL) /* - * CALL/JMP *%\reg + * [CS]{,3} CALL/JMP *%\reg [INT3]* */ -static int emit_indirect(int op, int reg, u8 *bytes) +static int emit_indirect(int op, int reg, u8 *bytes, int len) { + int cs = 0, bp = 0; int i = 0; u8 modrm; + /* + * Set @len to the excess bytes after writing the instruction. + */ + len -= 2 + (reg >= 8); + WARN_ON_ONCE(len < 0); + switch (op) { case CALL_INSN_OPCODE: modrm = 0x10; /* Reg = 2; CALL r/m */ + /* + * Additional NOP is better than prefix decode penalty. + */ + if (len <= 3) + cs = len; break; case JMP32_INSN_OPCODE: modrm = 0x20; /* Reg = 4; JMP r/m */ + bp = len; break; default: @@ -484,6 +726,9 @@ static int emit_indirect(int op, int reg, u8 *bytes) return -1; } + while (cs--) + bytes[i++] = 0x2e; /* CS-prefix */ + if (reg >= 8) { bytes[i++] = 0x41; /* REX.B prefix */ reg -= 8; @@ -495,10 +740,14 @@ static int emit_indirect(int op, int reg, u8 *bytes) bytes[i++] = 0xff; /* opcode */ bytes[i++] = modrm; + while (bp--) + bytes[i++] = 0xcc; /* INT3 */ + return i; } -static int emit_call_track_retpoline(void *addr, struct insn *insn, int reg, u8 *bytes) +static int __emit_trampoline(void *addr, struct insn *insn, u8 *bytes, + void *call_dest, void *jmp_dest) { u8 op = insn->opcode.bytes[0]; int i = 0; @@ -519,7 +768,7 @@ static int emit_call_track_retpoline(void *addr, struct insn *insn, int reg, u8 switch (op) { case CALL_INSN_OPCODE: __text_gen_insn(bytes+i, op, addr+i, - __x86_indirect_call_thunk_array[reg], + call_dest, CALL_INSN_SIZE); i += CALL_INSN_SIZE; break; @@ -527,7 +776,7 @@ static int emit_call_track_retpoline(void *addr, struct insn *insn, int reg, u8 case JMP32_INSN_OPCODE: clang_jcc: __text_gen_insn(bytes+i, op, addr+i, - __x86_indirect_jump_thunk_array[reg], + jmp_dest, JMP32_INSN_SIZE); i += JMP32_INSN_SIZE; break; @@ -542,6 +791,48 @@ clang_jcc: return i; } +static int emit_call_track_retpoline(void *addr, struct insn *insn, int reg, u8 *bytes) +{ + return __emit_trampoline(addr, insn, bytes, + __x86_indirect_call_thunk_array[reg], + __x86_indirect_jump_thunk_array[reg]); +} + +#ifdef CONFIG_MITIGATION_ITS +static int emit_its_trampoline(void *addr, struct insn *insn, int reg, u8 *bytes) +{ + u8 *thunk = __x86_indirect_its_thunk_array[reg]; + u8 *tmp = its_allocate_thunk(reg); + + if (tmp) + thunk = tmp; + + return __emit_trampoline(addr, insn, bytes, thunk, thunk); +} + +/* Check if an indirect branch is at ITS-unsafe address */ +static bool cpu_wants_indirect_its_thunk_at(unsigned long addr, int reg) +{ + if (!cpu_feature_enabled(X86_FEATURE_INDIRECT_THUNK_ITS)) + return false; + + /* Indirect branch opcode is 2 or 3 bytes depending on reg */ + addr += 1 + reg / 8; + + /* Lower-half of the cacheline? */ + return !(addr & 0x20); +} +#else /* CONFIG_MITIGATION_ITS */ + +#ifdef CONFIG_FINEIBT +static bool cpu_wants_indirect_its_thunk_at(unsigned long addr, int reg) +{ + return false; +} +#endif + +#endif /* CONFIG_MITIGATION_ITS */ + /* * Rewrite the compiler generated retpoline thunk calls. * @@ -616,20 +907,20 @@ static int patch_retpoline(void *addr, struct insn *insn, u8 *bytes) bytes[i++] = 0xe8; /* LFENCE */ } - ret = emit_indirect(op, reg, bytes + i); +#ifdef CONFIG_MITIGATION_ITS + /* + * Check if the address of last byte of emitted-indirect is in + * lower-half of the cacheline. Such branches need ITS mitigation. + */ + if (cpu_wants_indirect_its_thunk_at((unsigned long)addr + i, reg)) + return emit_its_trampoline(addr, insn, reg, bytes); +#endif + + ret = emit_indirect(op, reg, bytes + i, insn->length - i); if (ret < 0) return ret; i += ret; - /* - * The compiler is supposed to EMIT an INT3 after every unconditional - * JMP instruction due to AMD BTC. However, if the compiler is too old - * or SLS isn't enabled, we still need an INT3 after indirect JMPs - * even on Intel. - */ - if (op == JMP32_INSN_OPCODE && i < insn->length) - bytes[i++] = INT3_INSN_OPCODE; - for (; i < insn->length;) bytes[i++] = BYTES_NOP1; @@ -649,6 +940,7 @@ void __init_or_module noinline apply_retpolines(s32 *start, s32 *end) int len, ret; u8 bytes[16]; u8 op1, op2; + u8 *dest; ret = insn_decode_kernel(&insn, addr); if (WARN_ON_ONCE(ret < 0)) @@ -658,8 +950,19 @@ void __init_or_module noinline apply_retpolines(s32 *start, s32 *end) op2 = insn.opcode.bytes[1]; switch (op1) { + case 0x70 ... 0x7f: /* Jcc.d8 */ + /* See cfi_paranoid. */ + WARN_ON_ONCE(cfi_mode != CFI_FINEIBT); + continue; + case CALL_INSN_OPCODE: case JMP32_INSN_OPCODE: + /* Check for cfi_paranoid + ITS */ + dest = addr + insn.length + insn.immediate.value; + if (dest[-1] == 0xd6 && (dest[0] & 0xf0) == 0x70) { + WARN_ON_ONCE(cfi_mode != CFI_FINEIBT); + continue; + } break; case 0x0f: /* escape */ @@ -677,7 +980,7 @@ void __init_or_module noinline apply_retpolines(s32 *start, s32 *end) len = patch_retpoline(addr, &insn, bytes); if (len == insn.length) { - optimize_nops(bytes, len); + optimize_nops(addr, bytes, len); DUMP_BYTES(RETPOLINE, ((u8*)addr), len, "%px: orig: ", addr); DUMP_BYTES(RETPOLINE, ((u8*)bytes), len, "%px: repl: ", addr); text_poke_early(addr, bytes, len); @@ -685,7 +988,22 @@ void __init_or_module noinline apply_retpolines(s32 *start, s32 *end) } } -#ifdef CONFIG_RETHUNK +#ifdef CONFIG_MITIGATION_RETHUNK + +bool cpu_wants_rethunk(void) +{ + return cpu_feature_enabled(X86_FEATURE_RETHUNK); +} + +bool cpu_wants_rethunk_at(void *addr) +{ + if (!cpu_feature_enabled(X86_FEATURE_RETHUNK)) + return false; + if (x86_return_thunk != its_return_thunk) + return true; + + return !((unsigned long)addr & 0x20); +} /* * Rewrite the compiler generated return thunk tail-calls. @@ -703,7 +1021,7 @@ static int patch_return(void *addr, struct insn *insn, u8 *bytes) int i = 0; /* Patch the custom return thunks... */ - if (cpu_feature_enabled(X86_FEATURE_RETHUNK)) { + if (cpu_wants_rethunk_at(addr)) { i = JMP32_INSN_SIZE; __text_gen_insn(bytes, JMP32_INSN_OPCODE, addr, x86_return_thunk, i); } else { @@ -720,13 +1038,8 @@ void __init_or_module noinline apply_returns(s32 *start, s32 *end) { s32 *s; - /* - * Do not patch out the default return thunks if those needed are the - * ones generated by the compiler. - */ - if (cpu_feature_enabled(X86_FEATURE_RETHUNK) && - (x86_return_thunk == __x86_return_thunk)) - return; + if (cpu_wants_rethunk()) + static_call_force_reinit(); for (s = start; s < end; s++) { void *dest = NULL, *addr = (void *)s + *s; @@ -761,32 +1074,53 @@ void __init_or_module noinline apply_returns(s32 *start, s32 *end) } } } -#else +#else /* !CONFIG_MITIGATION_RETHUNK: */ void __init_or_module noinline apply_returns(s32 *start, s32 *end) { } -#endif /* CONFIG_RETHUNK */ +#endif /* !CONFIG_MITIGATION_RETHUNK */ -#else /* !CONFIG_RETPOLINE || !CONFIG_OBJTOOL */ +#else /* !CONFIG_MITIGATION_RETPOLINE || !CONFIG_OBJTOOL */ void __init_or_module noinline apply_retpolines(s32 *start, s32 *end) { } void __init_or_module noinline apply_returns(s32 *start, s32 *end) { } -#endif /* CONFIG_RETPOLINE && CONFIG_OBJTOOL */ +#endif /* !CONFIG_MITIGATION_RETPOLINE || !CONFIG_OBJTOOL */ #ifdef CONFIG_X86_KERNEL_IBT -static void poison_cfi(void *addr); +__noendbr bool is_endbr(u32 *val) +{ + u32 endbr; + + __get_kernel_nofault(&endbr, val, u32, Efault); + return __is_endbr(endbr); -static void __init_or_module poison_endbr(void *addr, bool warn) +Efault: + return false; +} + +#ifdef CONFIG_FINEIBT + +static __noendbr bool exact_endbr(u32 *val) { - u32 endbr, poison = gen_endbr_poison(); + u32 endbr; - if (WARN_ON_ONCE(get_kernel_nofault(endbr, addr))) - return; + __get_kernel_nofault(&endbr, val, u32, Efault); + return endbr == gen_endbr(); - if (!is_endbr(endbr)) { - WARN_ON_ONCE(warn); +Efault: + return false; +} + +#endif + +static void poison_cfi(void *addr); + +static void __init_or_module poison_endbr(void *addr) +{ + u32 poison = gen_endbr_poison(); + + if (WARN_ON_ONCE(!is_endbr(addr))) return; - } DPRINTK(ENDBR, "ENDBR at: %pS (%px)", addr, addr); @@ -811,28 +1145,74 @@ void __init_or_module noinline apply_seal_endbr(s32 *start, s32 *end) for (s = start; s < end; s++) { void *addr = (void *)s + *s; - poison_endbr(addr, true); + poison_endbr(addr); if (IS_ENABLED(CONFIG_FINEIBT)) poison_cfi(addr - 16); } } -#else +#else /* !CONFIG_X86_KERNEL_IBT: */ void __init_or_module apply_seal_endbr(s32 *start, s32 *end) { } -#endif /* CONFIG_X86_KERNEL_IBT */ +#endif /* !CONFIG_X86_KERNEL_IBT */ -#ifdef CONFIG_FINEIBT +#ifdef CONFIG_CFI_AUTO_DEFAULT +# define __CFI_DEFAULT CFI_AUTO +#elif defined(CONFIG_CFI) +# define __CFI_DEFAULT CFI_KCFI +#else +# define __CFI_DEFAULT CFI_OFF +#endif -enum cfi_mode { - CFI_DEFAULT, - CFI_OFF, - CFI_KCFI, - CFI_FINEIBT, -}; +enum cfi_mode cfi_mode __ro_after_init = __CFI_DEFAULT; +static bool cfi_debug __ro_after_init; + +#ifdef CONFIG_FINEIBT_BHI +bool cfi_bhi __ro_after_init = false; +#endif + +#ifdef CONFIG_CFI +u32 cfi_get_func_hash(void *func) +{ + u32 hash; + + func -= cfi_get_offset(); + switch (cfi_mode) { + case CFI_FINEIBT: + func += 7; + break; + case CFI_KCFI: + func += 1; + break; + default: + return 0; + } + + if (get_kernel_nofault(hash, func)) + return 0; + + return hash; +} + +int cfi_get_func_arity(void *func) +{ + bhi_thunk *target; + s32 disp; + + if (cfi_mode != CFI_FINEIBT && !cfi_bhi) + return 0; + + if (get_kernel_nofault(disp, func - 4)) + return 0; + + target = func + disp; + return target - __bhi_args; +} +#endif + +#ifdef CONFIG_FINEIBT -static enum cfi_mode cfi_mode __ro_after_init = CFI_DEFAULT; static bool cfi_rand __ro_after_init = true; static u32 cfi_seed __ro_after_init; @@ -843,7 +1223,7 @@ static u32 cfi_seed __ro_after_init; static u32 cfi_rehash(u32 hash) { hash ^= cfi_seed; - while (unlikely(is_endbr(hash) || is_endbr(-hash))) { + while (unlikely(__is_endbr(hash) || __is_endbr(-hash))) { bool lsb = hash & 1; hash >>= 1; if (lsb) @@ -865,18 +1245,39 @@ static __init int cfi_parse_cmdline(char *str) } if (!strcmp(str, "auto")) { - cfi_mode = CFI_DEFAULT; + cfi_mode = CFI_AUTO; } else if (!strcmp(str, "off")) { cfi_mode = CFI_OFF; cfi_rand = false; + } else if (!strcmp(str, "debug")) { + cfi_debug = true; } else if (!strcmp(str, "kcfi")) { cfi_mode = CFI_KCFI; } else if (!strcmp(str, "fineibt")) { cfi_mode = CFI_FINEIBT; } else if (!strcmp(str, "norand")) { cfi_rand = false; + } else if (!strcmp(str, "warn")) { + pr_alert("CFI: mismatch non-fatal!\n"); + cfi_warn = true; + } else if (!strcmp(str, "paranoid")) { + if (cfi_mode == CFI_FINEIBT) { + cfi_paranoid = true; + } else { + pr_err("CFI: ignoring paranoid; depends on fineibt.\n"); + } + } else if (!strcmp(str, "bhi")) { +#ifdef CONFIG_FINEIBT_BHI + if (cfi_mode == CFI_FINEIBT) { + cfi_bhi = true; + } else { + pr_err("CFI: ignoring bhi; depends on fineibt.\n"); + } +#else + pr_err("CFI: ignoring bhi; depends on FINEIBT_BHI=y.\n"); +#endif } else { - pr_err("Ignoring unknown cfi option (%s).", str); + pr_err("CFI: Ignoring unknown option (%s).", str); } str = next; @@ -891,10 +1292,8 @@ early_param("cfi", cfi_parse_cmdline); * * __cfi_\func: __cfi_\func: * movl $0x12345678,%eax // 5 endbr64 // 4 - * nop subl $0x12345678,%r10d // 7 - * nop jz 1f // 2 - * nop ud2 // 2 - * nop 1: nop // 1 + * nop subl $0x12345678,%eax // 5 + * nop jne.d32,pn \func+3 // 7 * nop * nop * nop @@ -902,39 +1301,70 @@ early_param("cfi", cfi_parse_cmdline); * nop * nop * nop + * nop + * nop + * \func: \func: + * endbr64 nopl -42(%rax) * * * caller: caller: - * movl $(-0x12345678),%r10d // 6 movl $0x12345678,%r10d // 6 - * addl $-15(%r11),%r10d // 4 sub $16,%r11 // 4 - * je 1f // 2 nop4 // 4 + * movl $(-0x12345678),%r10d // 6 movl $0x12345678,%eax // 5 + * addl $-15(%r11),%r10d // 4 lea -0x10(%r11),%r11 // 4 + * je 1f // 2 nop5 // 5 * ud2 // 2 - * 1: call __x86_indirect_thunk_r11 // 5 call *%r11; nop2; // 5 + * 1: cs call __x86_indirect_thunk_r11 // 6 call *%r11; nop3; // 6 + * + * + * Notably, the FineIBT sequences are crafted such that branches are presumed + * non-taken. This is based on Agner Fog's optimization manual, which states: * + * "Make conditional jumps most often not taken: The efficiency and throughput + * for not-taken branches is better than for taken branches on most + * processors. Therefore, it is good to place the most frequent branch first" */ -asm( ".pushsection .rodata \n" - "fineibt_preamble_start: \n" - " endbr64 \n" - " subl $0x12345678, %r10d \n" - " je fineibt_preamble_end \n" - " ud2 \n" - " nop \n" - "fineibt_preamble_end: \n" +/* + * <fineibt_preamble_start>: + * 0: f3 0f 1e fa endbr64 + * 4: 2d 78 56 34 12 sub $0x12345678, %eax + * 9: 2e 0f 85 03 00 00 00 jne,pn 13 <fineibt_preamble_start+0x13> + * 10: 0f 1f 40 d6 nopl -0x2a(%rax) + * + * Note that the JNE target is the 0xD6 byte inside the NOPL, this decodes as + * UDB on x86_64 and raises #UD. + */ +asm( ".pushsection .rodata \n" + "fineibt_preamble_start: \n" + " endbr64 \n" + " subl $0x12345678, %eax \n" + "fineibt_preamble_bhi: \n" + " cs jne.d32 fineibt_preamble_start+0x13 \n" + "#fineibt_func: \n" + " nopl -42(%rax) \n" + "fineibt_preamble_end: \n" ".popsection\n" ); extern u8 fineibt_preamble_start[]; +extern u8 fineibt_preamble_bhi[]; extern u8 fineibt_preamble_end[]; #define fineibt_preamble_size (fineibt_preamble_end - fineibt_preamble_start) -#define fineibt_preamble_hash 7 +#define fineibt_preamble_bhi (fineibt_preamble_bhi - fineibt_preamble_start) +#define fineibt_preamble_ud 0x13 +#define fineibt_preamble_hash 5 +/* + * <fineibt_caller_start>: + * 0: b8 78 56 34 12 mov $0x12345678, %eax + * 5: 4d 8d 5b f0 lea -0x10(%r11), %r11 + * 9: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1) + */ asm( ".pushsection .rodata \n" "fineibt_caller_start: \n" - " movl $0x12345678, %r10d \n" - " sub $16, %r11 \n" - ASM_NOP4 + " movl $0x12345678, %eax \n" + " lea -0x10(%r11), %r11 \n" + ASM_NOP5 "fineibt_caller_end: \n" ".popsection \n" ); @@ -943,17 +1373,66 @@ extern u8 fineibt_caller_start[]; extern u8 fineibt_caller_end[]; #define fineibt_caller_size (fineibt_caller_end - fineibt_caller_start) -#define fineibt_caller_hash 2 +#define fineibt_caller_hash 1 #define fineibt_caller_jmp (fineibt_caller_size - 2) -static u32 decode_preamble_hash(void *addr) +/* + * Since FineIBT does hash validation on the callee side it is prone to + * circumvention attacks where a 'naked' ENDBR instruction exists that + * is not part of the fineibt_preamble sequence. + * + * Notably the x86 entry points must be ENDBR and equally cannot be + * fineibt_preamble. + * + * The fineibt_paranoid caller sequence adds additional caller side + * hash validation. This stops such circumvention attacks dead, but at the cost + * of adding a load. + * + * <fineibt_paranoid_start>: + * 0: b8 78 56 34 12 mov $0x12345678, %eax + * 5: 41 3b 43 f5 cmp -0x11(%r11), %eax + * 9: 2e 4d 8d 5b <f0> cs lea -0x10(%r11), %r11 + * e: 75 fd jne d <fineibt_paranoid_start+0xd> + * 10: 41 ff d3 call *%r11 + * 13: 90 nop + * + * Notably LEA does not modify flags and