From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753582AbbDRP0V (ORCPT ); Sat, 18 Apr 2015 11:26:21 -0400 Received: from mail-wi0-f172.google.com ([209.85.212.172]:37262 "EHLO mail-wi0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753282AbbDRP0T (ORCPT ); Sat, 18 Apr 2015 11:26:19 -0400 Date: Sat, 18 Apr 2015 17:26:13 +0200 From: Ingo Molnar To: Linus Torvalds Cc: linux-kernel@vger.kernel.org, Borislav Petkov , Thomas Gleixner , "H. Peter Anvin" , Andrew Morton Subject: [GIT PULL] x86 fixes Message-ID: <20150418152613.GA4569@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Linus, Please pull the latest x86-urgent-for-linus git tree from: git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86-urgent-for-linus # HEAD: a6dfa128ce5c414ab46b1d690f7a1b8decb8526d config: Enable NEED_DMA_MAP_STATE by default when SWIOTLB is selected This tree includes: - an FPU related crash fix - a ptrace fix (with matching testcase in tools/testing/selftests/) - an x86 Kconfig DMA-config defaults tweak to better avoid non-working drivers out-of-topic modifications in x86-urgent-for-linus: ----------------------------------------------------- tools/testing/selftests/x86/Makefile# 0a15584d7276: x86, selftests: Add single_s tools/testing/selftests/x86/run_x86_tests.sh# 0a15584d7276: x86, selftests: Add single_s tools/testing/selftests/x86/single_step_syscall.c# 0a15584d7276: x86, selftests: Add single_s Thanks, Ingo ------------------> Andy Lutomirski (1): x86, selftests: Add single_step_syscall test Borislav Petkov (1): x86/fpu: Load xsave pointer *after* initialization Konrad Rzeszutek Wilk (1): config: Enable NEED_DMA_MAP_STATE by default when SWIOTLB is selected Oleg Nesterov (1): x86/ptrace: Fix the TIF_FORCED_TF logic in handle_signal() arch/x86/Kconfig | 2 +- arch/x86/kernel/i387.c | 8 +- arch/x86/kernel/signal.c | 22 +-- tools/testing/selftests/x86/Makefile | 2 +- tools/testing/selftests/x86/run_x86_tests.sh | 2 + tools/testing/selftests/x86/single_step_syscall.c | 181 ++++++++++++++++++++++ 6 files changed, 202 insertions(+), 15 deletions(-) create mode 100644 tools/testing/selftests/x86/single_step_syscall.c diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index faff6934c05a..eb79036e3503 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -177,7 +177,7 @@ config SBUS config NEED_DMA_MAP_STATE def_bool y - depends on X86_64 || INTEL_IOMMU || DMA_API_DEBUG + depends on X86_64 || INTEL_IOMMU || DMA_API_DEBUG || SWIOTLB config NEED_SG_DMA_LENGTH def_bool y diff --git a/arch/x86/kernel/i387.c b/arch/x86/kernel/i387.c index 367f39d35e9c..009183276bb7 100644 --- a/arch/x86/kernel/i387.c +++ b/arch/x86/kernel/i387.c @@ -341,7 +341,7 @@ int xstateregs_get(struct task_struct *target, const struct user_regset *regset, unsigned int pos, unsigned int count, void *kbuf, void __user *ubuf) { - struct xsave_struct *xsave = &target->thread.fpu.state->xsave; + struct xsave_struct *xsave; int ret; if (!cpu_has_xsave) @@ -351,6 +351,8 @@ int xstateregs_get(struct task_struct *target, const struct user_regset *regset, if (ret) return ret; + xsave = &target->thread.fpu.state->xsave; + /* * Copy the 48bytes defined by the software first into the xstate * memory layout in the thread struct, so that we can copy the entire @@ -369,7 +371,7 @@ int xstateregs_set(struct task_struct *target, const struct user_regset *regset, unsigned int pos, unsigned int count, const void *kbuf, const void __user *ubuf) { - struct xsave_struct *xsave = &target->thread.fpu.state->xsave; + struct xsave_struct *xsave; int ret; if (!cpu_has_xsave) @@ -379,6 +381,8 @@ int xstateregs_set(struct task_struct *target, const struct user_regset *regset, if (ret) return ret; + xsave = &target->thread.fpu.state->xsave; + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, xsave, 