diff options
Diffstat (limited to 'arch/x86/mm/extable.c')
| -rw-r--r-- | arch/x86/mm/extable.c | 203 |
1 files changed, 178 insertions, 25 deletions
diff --git a/arch/x86/mm/extable.c b/arch/x86/mm/extable.c index dba2197c05c3..2fdc1f1f5adb 100644 --- a/arch/x86/mm/extable.c +++ b/arch/x86/mm/extable.c @@ -6,6 +6,7 @@ #include <xen/xen.h> #include <asm/fpu/api.h> +#include <asm/fred.h> #include <asm/sev.h> #include <asm/traps.h> #include <asm/kdebug.h> @@ -41,6 +42,59 @@ static bool ex_handler_default(const struct exception_table_entry *e, return true; } +/* + * This is the *very* rare case where we do a "load_unaligned_zeropad()" + * and it's a page crosser into a non-existent page. + * + * This happens when we optimistically load a pathname a word-at-a-time + * and the name is less than the full word and the next page is not + * mapped. Typically that only happens for CONFIG_DEBUG_PAGEALLOC. + * + * NOTE! The faulting address is always a 'mov mem,reg' type instruction + * of size 'long', and the exception fixup must always point to right + * after the instruction. + */ +static bool ex_handler_zeropad(const struct exception_table_entry *e, + struct pt_regs *regs, + unsigned long fault_addr) +{ + struct insn insn; + const unsigned long mask = sizeof(long) - 1; + unsigned long offset, addr, next_ip, len; + unsigned long *reg; + + next_ip = ex_fixup_addr(e); + len = next_ip - regs->ip; + if (len > MAX_INSN_SIZE) + return false; + + if (insn_decode(&insn, (void *) regs->ip, len, INSN_MODE_KERN)) + return false; + if (insn.length != len) + return false; + + if (insn.opcode.bytes[0] != 0x8b) + return false; + if (insn.opnd_bytes != sizeof(long)) + return false; + + addr = (unsigned long) insn_get_addr_ref(&insn, regs); + if (addr == ~0ul) + return false; + + offset = addr & mask; + addr = addr & ~mask; + if (fault_addr != addr + sizeof(long)) + return false; + + reg = insn_get_modrm_reg_ptr(&insn, regs); + if (!reg) + return false; + + *reg = *(unsigned long *)addr >> (offset * 8); + return ex_handler_default(e, regs); +} + static bool ex_handler_fault(const struct exception_table_entry *fixup, struct pt_regs *regs, int trapnr) { @@ -57,7 +111,7 @@ static bool ex_handler_sgx(const struct exception_table_entry *fixup, /* * Handler for when we fail to restore a task's FPU state. We should never get - * here because the FPU state of a task using the FPU (task->thread.fpu.state) + * here because the FPU state of a task using the FPU (struct fpu::fpstate) * should always be valid. However, past bugs have allowed userspace to set * reserved bits in the XSAVE area using PTRACE_SETREGSET or sys_rt_sigreturn(). * These caused XRSTOR to fail when switching to the task, leaking the FPU @@ -68,42 +122,62 @@ static bool ex_handler_sgx(const struct exception_table_entry *fixup, static bool ex_handler_fprestore(const struct exception_table_entry *fixup, struct pt_regs *regs) { - regs->ip = ex_fixup_addr(fixup); - WARN_ONCE(1, "Bad FPU state detected at %pB, reinitializing FPU registers.", (void *)instruction_pointer(regs)); fpu_reset_from_exception_fixup(); - return true; + + return ex_handler_default(fixup, regs); } -static bool ex_handler_uaccess(const struct exception_table_entry *fixup, - struct pt_regs *regs, int trapnr) +/* + * On x86-64, we end up being imprecise with 'access_ok()', and allow + * non-canonical user addresses to make the range comparisons simpler, + * and to not have to worry about LAM being enabled. + * + * In fact, we allow up to one page of "slop" at the sign boundary, + * which means that we can do access_ok() by just checking the sign + * of the pointer for the common case of having a small access size. + */ +static bool gp_fault_address_ok(unsigned long fault_address) { - WARN_ONCE(trapnr == X86_TRAP_GP, "General protection fault in user access. Non-canonical address?"); - return ex_handler_default(fixup, regs); +#ifdef CONFIG_X86_64 + /* Is it in the "user space" part of the non-canonical space? */ + if (valid_user_address(fault_address)) + return true; + + /* .. or just above it? */ + fault_address -= PAGE_SIZE; + if (valid_user_address(fault_address)) + return true; +#endif + return false; } -static bool ex_handler_copy(const struct exception_table_entry *fixup, - struct pt_regs *regs, int trapnr) +static bool ex_handler_uaccess(const struct exception_table_entry *fixup, + struct pt_regs *regs, int trapnr, + unsigned long fault_address) { - WARN_ONCE(trapnr == X86_TRAP_GP, "General protection fault in user access. Non-canonical address?"); - return ex_handler_fault(fixup, regs, trapnr); + WARN_ONCE(trapnr == X86_TRAP_GP && !gp_fault_address_ok(fault_address), + "General protection fault in user access. Non-canonical address?"); + return ex_handler_default(fixup, regs); } static bool ex_handler_msr(const struct exception_table_entry *fixup, struct pt_regs *regs, bool wrmsr, bool safe, int reg) { - if (!safe && wrmsr && - pr_warn_once("unchecked MSR access error: WRMSR to 0x%x (tried to write 0x%08x%08x) at rIP: 0x%lx (%pS)\n", - (unsigned int)regs->cx, (unsigned int)regs->dx, - (unsigned int)regs->ax, regs->ip, (void *)regs->ip)) + if (__ONCE_LITE_IF(!safe && wrmsr)) { + pr_warn("unchecked MSR access error: WRMSR to 0x%x (tried to write 0x%08x%08x) at rIP: 0x%lx (%pS)\n", + (unsigned int)regs->cx, (unsigned int)regs->dx, + (unsigned int)regs->ax, regs->ip, (void *)regs->ip); show_stack_regs(regs); + } - if (!safe && !wrmsr && - pr_warn_once("unchecked MSR access error: RDMSR from 0x%x at rIP: 0x%lx (%pS)\n", - (unsigned int)regs->cx, regs->ip, (void *)regs->ip)) + if (__ONCE_LITE_IF(!safe && !wrmsr)) { + pr_warn("unchecked MSR access error: RDMSR from 0x%x at rIP: 0x%lx (%pS)\n", + (unsigned int)regs->cx, regs->ip, (void *)regs->ip); show_stack_regs(regs); + } if (!wrmsr) { /* Pretend that the read succeeded and returned 0. */ @@ -134,12 +208,87 @@ static bool ex_handler_imm_reg(const struct exception_table_entry *fixup, } static bool ex_handler_ucopy_len(const struct exception_table_entry *fixup, - struct pt_regs *regs, int trapnr, int reg, int imm) + struct pt_regs *regs, int trapnr, + unsigned long fault_address, + int reg, int imm) { regs->cx = imm * regs->cx + *pt_regs_nr(regs, reg); - return ex_handler_uaccess(fixup, regs, trapnr); + return ex_handler_uaccess(fixup, regs, trapnr, fault_address); } +#ifdef CONFIG_X86_FRED +static bool ex_handler_eretu(const struct exception_table_entry *fixup, + struct pt_regs *regs, unsigned long error_code) +{ + struct pt_regs *uregs = (struct pt_regs *)(regs->sp - offsetof(struct pt_regs, orig_ax)); + unsigned short ss = uregs->ss; + unsigned short cs = uregs->cs; + + /* + * Move the NMI bit from the invalid stack frame, which caused ERETU + * to fault, to the fault handler's stack frame, thus to unblock NMI + * with the fault handler's ERETS instruction ASAP if NMI is blocked. + */ + regs->fred_ss.nmi = uregs->fred_ss.nmi; + + /* + * Sync event information to uregs, i.e., the ERETU return frame, but + * is it safe to write to the ERETU return frame which is just above + * current event stack frame? + * + * The RSP used by FRED to push a stack frame is not the value in %rsp, + * it is calculated from %rsp with the following 2 steps: + * 1) RSP = %rsp - (IA32_FRED_CONFIG & 0x1c0) // Reserve N*64 bytes + * 2) RSP = RSP & ~0x3f // Align to a 64-byte cache line + * when an event delivery doesn't trigger a stack level change. + * + * Here is an example with N*64 (N=1) bytes reserved: + * + * 64-byte cache line ==> ______________ + * |___Reserved___| + * |__Event_data__| + * |_____SS_______| + * |_____RSP______| + * |_____FLAGS____| + * |_____CS_______| + * |_____IP_______| + * 64-byte cache line ==> |__Error_code__| <== ERETU return frame + * |______________| + * |______________| + * |______________| + * |______________| + * |______________| + * |______________| + * |______________| + * 64-byte cache line ==> |______________| <== RSP after step 1) and 2) + * |___Reserved___| + * |__Event_data__| + * |_____SS_______| + * |_____RSP______| + * |_____FLAGS____| + * |_____CS_______| + * |_____IP_______| + * 64-byte cache line ==> |__Error_code__| <== ERETS return frame + * + * Thus a new FRED stack frame will always be pushed below a previous + * FRED stack frame ((N*64) bytes may be reserved between), and it is + * safe to write to a previous FRED stack frame as they never overlap. + */ + fred_info(uregs)->edata = fred_event_data(regs); + uregs->ssx = regs->ssx; + uregs->fred_ss.ss = ss; + /* The NMI bit was moved away above */ + uregs->fred_ss.nmi = 0; + uregs->csx = regs->csx; + uregs->fred_cs.sl = 0; + uregs->fred_cs.wfe = 0; + uregs->cs = cs; + uregs->orig_ax = error_code; + + return ex_handler_default(fixup, regs); +} +#endif + int ex_get_fixup_type(unsigned long ip) { const struct exception_table_entry *e = search_exception_tables(ip); @@ -183,9 +332,7 @@ int fixup_exception(struct pt_regs *regs, int trapnr, unsigned long error_code, case EX_TYPE_FAULT_MCE_SAFE: return ex_handler_fault(e, regs, trapnr); case EX_TYPE_UACCESS: - return ex_handler_uaccess(e, regs, trapnr); - case EX_TYPE_COPY: - return ex_handler_copy(e, regs, trapnr); + return ex_handler_uaccess(e, regs, trapnr, fault_addr); case EX_TYPE_CLEAR_FS: return ex_handler_clear_fs(e, regs); case EX_TYPE_FPU_RESTORE: @@ -214,7 +361,13 @@ int fixup_exception(struct pt_regs *regs, int trapnr, unsigned long error_code, case EX_TYPE_FAULT_SGX: return ex_handler_sgx(e, regs, trapnr); case EX_TYPE_UCOPY_LEN: - return ex_handler_ucopy_len(e, regs, trapnr, reg, imm); + return ex_handler_ucopy_len(e, regs, trapnr, fault_addr, reg, imm); + case EX_TYPE_ZEROPAD: + return ex_handler_zeropad(e, regs, fault_addr); +#ifdef CONFIG_X86_FRED + case EX_TYPE_ERETU: + return ex_handler_eretu(e, regs, error_code); +#endif } BUG(); } |