can be reordered with the CMP, + * avoiding a dependency. Again, using a non-taken (backwards) branch + * for the failure case, abusing LEA's immediate 0xf0 as LOCK prefix for the + * Jcc.d8, causing #UD. + */ +asm( ".pushsection .rodata \n" + "fineibt_paranoid_start: \n" + " mov $0x12345678, %eax \n" + " cmpl -11(%r11), %eax \n" + " cs lea -0x10(%r11), %r11 \n" + "#fineibt_caller_size: \n" + " jne fineibt_paranoid_start+0xd \n" + "fineibt_paranoid_ind: \n" + " cs call *%r11 \n" + "fineibt_paranoid_end: \n" + ".popsection \n" +); + +extern u8 fineibt_paranoid_start[]; +extern u8 fineibt_paranoid_ind[]; +extern u8 fineibt_paranoid_end[]; + +#define fineibt_paranoid_size (fineibt_paranoid_end - fineibt_paranoid_start) +#define fineibt_paranoid_ind (fineibt_paranoid_ind - fineibt_paranoid_start) +#define fineibt_paranoid_ud 0xd + +static u32 decode_preamble_hash(void *addr, int *reg) { u8 *p = addr; - /* b8 78 56 34 12 mov $0x12345678,%eax */ - if (p[0] == 0xb8) + /* b8+reg 78 56 34 12 movl $0x12345678,\reg */ + if (p[0] >= 0xb8 && p[0] < 0xc0) { + if (reg) + *reg = p[0] - 0xb8; return *(u32 *)(addr + 1); + } return 0; /* invalid hash value */ } @@ -962,11 +1441,11 @@ static u32 decode_caller_hash(void *addr) { u8 *p = addr; - /* 41 ba 78 56 34 12 mov $0x12345678,%r10d */ + /* 41 ba 88 a9 cb ed mov $(-0x12345678),%r10d */ if (p[0] == 0x41 && p[1] == 0xba) return -*(u32 *)(addr + 2); - /* e8 0c 78 56 34 12 jmp.d8 +12 */ + /* e8 0c 88 a9 cb ed jmp.d8 +12 */ if (p[0] == JMP8_INSN_OPCODE && p[1] == fineibt_caller_jmp) return -*(u32 *)(addr + 2); @@ -1031,7 +1510,7 @@ static int cfi_rand_preamble(s32 *start, s32 *end) void *addr = (void *)s + *s; u32 hash; - hash = decode_preamble_hash(addr); + hash = decode_preamble_hash(addr, NULL); if (WARN(!hash, "no CFI hash found at: %pS %px %*ph\n", addr, addr, 5, addr)) return -EINVAL; @@ -1043,15 +1522,87 @@ static int cfi_rand_preamble(s32 *start, s32 *end) return 0; } +/* + * Inline the bhi-arity 1 case: + * + * __cfi_foo: + * 0: f3 0f 1e fa endbr64 + * 4: 2d 78 56 34 12 sub $0x12345678, %eax + * 9: 49 0f 45 fa cmovne %rax, %rdi + * d: 2e 75 03 jne,pn foo+0x3 + * + * foo: + * 10: 0f 1f 40 <d6> nopl -42(%rax) + * + * Notably, this scheme is incompatible with permissive CFI + * because the CMOVcc is unconditional and RDI will have been + * clobbered. + */ +asm( ".pushsection .rodata \n" + "fineibt_bhi1_start: \n" + " cmovne %rax, %rdi \n" + " cs jne fineibt_bhi1_func + 0x3 \n" + "fineibt_bhi1_func: \n" + " nopl -42(%rax) \n" + "fineibt_bhi1_end: \n" + ".popsection \n" +); + +extern u8 fineibt_bhi1_start[]; +extern u8 fineibt_bhi1_end[]; + +#define fineibt_bhi1_size (fineibt_bhi1_end - fineibt_bhi1_start) + +static void cfi_fineibt_bhi_preamble(void *addr, int arity) +{ + u8 bytes[MAX_INSN_SIZE]; + + if (!arity) + return; + + if (!cfi_warn && arity == 1) { + text_poke_early(addr + fineibt_preamble_bhi, + fineibt_bhi1_start, fineibt_bhi1_size); + return; + } + + /* + * Replace the bytes at fineibt_preamble_bhi with a CALL instruction + * that lines up exactly with the end of the preamble, such that the + * return address will be foo+0. + * + * __cfi_foo: + * 0: f3 0f 1e fa endbr64 + * 4: 2d 78 56 34 12 sub $0x12345678, %eax + * 9: 2e 2e e8 DD DD DD DD cs cs call __bhi_args[arity] + */ + bytes[0] = 0x2e; + bytes[1] = 0x2e; + __text_gen_insn(bytes + 2, CALL_INSN_OPCODE, + addr + fineibt_preamble_bhi + 2, + __bhi_args[arity], CALL_INSN_SIZE); + + text_poke_early(addr + fineibt_preamble_bhi, bytes, 7); +} + static int cfi_rewrite_preamble(s32 *start, s32 *end) { s32 *s; for (s = start; s < end; s++) { void *addr = (void *)s + *s; + int arity; u32 hash; - hash = decode_preamble_hash(addr); + /* + * When the function doesn't start with ENDBR the compiler will + * have determined there are no indirect calls to it and we + * don't need no CFI either. + */ + if (!is_endbr(addr + 16)) + continue; + + hash = decode_preamble_hash(addr, &arity); if (WARN(!hash, "no CFI hash found at: %pS %px %*ph\n", addr, addr, 5, addr)) return -EINVAL; @@ -1059,6 +1610,13 @@ static int cfi_rewrite_preamble(s32 *start, s32 *end) text_poke_early(addr, fineibt_preamble_start, fineibt_preamble_size); WARN_ON(*(u32 *)(addr + fineibt_preamble_hash) != 0x12345678); text_poke_early(addr + fineibt_preamble_hash, &hash, 4); + + WARN_ONCE(!IS_ENABLED(CONFIG_FINEIBT_BHI) && arity, + "kCFI preamble has wrong register at: %pS %*ph\n", + addr, 5, addr); + + if (cfi_bhi) + cfi_fineibt_bhi_preamble(addr, arity); } return 0; @@ -1071,7 +1629,10 @@ static void cfi_rewrite_endbr(s32 *start, s32 *end) for (s = start; s < end; s++) { void *addr = (void *)s + *s; - poison_endbr(addr+16, false); + if (!exact_endbr(addr + 16)) + continue; + + poison_endbr(addr + 16); } } @@ -1095,40 +1656,101 @@ static int cfi_rand_callers(s32 *start, s32 *end) return 0; } +static int emit_paranoid_trampoline(void *addr, struct insn *insn, int reg, u8 *bytes) +{ + u8 *thunk = (void *)__x86_indirect_its_thunk_array[reg] - 2; + +#ifdef CONFIG_MITIGATION_ITS + u8 *tmp = its_allocate_thunk(reg); + if (tmp) + thunk = tmp; +#endif + + return __emit_trampoline(addr, insn, bytes, thunk, thunk); +} + static int cfi_rewrite_callers(s32 *start, s32 *end) { s32 *s; for (s = start; s < end; s++) { void *addr = (void *)s + *s; + struct insn insn; + u8 bytes[20]; u32 hash; + int ret; + u8 op; addr -= fineibt_caller_size; hash = decode_caller_hash(addr); - if (hash) { + if (!hash) + continue; + + if (!cfi_paranoid) { text_poke_early(addr, fineibt_caller_start, fineibt_caller_size); WARN_ON(*(u32 *)(addr + fineibt_caller_hash) != 0x12345678); text_poke_early(addr + fineibt_caller_hash, &hash, 4); + /* rely on apply_retpolines() */ + continue; + } + + /* cfi_paranoid */ + ret = insn_decode_kernel(&insn, addr + fineibt_caller_size); + if (WARN_ON_ONCE(ret < 0)) + continue; + + op = insn.opcode.bytes[0]; + if (op != CALL_INSN_OPCODE && op != JMP32_INSN_OPCODE) { + WARN_ON_ONCE(1); + continue; } - /* rely on apply_retpolines() */ + + memcpy(bytes, fineibt_paranoid_start, fineibt_paranoid_size); + memcpy(bytes + fineibt_caller_hash, &hash, 4); + + if (cpu_wants_indirect_its_thunk_at((unsigned long)addr + fineibt_paranoid_ind, 11)) { + emit_paranoid_trampoline(addr + fineibt_caller_size, + &insn, 11, bytes + fineibt_caller_size); + } else { + int len = fineibt_paranoid_size - fineibt_paranoid_ind; + ret = emit_indirect(op, 11, bytes + fineibt_paranoid_ind, len); + if (WARN_ON_ONCE(ret != len)) + continue; + } + + text_poke_early(addr, bytes, fineibt_paranoid_size); } return 0; } +#define pr_cfi_debug(X...) if (cfi_debug) pr_info(X) + +#define FINEIBT_WARN(_f, _v) \ + WARN_ONCE((_f) != (_v), "FineIBT: " #_f " %ld != %d\n", _f, _v) + static void __apply_fineibt(s32 *start_retpoline, s32 *end_retpoline, s32 *start_cfi, s32 *end_cfi, bool builtin) { int ret; - if (WARN_ONCE(fineibt_preamble_size != 16, - "FineIBT preamble wrong size: %ld", fineibt_preamble_size)) + if (FINEIBT_WARN(fineibt_preamble_size, 20) || + FINEIBT_WARN(fineibt_preamble_bhi + fineibt_bhi1_size, 20) || + FINEIBT_WARN(fineibt_caller_size, 14) || + FINEIBT_WARN(fineibt_paranoid_size, 20)) return; - if (cfi_mode == CFI_DEFAULT) { + if (cfi_mode == CFI_AUTO) { cfi_mode = CFI_KCFI; - if (HAS_KERNEL_IBT && cpu_feature_enabled(X86_FEATURE_IBT)) + if (HAS_KERNEL_IBT && cpu_feature_enabled(X86_FEATURE_IBT)) { + /* + * FRED has much saner context on exception entry and + * is less easy to