0, -1); /* * mxcsr reserved bits must be masked to zero for security reasons. diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c index 3e581865c8e2..d185bdd95a4b 100644 --- a/arch/x86/kernel/signal.c +++ b/arch/x86/kernel/signal.c @@ -630,7 +630,8 @@ setup_rt_frame(struct ksignal *ksig, struct pt_regs *regs) static void handle_signal(struct ksignal *ksig, struct pt_regs *regs) { - bool failed; + bool stepping, failed; + /* Are we from a system call? */ if (syscall_get_nr(current, regs) >= 0) { /* If so, check system call restarting.. */ @@ -654,12 +655,13 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs) } /* - * If TF is set due to a debugger (TIF_FORCED_TF), clear the TF - * flag so that register information in the sigcontext is correct. + * If TF is set due to a debugger (TIF_FORCED_TF), clear TF now + * so that register information in the sigcontext is correct and + * then notify the tracer before entering the signal handler. */ - if (unlikely(regs->flags & X86_EFLAGS_TF) && - likely(test_and_clear_thread_flag(TIF_FORCED_TF))) - regs->flags &= ~X86_EFLAGS_TF; + stepping = test_thread_flag(TIF_SINGLESTEP); + if (stepping) + user_disable_single_step(current); failed = (setup_rt_frame(ksig, regs) < 0); if (!failed) { @@ -670,10 +672,8 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs) * it might disable possible debug exception from the * signal handler. * - * Clear TF when entering the signal handler, but - * notify any tracer that was single-stepping it. - * The tracer may want to single-step inside the - * handler too. + * Clear TF for the case when it wasn't set by debugger to + * avoid the recursive send_sigtrap() in SIGTRAP handler. */ regs->flags &= ~(X86_EFLAGS_DF|X86_EFLAGS_RF|X86_EFLAGS_TF); /* @@ -682,7 +682,7 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs) if (used_math()) fpu_reset_state(current); } - signal_setup_done(failed, ksig, test_thread_flag(TIF_SINGLESTEP)); + signal_setup_done(failed, ksig, stepping); } #ifdef CONFIG_X86_32 diff --git a/tools/testing/selftests/x86/Makefile b/tools/testing/selftests/x86/Makefile index f0a7918178dd..ddf63569df5a 100644 --- a/tools/testing/selftests/x86/Makefile +++ b/tools/testing/selftests/x86/Makefile @@ -1,6 +1,6 @@ .PHONY: all all_32 all_64 check_build32 clean run_tests -TARGETS_C_BOTHBITS := sigreturn +TARGETS_C_BOTHBITS := sigreturn single_step_syscall BINARIES_32 := $(TARGETS_C_BOTHBITS:%=%_32) BINARIES_64 := $(TARGETS_C_BOTHBITS:%=%_64) diff --git a/tools/testing/selftests/x86/run_x86_tests.sh b/tools/testing/selftests/x86/run_x86_tests.sh index 3d3ec65f3e7c..3fc19b376812 100644 --- a/tools/testing/selftests/x86/run_x86_tests.sh +++ b/tools/testing/selftests/x86/run_x86_tests.sh @@ -3,9 +3,11 @@ # This is deliberately minimal. IMO kselftests should provide a standard # script here. ./sigreturn_32 || exit 1 +./single_step_syscall_32 || exit 1 if [[ "$uname -p" -eq "x86_64" ]]; then ./sigreturn_64 || exit 1 + ./single_step_syscall_64 || exit 1 fi exit 0 diff --git a/tools/testing/selftests/x86/single_step_syscall.c b/tools/testing/selftests/x86/single_step_syscall.c new file mode 100644 index 000000000000..50c26358e8b7 --- /dev/null +++ b/tools/testing/selftests/x86/single_step_syscall.c @@ -0,0 +1,181 @@ +/* + * single_step_syscall.c - single-steps various x86 syscalls + * Copyright (c) 2014-2015 Andrew Lutomirski + * + * This program is free software; you can redistribute it and/or modify + * it under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * This is a very simple series of tests that makes system calls with + * the TF flag set. This exercises some nasty kernel code in the + * SYSENTER case: SYSENTER does not clear TF, so SYSENTER with TF set + * immediately issues #DB from CPL 0. This requires special handling in + * the kernel. + */ + +#define _GNU_SOURCE + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *), + int flags) +{ + struct sigaction sa; + memset(&sa, 0, sizeof(sa)); + sa.sa_sigaction = handler; + sa.sa_flags = SA_SIGINFO | flags; + sigemptyset(&sa.sa_mask); + if (sigaction(sig, &sa, 0)) + err(1, "sigaction"); +} + +static volatile sig_atomic_t sig_traps; + +#ifdef __x86_64__ +# define REG_IP REG_RIP +# define WIDTH "q" +#else +# define REG_IP REG_EIP +# define WIDTH "l" +#endif + +static unsigned long get_eflags(void) +{ + unsigned long eflags; + asm volatile ("pushf" WIDTH "\n\tpop" WIDTH " %0" : "=rm" (eflags)); + return eflags; +} + +static void set_eflags(unsigned long eflags) +{ + asm volatile ("push" WIDTH " %0\n\tpopf" WIDTH + : : "rm" (eflags) : "flags"); +} + +#define X86_EFLAGS_TF (1UL << 8) + +static void sigtrap(int sig, siginfo_t *info, void *ctx_void) +{ + ucontext_t *ctx = (ucontext_t*)ctx_void; + + if (get_eflags() & X86_EFLAGS_TF) { + set_eflags(get_eflags() & ~X86_EFLAGS_TF); + printf("[WARN]\tSIGTRAP handler had TF set\n"); + _exit(1); + } + + sig_traps++; + + if (sig_traps == 10000 || sig_traps == 10001) { + printf("[WARN]\tHit %d SIGTRAPs with si_addr 0x%lx, ip 0x%lx\n", + (int)sig_traps, + (unsigned long)info->si_addr, + (unsigned long)ctx->uc_mcontext.gregs[REG_IP]); + } +} + +static void check_result(void) +{ + unsigned long new_eflags = get_eflags(); + set_eflags(new_eflags & ~X86_EFLAGS_TF); + + if (!sig_traps) { + printf("[FAIL]\tNo SIGTRAP\n"); + exit(1); + } + + if (!(new_eflags & X86_EFLAGS_TF)) { + printf("[FAIL]\tTF was cleared\n"); + exit(1); + } + + printf("[OK]\tSurvived with TF set and %d traps\n", (int)sig_traps); + sig_traps = 0; +} + +int main() +{ + int tmp; + + sethandler(SIGTRAP, sigtrap, 0); + + printf("[RUN]\tSet TF and check nop\n"); + set_eflags(get_eflags() | X86_EFLAGS_TF); + asm volatile ("nop"); + check_result(); + +#ifdef __x86_64__ + printf("[RUN]\tSet TF and check syscall-less opportunistic sysret\n"); + set_eflags(get_eflags() | X86_EFLAGS_TF); + extern unsigned char post_nop[]; + asm volatile ("pushf" WIDTH "\n\t" + "pop" WIDTH " %%r11\n\t" + "nop\n\t" + "post_nop:" + : : "c" (post_nop) : "r11"); + check_result(); +#endif + + printf("[RUN]\tSet TF and check int80\n"); + set_eflags(get_eflags() | X86_EFLAGS_TF); + asm volatile ("int $0x80" : "=a" (tmp) : "a" (SYS_getpid)); + check_result(); + + /* + * This test is particularly interesting if fast syscalls use + * SYSENTER: it triggers a nasty design flaw in SYSENTER. + * Specifically, SYSENTER does not clear TF, so either SYSENTER + * or the next instruction traps at CPL0. (Of course, Intel + * mostly forgot to document exactly what happens here.) So we + * get a CPL0 fault with usergs (on 64-bit kernels) and possibly + * no stack. The only sane way the kernel can possibly handle + * it is to clear TF on return from the #DB handler, but this + * happens way too early to set TF in the saved pt_regs, so the + * kernel has to do something clever to avoid losing track of + * the TF bit. + * + * Needless to say, we've had bugs in this area. + */ + syscall(SYS_getpid); /* Force symbol binding without TF set. */ + printf("[RUN]\tSet TF and check a fast syscall\n"); + set_eflags(get_eflags() | X86_EFLAGS_TF); + syscall(SYS_getpid); + check_result(); + + /* Now make sure that another fast syscall doesn't set TF again. */ + printf("[RUN]\tFast syscall with TF cleared\n"); + fflush(stdout); /* Force a syscall */ + if (get_eflags() & X86_EFLAGS_TF) { + printf("[FAIL]\tTF is now set\n"); + exit(1); + } + if (sig_traps) { + printf("[FAIL]\tGot SIGTRAP\n"); + exit(1); + } + printf("[OK]\tNothing unexpected happened\n"); + + return 0; +}