take advantage of. + */ + if (!cpu_feature_enabled(X86_FEATURE_FRED)) + cfi_paranoid = true; cfi_mode = CFI_FINEIBT; + } } /* @@ -1136,54 +1758,71 @@ static void __apply_fineibt(s32 *start_retpoline, s32 *end_retpoline, * rewrite them. This disables all CFI. If this succeeds but any of the * later stages fails, we're without CFI. */ + pr_cfi_debug("CFI: disabling all indirect call checking\n"); ret = cfi_disable_callers(start_retpoline, end_retpoline); if (ret) goto err; if (cfi_rand) { - if (builtin) + if (builtin) { cfi_seed = get_random_u32(); + cfi_bpf_hash = cfi_rehash(cfi_bpf_hash); + cfi_bpf_subprog_hash = cfi_rehash(cfi_bpf_subprog_hash); + } + pr_cfi_debug("CFI: cfi_seed: 0x%08x\n", cfi_seed); + pr_cfi_debug("CFI: rehashing all preambles\n"); ret = cfi_rand_preamble(start_cfi, end_cfi); if (ret) goto err; + pr_cfi_debug("CFI: rehashing all indirect calls\n"); ret = cfi_rand_callers(start_retpoline, end_retpoline); if (ret) goto err; + } else { + pr_cfi_debug("CFI: rehashing disabled\n"); } switch (cfi_mode) { case CFI_OFF: if (builtin) - pr_info("Disabling CFI\n"); + pr_info("CFI: disabled\n"); return; case CFI_KCFI: + pr_cfi_debug("CFI: re-enabling all indirect call checking\n"); ret = cfi_enable_callers(start_retpoline, end_retpoline); if (ret) goto err; if (builtin) - pr_info("Using kCFI\n"); + pr_info("CFI: Using %sretpoline kCFI\n", + cfi_rand ? "rehashed " : ""); return; case CFI_FINEIBT: + pr_cfi_debug("CFI: adding FineIBT to all preambles\n"); /* place the FineIBT preamble at func()-16 */ ret = cfi_rewrite_preamble(start_cfi, end_cfi); if (ret) goto err; /* rewrite the callers to target func()-16 */ + pr_cfi_debug("CFI: rewriting indirect call sites to use FineIBT\n"); ret = cfi_rewrite_callers(start_retpoline, end_retpoline); if (ret) goto err; /* now that nobody targets func()+0, remove ENDBR there */ + pr_cfi_debug("CFI: removing old endbr insns\n"); cfi_rewrite_endbr(start_cfi, end_cfi); - if (builtin) - pr_info("Using FineIBT CFI\n"); + if (builtin) { + pr_info("Using %sFineIBT%s CFI\n", + cfi_paranoid ? "paranoid " : "", + cfi_bhi ? "+BHI" : ""); + } return; default: @@ -1201,22 +1840,42 @@ static inline void poison_hash(void *addr) static void poison_cfi(void *addr) { + /* + * Compilers manage to be inconsistent with ENDBR vs __cfi prefixes, + * some (static) functions for which they can determine the address + * is never taken do not get a __cfi prefix, but *DO* get an ENDBR. + * + * As such, these functions will get sealed, but we need to be careful + * to not unconditionally scribble the previous function. + */ switch (cfi_mode) { case CFI_FINEIBT: /* + * FineIBT prefix should start with an ENDBR. + */ + if (!is_endbr(addr)) + break; + + /* * __cfi_\func: - * osp nopl (%rax) - * subl $0, %r10d - * jz 1f - * ud2 - * 1: nop + * nopl -42(%rax) + * sub $0, %eax + * jne \func+3 + * \func: + * nopl -42(%rax) */ - poison_endbr(addr, false); + poison_endbr(addr); poison_hash(addr + fineibt_preamble_hash); break; case CFI_KCFI: /* + * kCFI prefix should start with a valid hash. + */ + if (!decode_preamble_hash(addr, NULL)) + break; + + /* * __cfi_\func: * movl $0, %eax * .skip 11, 0x90 @@ -1229,18 +1888,168 @@ static void poison_cfi(void *addr) } } -#else +#define fineibt_prefix_size (fineibt_preamble_size - ENDBR_INSN_SIZE) + +/* + * When regs->ip points to a 0xD6 byte in the FineIBT preamble, + * return true and fill out target and type. + * + * We check the preamble by checking for the ENDBR instruction relative to the + * UDB instruction. + */ +static bool decode_fineibt_preamble(struct pt_regs *regs, unsigned long *target, u32 *type) +{ + unsigned long addr = regs->ip - fineibt_preamble_ud; + u32 hash; + + if (!exact_endbr((void *)addr)) + return false; + + *target = addr + fineibt_prefix_size; + + __get_kernel_nofault(&hash, addr + fineibt_preamble_hash, u32, Efault); + *type = (u32)regs->ax + hash; + + /* + * Since regs->ip points to the middle of an instruction; it cannot + * continue with the normal fixup. + */ + regs->ip = *target; + + return true; + +Efault: + return false; +} + +/* + * regs->ip points to one of the UD2 in __bhi_args[]. + */ +static bool decode_fineibt_bhi(struct pt_regs *regs, unsigned long *target, u32 *type) +{ + unsigned long addr; + u32 hash; + + if (!cfi_bhi) + return false; + + if (regs->ip < (unsigned long)__bhi_args || + regs->ip >= (unsigned long)__bhi_args_end) + return false; + + /* + * Fetch the return address from the stack, this points to the + * FineIBT preamble. Since the CALL instruction is in the 5 last + * bytes of the preamble, the return address is in fact the target + * address. + */ + __get_kernel_nofault(&addr, regs->sp, unsigned long, Efault); + *target = addr; + + addr -= fineibt_prefix_size; + if (!exact_endbr((void *)addr)) + return false; + + __get_kernel_nofault(&hash, addr + fineibt_preamble_hash, u32, Efault); + *type = (u32)regs->ax + hash; + + /* + * The UD2 sites are constructed with a RET immediately following, + * as such the non-fatal case can use the regular fixup. + */ + return true; + +Efault: + return false; +} + +static bool is_paranoid_thunk(unsigned long addr) +{ + u32 thunk; + + __get_kernel_nofault(&thunk, (u32 *)addr, u32, Efault); + return (thunk & 0x00FFFFFF) == 0xfd75d6; + +Efault: + return false; +} + +/* + * regs->ip points to a LOCK Jcc.d8 instruction from the fineibt_paranoid_start[] + * sequence, or to UDB + Jcc.d8 for cfi_paranoid + ITS thunk. + */ +static bool decode_fineibt_paranoid(struct pt_regs *regs, unsigned long *target, u32 *type) +{ + unsigned long addr = regs->ip - fineibt_paranoid_ud; + + if (!cfi_paranoid) + return false; + + if (is_cfi_trap(addr + fineibt_caller_size - LEN_UD2)) { + *target = regs->r11 + fineibt_prefix_size; + *type = regs->ax; + + /* + * Since the trapping instruction is the exact, but LOCK prefixed, + * Jcc.d8 that got us here, the normal fixup will work. + */ + return true; + } + + /* + * The cfi_paranoid + ITS thunk combination results in: + * + * 0: b8 78 56 34 12 mov $0x12345678, %eax + * 5: 41 3b 43 f7 cmp -11(%r11), %eax + * a: 2e 3d 8d 5b f0 cs lea -0x10(%r11), %r11 + * e: 2e e8 XX XX XX XX cs call __x86_indirect_paranoid_thunk_r11 + * + * Where the paranoid_thunk looks like: + * + * 1d: <d6> udb + * __x86_indirect_paranoid_thunk_r11: + * 1e: 75 fd jne 1d + * __x86_indirect_its_thunk_r11: + * 20: 41 ff eb jmp *%r11 + * 23: cc int3 + * + */ + if (is_paranoid_thunk(regs->ip)) { + *target = regs->r11 + fineibt_prefix_size; + *type = regs->ax; + + regs->ip = *target; + return true; + } + + return false; +} + +bool decode_fineibt_insn(struct pt_regs *regs, unsigned long *target, u32 *type) +{ + if (decode_fineibt_paranoid(regs, target, type)) + return true; + + if (decode_fineibt_bhi(regs, target, type)) + return true; + + return decode_fineibt_preamble(regs, target, type); +} + +#else /* !CONFIG_FINEIBT: */ static void __apply_fineibt(s32 *start_retpoline, s32 *end_retpoline, s32 *start_cfi, s32 *end_cfi, bool builtin) { + if (IS_ENABLED(CONFIG_CFI) && builtin) + pr_info("CFI: Using standard kCFI\n"); } #ifdef CONFIG_X86_KERNEL_IBT static void poison_cfi(void *addr) { } #endif -#endif +#endif /* !CONFIG_FINEIBT */ void apply_fineibt(s32 *start_retpoline, s32 *end_retpoline, s32 *start_cfi, s32 *end_cfi) @@ -1403,46 +2212,6 @@ int alternatives_text_reserved(void *start, void *end) } #endif /* CONFIG_SMP */ -#ifdef CONFIG_PARAVIRT - -/* Use this to add nops to a buffer, then text_poke the whole buffer. */ -static void __init_or_module add_nops(void *insns, unsigned int len) -{ - while (len > 0) { - unsigned int noplen = len; - if (noplen > ASM_NOP_MAX) - noplen = ASM_NOP_MAX; - memcpy(insns, x86_nops[noplen], noplen); - insns += noplen; - len -= noplen; - } -} - -void __init_or_module apply_paravirt(struct paravirt_patch_site *start, - struct paravirt_patch_site *end) -{ - struct paravirt_patch_site *p; - char insn_buff[MAX_PATCH_LEN]; - - for (p = start; p < end; p++) { - unsigned int used; - - BUG_ON(p->len > MAX_PATCH_LEN); - /* prep the buffer with the original instructions */ - memcpy(insn_buff, p->instr, p->len); - used = paravirt_patch(p->type, insn_buff, (unsigned long)p->instr, p->len); - - BUG_ON(used > p->len); - - /* Pad the rest with nops */ - add_nops(insn_buff + used, p->len - used); - text_poke_early(p->instr, insn_buff, p->len); - } -} -extern struct paravirt_patch_site __start_parainstructions[], - __stop_parainstructions[]; -#endif /* CONFIG_PARAVIRT */ - /* * Self-test for the INT3 based CALL emulation code. * @@ -1454,21 +2223,34 @@ extern struct paravirt_patch_site __start_parainstructions[], * See entry_{32,64}.S for more details. */ -/* - * We define the int3_magic() function in assembly to control the calling - * convention such that we can 'call' it from assembly. - */ +extern void int3_selftest_asm(unsigned int *ptr); + +asm ( +" .pushsection .init.text, \"ax\", @progbits\n" +" .type int3_selftest_asm, @function\n" +"int3_selftest_asm:\n" + ANNOTATE_NOENDBR "\n" + /* + * INT3 padded with NOP to CALL_INSN_SIZE. The INT3 triggers an + * exception, then the int3_exception_nb notifier emulates a call to + * int3_selftest_callee(). + */ +" int3; nop; nop; nop; nop\n" + ASM_RET +" .size int3_selftest_asm, . - int3_selftest_asm\n" +" .popsection\n" +); -extern void int3_magic(unsigned int *ptr); /* defined in asm */ +extern void int3_selftest_callee(unsigned int *ptr); asm ( " .pushsection .init.text, \"ax\", @progbits\n" -" .type int3_magic, @function\n" -"int3_magic:\n" - ANNOTATE_NOENDBR -" movl $1, (%" _ASM_ARG1 ")\n" +" .type int3_selftest_callee, @function\n" +"int3_selftest_callee:\n" + ANNOTATE_NOENDBR "\n" +" movl $0x1234, (%" _ASM_ARG1 ")\n" ASM_RET -" .size int3_magic, .-int3_magic\n" +" .size int3_selftest_callee, . - int3_selftest_callee\n" " .popsection\n" ); @@ -1477,7 +2259,7 @@ extern void int3_selftest_ip(void); /* defined in asm below */ static int __init int3_exception_notify(struct notifier_block *self, unsigned long val, void *data) { - unsigned long selftest = (unsigned long)&int3_selftest_ip; + unsigned long selftest = (unsigned long)&int3_selftest_asm; struct die_args *args = data; struct pt_regs *regs = args->regs; @@ -1492,7 +2274,7 @@ int3_exception_notify(struct notifier_block *self, unsigned long val, void *data if (regs->ip - INT3_INSN_SIZE != selftest) return NOTIFY_DONE; - int3_emulate_call(regs, (unsigned long)&int3_magic); + int3_emulate_call(regs, (unsigned long)&int3_selftest_callee); return NOTIFY_STOP; } @@ -1508,19 +2290,11 @@ static noinline void __init int3_selftest(void) BUG_ON(register_die_notifier(&int3_exception_nb)); /* - * Basically: int3_magic(&val); but really complicated :-) - * - * INT3 padded with NOP to CALL_INSN_SIZE. The int3_exception_nb - * notifier above will emulate CALL for us. + * Basically: int3_selftest_callee(&val); but really complicated :-) */ - asm volatile ("int3_selftest_ip:\n\t" - ANNOTATE_NOENDBR - " int3; nop; nop; nop; nop\n\t" - : ASM_CALL_CONSTRAINT - : __ASM_SEL_RAW(a, D) (&val) - : "memory"); + int3_selftest_asm(&val); - BUG_ON(val != 1); + BUG_ON(val != 0x1234); unregister_die_notifier(&int3_exception_nb); } @@ -1536,7 +2310,7 @@ __visible noinline void __init __alt_reloc_selftest(void *arg) static noinline void __init alt_reloc_selftest(void) { /* - * Tests apply_relocation(). + * Tests text_poke_apply_relocation(). * * This has a relative immediate (CALL) in a place other than the first * instruction and additionally on x86_64 we get a RIP-relative LEA: @@ -1549,7 +2323,7 @@ static noinline void __init alt_reloc_selftest(void) */ asm_inline volatile ( ALTERNATIVE("", "lea %[mem], %%" _ASM_ARG1 "; call __alt_reloc_selftest;", X86_FEATURE_ALWAYS) - : /* output */ + : ASM_CALL_CONSTRAINT : [mem] "m" (__alt_reloc_selftest_addr) : _ASM_ARG1 ); @@ -1557,6 +2331,8 @@ static noinline void __init alt_reloc_selftest(void) void __init alternative_instructions(void) { + u64 ibt; + int3_selftest(); /* @@ -1578,30 +2354,17 @@ void __init alternative_instructions(void) */ /* - * Paravirt patching and alternative patching can be combined to - * replace a function call with a short direct code sequence (e.g. - * by setting a constant return value instead of doing that in an - * external function). - * In order to make this work the following sequence is required: - * 1. set (artificial) features depending on used paravirt - * functions which can later influence alternative patching - * 2. apply paravirt patching (generally replacing an indirect - * function call with a direct one) - * 3. apply alternative patching (e.g. replacing a direct function - * call with a custom code sequence) - * Doing paravirt patching after alternative patching would clobber - * the optimization of the custom code with a function call again. + * Make sure to set (artificial) features depending on used paravirt + * functions which can later influence alternative patching. */ paravirt_set_cap(); - /* - * First patch paravirt functions, such that we overwrite the indirect - * call with the direct call. - */ - apply_paravirt(__parainstructions, __parainstructions_end); + /* Keep CET-IBT disabled until caller/callee are patched */ + ibt = ibt_save(/*disable*/ true); __apply_fineibt(__retpoline_sites, __retpoline_sites_end, __cfi_sites, __cfi_sites_end, true); + cfi_debug = false; /* * Rewrite the retpolines, must be done before alternatives since @@ -1610,23 +2373,23 @@ void __init alternative_instructions(void) apply_retpolines(__retpoline_sites, __retpoline_sites_end); apply_returns(__return_sites, __return_sites_end); - /* - * Then patch alternatives, such that those paravirt calls that are in - * alternatives can be overwritten by their immediate fragments. - */ - apply_alternatives(__alt_instructions, __alt_instructions_end); + its_fini_core(); /* - * Now all calls are established. Apply the call thunks if - * required. + * Adjust all CALL instructions to point to func()-10, including + * those in .altinstr_replacement. */ callthunks_patch_builtin_calls(); + apply_alternatives(__alt_instructions, __alt_instructions_end); + /* * Seal all functions that do not have their address taken. */ apply_seal_endbr(__ibt_endbr_seal, __ibt_endbr_seal_end); + ibt_restore(ibt); + #ifdef CONFIG_SMP /* Patch to UP if other cpus not imminent. */ if (!noreplace_smp && (num_present_cpus() == 1 || setup_max_cpus <= 1)) { @@ -1677,8 +2440,8 @@ void __init_or_module text_poke_early(void *addr, const void *opcode, } else { local_irq_save(flags); memcpy(addr, opcode, len); - local_irq_restore(flags); sync_core(); + local_irq_restore(flags); /* * Could also do a CLFLUSH here to speed up CPU recovery; but @@ -1687,83 +2450,33 @@ void __init_or_module text_poke_early(void *addr, const void *opcode, } } -typedef struct { - struct mm_struct *mm; -} temp_mm_state_t; +__ro_after_init struct mm_struct *text_poke_mm; +__ro_after_init unsigned long text_poke_mm_addr; /* - * Using a temporary mm allows to set temporary mappings that are not accessible - * by other CPUs. Such mappings are needed to perform sensitive memory writes - * that override the kernel memory protections (e.g., W^X), without exposing the - * temporary page-table mappings that are required for these write operations to - * other CPUs. Using a temporary mm also allows to avoid TLB shootdowns when the - * mapping is torn down. + * Text poking creates and uses a mapping in the lower half of the + * address space. Relax LASS enforcement when accessing the poking + * address. * - * Context: The temporary mm needs to be used exclusively by a single core. To - * harden security IRQs must be disabled while the temporary mm is - * loaded, thereby preventing interrupt handler bugs from overriding - * the kernel memory protection. + * objtool enforces a strict policy of "no function calls within AC=1 + * regions". Adhere to the policy by using inline versions of + * memcpy()/memset() that will never result in a function call. */ -static inline temp_mm_state_t use_temporary_mm(struct mm_struct *mm) -{ - temp_mm_state_t temp_state; - - lockdep_assert_irqs_disabled(); - - /* - * Make sure not to be in TLB lazy mode, as otherwise we'll end up - * with a stale address space WITHOUT being in lazy mode after - * restoring the previous mm. - */ - if (this_cpu_read(cpu_tlbstate_shared.is_lazy)) - leave_mm(smp_processor_id()); - - temp_state.mm = this_cpu_read(cpu_tlbstate.loaded_mm); - switch_mm_irqs_off(NULL, mm, current); - - /* - * If breakpoints are enabled, disable them while the temporary mm is - * used. Userspace might set up watchpoints on addresses that are used - * in the temporary mm, which would lead to wrong signals being sent or - * crashes. - * - * Note that breakpoints are not disabled selectively, which also causes - * kernel breakpoints (e.g., perf's) to be disabled. This might be - * undesirable, but still seems reasonable as the code that runs in the - * temporary mm should be short. - */ - if (hw_breakpoint_active()) - hw_breakpoint_disable(); - - return temp_state; -} - -static inline void unuse_temporary_mm(temp_mm_state_t prev_state) -{ - lockdep_assert_irqs_disabled(); - switch_mm_irqs_off(NULL, prev_state.mm, current); - - /* - * Restore the breakpoints if they were disabled before the temporary mm - * was loaded. - */ - if (hw_breakpoint_active()) - hw_breakpoint_restore(); -} - -__ro_after_init struct mm_struct *poking_mm; -__ro_after_init unsigned long poking_addr; static void text_poke_memcpy(void *dst, const void *src, size_t len) { - memcpy(dst, src, len); + lass_stac(); + __inline_memcpy(dst, src, len); + lass_clac(); } static void text_poke_memset(void *dst, const void *src, size_t len) { int c = *(const int *)src; - memset(dst, c, len); + lass_stac(); + __inline_memset(dst, c, len); + lass_clac(); } typedef void text_poke_f(void *dst, const void *src, size_t len); @@ -1772,7 +2485,7 @@ static void *__text_poke(text_poke_f func, void *addr, const void *src, size_t l { bool cross_page_boundary = offset_in_page(addr) + len > PAGE_SIZE; struct page *pages[2] = {NULL}; - temp_mm_state_t prev; + struct mm_struct *prev_mm; unsigned long flags; pte_t pte, *ptep; spinlock_t *ptl; @@ -1809,7 +2522,7 @@ static void *__text_poke(text_poke_f func, void *addr, const void *src, size_t l /* * The lock is not really needed, but this allows to avoid open-coding. */ - ptep = get_locked_pte(poking_mm, poking_addr, &ptl); + ptep = get_locked_pte(text_poke_mm, text_poke_mm_addr, &ptl); /* * This must not fail; preallocated in poking_init(). @@ -1819,21 +2532,21 @@ static void *__text_poke(text_poke_f func, void *addr, const void *src, size_t l local_irq_save(flags); pte = mk_pte(pages[0], pgprot); - set_pte_at(poking_mm, poking_addr, ptep, pte); + set_pte_at(text_poke_mm, text_poke_mm_addr, ptep, pte); if (cross_page_boundary) { pte = mk_pte(pages[1], pgprot); - set_pte_at(poking_mm, poking_addr + PAGE_SIZE, ptep + 1, pte); + set_pte_at(text_poke_mm, text_poke_mm_addr + PAGE_SIZE, ptep + 1, pte); } /* * Loading the temporary mm behaves as a compiler barrier, which * guarantees that the PTE will be set at the time memcpy() is done. */ - prev = use_temporary_mm(poking_mm); + prev_mm = use_temporary_mm(text_poke_mm); kasan_disable_current(); - func((u8 *)poking_addr + offset_in_page(addr), src, len); + func((u8 *)text_poke_mm_addr + offset_in_page(addr), src, len); kasan_enable_current(); /* @@ -1842,22 +2555,22 @@ static void *__text_poke(text_poke_f func, void *addr, const void *src, size_t l */ barrier(); - pte_clear(poking_mm, poking_addr, ptep); + pte_clear(text_poke_mm, text_poke_mm_addr, ptep); if (cross_page_boundary) - pte_clear(poking_mm, poking_addr + PAGE_SIZE, ptep + 1); + pte_clear(text_poke_mm, text_poke_mm_addr + PAGE_SIZE, ptep + 1); /* * Loading the previous page-table hierarchy requires a serializing * instruction that already allows the core to see the updated version. * Xen-PV is assumed to serialize execution in a similar manner. */ - unuse_temporary_mm(prev); + unuse_temporary_mm(prev_mm); /* * Flushing the TLB might involve IPIs, which would require enabled * IRQs, but not if the mm is not used, as it is in this point. */ - flush_tlb_mm_range(poking_mm, poking_addr, poking_addr + + flush_tlb_mm_range(text_poke_mm, text_poke_mm_addr, text_poke_mm_addr + (cross_page_boundary ? 2 : 1) * PAGE_SIZE, PAGE_SHIFT, false); @@ -1888,7 +2601,7 @@ static void *__text_poke(text_poke_f func, void *addr, const void *src, size_t l * Note that the caller must ensure that if the modified code is part of a * module, the module would not be removed during poking. This can be achieved * by registering a module notifier, and ordering module removal and patching - * trough a mutex. + * through a mutex. */ void *text_poke(void *addr, const void *opcode, size_t len) { @@ -1993,7 +2706,7 @@ static void do_sync_core(void *info) sync_core(); } -void text_poke_sync(void) +void smp_text_poke_sync_each_cpu(void) { on_each_cpu(do_sync_core, NULL, 1); } @@ -2003,64 +2716,66 @@ void text_poke_sync(void) * this thing. When len == 6 everything is prefixed with 0x0f and we map * opcode to Jcc.d8, using len to distinguish. */ -struct text_poke_loc { +struct smp_text_poke_loc { /* addr := _stext + rel_addr */ s32 rel_addr; s32 disp; u8 len; u8 opcode; - const u8 text[POKE_MAX_OPCODE_SIZE]; - /* see text_poke_bp_batch() */ + const u8 text[TEXT_POKE_MAX_OPCODE_SIZE]; + /* see smp_text_poke_batch_finish() */ u8 old; }; -struct bp_patching_desc { - struct text_poke_loc *vec; +#define TEXT_POKE_ARRAY_MAX (PAGE_SIZE / sizeof(struct smp_text_poke_loc)) + +static struct smp_text_poke_array { + struct smp_text_poke_loc vec[TEXT_POKE_ARRAY_MAX]; int nr_entries; - atomic_t refs; -}; +} text_poke_array; -static struct bp_patching_desc bp_desc; +static DEFINE_PER_CPU(atomic_t, text_poke_array_refs); -static __always_inline -struct bp_patching_desc *try_get_desc(void) +/* + * These four __always_inline annotations imply noinstr, necessary + * due to smp_text_poke_int3_handler() being noinstr: + */ + +static __always_inline bool try_get_text_poke_array(void) { - struct bp_patching_desc *desc = &bp_desc; + atomic_t *refs = this_cpu_ptr(&text_poke_array_refs); - if (!raw_atomic_inc_not_zero(&desc->refs)) - return NULL; + if (!raw_atomic_inc_not_zero(refs)) + return false; - return desc; + return true; } -static __always_inline void put_desc(void) +static __always_inline void put_text_poke_array(void) { - struct bp_patching_desc *desc = &bp_desc; + atomic_t *refs = this_cpu_ptr(&text_poke_array_refs); smp_mb__before_atomic(); - raw_atomic_dec(&desc->refs); + raw_atomic_dec(refs); } -static __always_inline void *text_poke_addr(struct text_poke_loc *tp) +static __always_inline void *text_poke_addr(const struct smp_text_poke_loc *tpl) { - return _stext + tp->rel_addr; + return _stext + tpl->rel_addr; } -static __always_inline int patch_cmp(const void *key, const void *elt) +static __always_inline int patch_cmp(const void *tpl_a, const void *tpl_b) { - struct text_poke_loc *tp = (struct text_poke_loc *) elt; - - if (key < text_poke_addr(tp)) + if (tpl_a < text_poke_addr(tpl_b)) return -1; - if (key > text_poke_addr(tp)) + if (tpl_a > text_poke_addr(tpl_b)) return 1; return 0; } -noinstr int poke_int3_handler(struct pt_regs *regs) +noinstr int smp_text_poke_int3_handler(struct pt_regs *regs) { - struct bp_patching_desc *desc; - struct text_poke_loc *tp; + struct smp_text_poke_loc *tpl; int ret = 0; void *ip; @@ -2069,41 +2784,40 @@ noinstr int poke_int3_handler(struct pt_regs *regs) /* * Having observed our INT3 instruction, we now must observe - * bp_desc with non-zero refcount: + * text_poke_array with non-zero refcount: * - * bp_desc.refs = 1 INT3 - * WMB RMB - * write INT3 if (bp_desc.refs != 0) + * text_poke_array_refs = 1 INT3 + * WMB RMB + * write INT3 if (text_poke_array_refs != 0) */ smp_rmb(); - desc = try_get_desc(); - if (!desc) + if (!try_get_text_poke_array()) return 0; /* - * Discount the INT3. See text_poke_bp_batch(). + * Discount the INT3. See smp_text_poke_batch_finish(). */ ip = (void *) regs->ip - INT3_INSN_SIZE; /* * Skip the binary search if there is a single member in the vector. */ - if (unlikely(desc->nr_entries > 1)) { - tp = __inline_bsearch(ip, desc->vec, desc->nr_entries, - sizeof(struct text_poke_loc), + if (unlikely(text_poke_array.nr_entries > 1)) { + tpl = __inline_bsearch(ip, text_poke_array.vec, text_poke_array.nr_entries, + sizeof(struct smp_text_poke_loc), patch_cmp); - if (!tp) + if (!tpl) goto out_put; } else { - tp = desc->vec; - if (text_poke_addr(tp) != ip) + tpl = text_poke_array.vec; + if (text_poke_addr(tpl) != ip) goto out_put; } - ip += tp->len; + ip += tpl->len; - switch (tp->opcode) { + switch (tpl->opcode) { case INT3_INSN_OPCODE: /* * Someone poked an explicit INT3, they'll want to handle it, @@ -2116,16 +2830,16 @@ noinstr int poke_int3_handler(struct pt_regs *regs) break; case CALL_INSN_OPCODE: - int3_emulate_call(regs, (long)ip + tp->disp); + int3_emulate_call(regs, (long)ip + tpl->disp); break; case JMP32_INSN_OPCODE: case JMP8_INSN_OPCODE: - int3_emulate_jmp(regs, (long)ip + tp->disp); + int3_emulate_jmp(regs, (long)ip + tpl->disp); break; case 0x70 ... 0x7f: /* Jcc */ - int3_emulate_jcc(regs, tp->opcode & 0xf, (long)ip, tp->disp); + int3_emulate_jcc(regs, tpl->opcode & 0xf, (long)ip, tpl->disp); break; default: @@ -2135,51 +2849,50 @@ noinstr int poke_int3_handler(struct pt_regs *regs) ret = 1; out_put: - put_desc(); + put_text_poke_array(); return ret; } -#define TP_VEC_MAX (PAGE_SIZE / sizeof(struct text_poke_loc)) -static struct text_poke_loc tp_vec[TP_VEC_MAX]; -static int tp_vec_nr; - /** - * text_poke_bp_batch() -- update instructions on live kernel on SMP - * @tp: vector of instructions to patch - * @nr_entries: number of entries in the vector + * smp_text_poke_batch_finish() -- update instructions on live kernel on SMP + * + * Input state: + * text_poke_array.vec: vector of instructions to patch + * text_poke_array.nr_entries: number of entries in the vector * - * Modify multi-byte instruction by using int3 breakpoint on SMP. - * We completely avoid stop_machine() here, and achieve the - * synchronization using int3 breakpoint. + * Modify multi-byte instructions by using INT3 breakpoints on SMP. + * We completely avoid using stop_machine() here, and achieve the + * synchronization using INT3 breakpoints and SMP cross-calls. * * The way it is done: * - For each entry in the vector: - * - add a int3 trap to the address that will be patched - * - sync cores + * - add an INT3 trap to the address that will be patched + * - SMP sync all CPUs * - For each entry in the vector: * - update all but the first byte of the patched range - * - sync cores + * - SMP sync all CPUs * - For each entry in the vector: - * - replace the first byte (int3) by the first byte of + * - replace the first byte (INT3) by the first byte of the * replacing opcode - * - sync cores + * - SMP sync all CPUs */ -static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries) +void smp_text_poke_batch_finish(void) { unsigned char int3 = INT3_INSN_OPCODE; unsigned int i; int do_sync; - lockdep_assert_held(&text_mutex); + if (!text_poke_array.nr_entries) + return; - bp_desc.vec = tp; - bp_desc.nr_entries = nr_entries; + lockdep_assert_held(&text_mutex); /* - * Corresponds to the implicit memory barrier in try_get_desc() to - * ensure reading a non-zero refcount provides up to date bp_desc data. + * Corresponds to the implicit memory barrier in try_get_text_poke_array() to + * ensure reading a non-zero refcount provides up to date text_poke_array data. */ - atomic_set_release(&bp_desc.refs, 1); + for_each_possible_cpu(i) + atomic_set_release(per_cpu_ptr(&text_poke_array_refs, i), 1); /* * Function tracing can enable thousands of places that need to be @@ -2192,33 +2905,33 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries cond_resched(); /* - * Corresponding read barrier in int3 notifier for making sure the - * nr_entries and handler are correctly ordered wrt. patching. + * Corresponding read barrier in INT3 notifier for making sure the + * text_poke_array.nr_entries and handler are correctly ordered wrt. patching. */ smp_wmb(); /* - * First step: add a int3 trap to the address that will be patched. + * First step: add a INT3 trap to the address that will be patched. */ - for (i = 0; i < nr_entries; i++) { - tp[i].old = *(u8 *)text_poke_addr(&tp[i]); - text_poke(text_poke_addr(&tp[i]), &int3, INT3_INSN_SIZE); + for (i = 0; i < text_poke_array.nr_entries; i++) { + text_poke_array.vec[i].old = *(u8 *)text_poke_addr(&text_poke_array.vec[i]); + text_poke(text_poke_addr(&text_poke_array.vec[i]), &int3, INT3_INSN_SIZE); } - text_poke_sync(); + smp_text_poke_sync_each_cpu(); /* * Second step: update all but the first byte of the patched range. */ - for (do_sync = 0, i = 0; i < nr_entries; i++) { - u8 old[POKE_MAX_OPCODE_SIZE+1] = { tp[i].old, }; - u8 _new[POKE_MAX_OPCODE_SIZE+1]; - const u8 *new = tp[i].text; - int len = tp[i].len; + for (do_sync = 0, i = 0; i < text_poke_array.nr_entries; i++) { + u8 old[TEXT_POKE_MAX_OPCODE_SIZE+1] = { text_poke_array.vec[i].old, }; + u8 _new[TEXT_POKE_MAX_OPCODE_SIZE+1]; + const u8 *new = text_poke_array.vec[i].text; + int len = text_poke_array.vec[i].len; if (len - INT3_INSN_SIZE > 0) { memcpy(old + INT3_INSN_SIZE, - text_poke_addr(&tp[i]) + INT3_INSN_SIZE, + text_poke_addr(&text_poke_array.vec[i]) + INT3_INSN_SIZE, len - INT3_INSN_SIZE); if (len == 6) { @@ -2227,7 +2940,7 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries new = _new; } - text_poke(text_poke_addr(&tp[i]) + INT3_INSN_SIZE, + text_poke(text_poke_addr(&text_poke_array.vec[i]) + INT3_INSN_SIZE, new + INT3_INSN_SIZE, len - INT3_INSN_SIZE); @@ -2258,7 +2971,7 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries * The old instruction is recorded so that the event can be * processed forwards or backwards. */ - perf_event_text_poke(text_poke_addr(&tp[i]), old, len, new, len); + perf_event_text_poke(text_poke_addr(&text_poke_array.vec[i]), old, len, new, len); } if (do_sync) { @@ -2267,63 +2980,79 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries * not necessary and we'd be safe even without it. But * better safe than sorry (plus there's not only Intel). */ - text_poke_sync(); + smp_text_poke_sync_each_cpu(); } /* - * Third step: replace the first byte (int3) by the first byte of + * Third step: replace the first byte (INT3) by the first byte of the * replacing opcode. */ - for (do_sync = 0, i = 0; i < nr_entries; i++) { - u8 byte = tp[i].text[0]; + for (do_sync = 0, i = 0; i < text_poke_array.nr_entries; i++) { + u8 byte = text_poke_array.vec[i].text[0]; - if (tp[i].len == 6) + if (text_poke_array.vec[i].len == 6) byte = 0x0f; if (byte == INT3_INSN_OPCODE) continue; - text_poke(text_poke_addr(&tp[i]), &byte, INT3_INSN_SIZE); + text_poke(text_poke_addr(&text_poke_array.vec[i]), &byte, INT3_INSN_SIZE); do_sync++; } if (do_sync) - text_poke_sync(); + smp_text_poke_sync_each_cpu(); /* * Remove and wait for refs to be zero. + * + * Notably, if after step-3 above the INT3 got removed, then the + * smp_text_poke_sync_each_cpu() will have serialized against any running INT3 + * handlers and the below spin-wait will not happen. + * + * IOW. unless the replacement instruction is INT3, this case goes + * unused. */ - if (!atomic_dec_and_test(&bp_desc.refs)) - atomic_cond_read_acquire(&bp_desc.refs, !VAL); + for_each_possible_cpu(i) { + atomic_t *refs = per_cpu_ptr(&text_poke_array_refs, i); + + if (unlikely(!atomic_dec_and_test(refs))) + atomic_cond_read_acquire(refs, !VAL); + } + + /* They are all completed: */ + text_poke_array.nr_entries = 0; } -static void text_poke_loc_init(struct text_poke_loc *tp, void *addr, - const void *opcode, size_t len, const void *emulate) +static void __smp_text_poke_batch_add(void *addr, const void *opcode, size_t len, const void *emulate) { + struct smp_text_poke_loc *tpl; struct insn insn; int ret, i = 0; + tpl = &text_poke_array.vec[text_poke_array.nr_entries++]; + if (len == 6) i = 1; - memcpy((void *)tp->text, opcode+i, len-i); + memcpy((void *)tpl->text, opcode+i, len-i); if (!emulate) emulate = opcode; ret = insn_decode_kernel(&insn, emulate); BUG_ON(ret < 0); - tp->rel_addr = addr - (void *)_stext; - tp->len = len; - tp->opcode = insn.opcode.bytes[0]; + tpl->rel_addr = addr - (void *)_stext; + tpl->len = len; + tpl->opcode = insn.opcode.bytes[0]; if (is_jcc32(&insn)) { /* * Map Jcc.d32 onto Jcc.d8 and use len to distinguish. */ - tp->opcode = insn.opcode.bytes[1] - 0x10; + tpl->opcode = insn.opcode.bytes[1] - 0x10; } - switch (tp->opcode) { + switch (tpl->opcode) { case RET_INSN_OPCODE: case JMP32_INSN_OPCODE: case JMP8_INSN_OPCODE: @@ -2332,14 +3061,14 @@ static void text_poke_loc_init(struct text_poke_loc *tp, void *addr, * next instruction can be padded with INT3. */ for (i = insn.length; i < len; i++) - BUG_ON(tp->text[i] != INT3_INSN_OPCODE); + BUG_ON(tpl->text[i] != INT3_INSN_OPCODE); break; default: BUG_ON(len != insn.length); } - switch (tp->opcode) { + switch (tpl->opcode) { case INT3_INSN_OPCODE: case RET_INSN_OPCODE: break; @@ -2348,21 +3077,21 @@ static void text_poke_loc_init(struct text_poke_loc *tp, void *addr, case JMP32_INSN_OPCODE: case JMP8_INSN_OPCODE: case 0x70 ... 0x7f: /* Jcc */ - tp->disp = insn.immediate.value; + tpl->disp = insn.immediate.value; break; default: /* assume NOP */ switch (len) { case 2: /* NOP2 -- emulate as JMP8+0 */ BUG_ON(memcmp(emulate, x86_nops[len], len)); - tp->opcode = JMP8_INSN_OPCODE; - tp->disp = 0; + tpl->opcode = JMP8_INSN_OPCODE; + tpl->disp = 0; break; case 5: /* NOP5 -- emulate as JMP32+0 */ BUG_ON(memcmp(emulate, x86_nops[len], len)); - tp->opcode = JMP32_INSN_OPCODE; - tp->disp = 0; + tpl->opcode = JMP32_INSN_OPCODE; + tpl->disp = 0; break; default: /* unknown instruction */ @@ -2373,51 +3102,50 @@ static void text_poke_loc_init(struct text_poke_loc *tp, void *addr, } /* - * We hard rely on the tp_vec being ordered; ensure this is so by flushing + * We hard rely on the text_poke_array.vec being ordered; ensure this is so by flushing * early if needed. */ -static bool tp_order_fail(void *addr) +static bool text_poke_addr_ordered(void *addr) { - struct text_poke_loc *tp; + WARN_ON_ONCE(!addr); - if (!tp_vec_nr) - return false; - - if (!addr) /* force */ + if (!text_poke_array.nr_entries) return true; - tp = &tp_vec[tp_vec_nr - 1]; - if ((unsigned long)text_poke_addr(tp) > (unsigned long)addr) - return true; - - return false; -} - -static void text_poke_flush(void *addr) -{ - if (tp_vec_nr == TP_VEC_MAX || tp_order_fail(addr)) { - text_poke_bp_batch(tp_vec, tp_vec_nr); - tp_vec_nr = 0; - } -} + /* + * If the last current entry's address is higher than the + * new entry's address we'd like to add, then ordering + * is violated and we must first flush all pending patching + * requests: + */ + if (text_poke_addr(text_poke_array.vec + text_poke_array.nr_entries-1) > addr) + return false; -void text_poke_finish(void) -{ - text_poke_flush(NULL); + return true; } -void __ref text_poke_queue(void *addr, const void *opcode, size_t len, const void *emulate) +/** + * smp_text_poke_batch_add() -- update instruction on live kernel on SMP, batched + * @addr: address to patch + * @opcode: opcode of new instruction + * @len: length to copy + * @emulate: instruction to be emulated + * + * Add a new instruction to the current queue of to-be-patched instructions + * the kernel maintains. The patching request will not be executed immediately, + * but becomes part of an array of patching requests, optimized for batched + * execution. All pending patching requests will be executed on the next + * smp_text_poke_batch_finish() call. + */ +void __ref smp_text_poke_batch_add(void *addr, const void *opcode, size_t len, const void *emulate) { - struct text_poke_loc *tp; - - text_poke_flush(addr); - - tp = &tp_vec[tp_vec_nr++]; - text_poke_loc_init(tp, addr, opcode, len, emulate); + if (text_poke_array.nr_entries == TEXT_POKE_ARRAY_MAX || !text_poke_addr_ordered(addr)) + smp_text_poke_batch_finish(); + __smp_text_poke_batch_add(addr, opcode, len, emulate); } /** - * text_poke_bp() -- update instructions on live kernel on SMP + * smp_text_poke_single() -- update instruction on live kernel on SMP immediately * @addr: address to patch * @opcode: opcode of new instruction * @len: length to copy @@ -2425,12 +3153,11 @@ void __ref text_poke_queue(void *addr, const void *opcode, size_t len, const voi * * Update a single instruction with the vector in the stack, avoiding * dynamically allocated memory. This function should be used when it is - * not possible to allocate memory. + * not possible to allocate memory for a vector. The single instruction + * is patched in immediately. */ -void __ref text_poke_bp(void *addr, const void *opcode, size_t len, const void *emulate) +void __ref smp_text_poke_single(void *addr, const void *opcode, size_t len, const void *emulate) { - struct text_poke_loc tp; - - text_poke_loc_init(&tp, addr, opcode, len, emulate); - text_poke_bp_batch(&tp, 1); + smp_text_poke_batch_add(addr, opcode, len, emulate); + smp_text_poke_batch_finish(); } |