| Message ID | 519fee5268faea09ae550776ce969fa6e88668b0.1520449896.git.luto@kernel.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Mar 7, 2018 at 11:12 AM, Andy Lutomirski <luto@kernel.org> wrote: > Since Linux 3.2, vsyscalls have been deprecated and slow. From 3.2 > on, Linux had three vsyscall modes: "native", "emulate", and "none". > > "emulate" is the default. All known user programs work correctly in > emulate mode, but vsyscalls turn into page faults and are emulated. > This is very slow. In "native" mode, the vsyscall page is easily > usable as an exploit gadget, but vsyscalls are a bit faster -- they > turn into normal syscalls. (This is in contrast to vDSO functions, > which can be much faster than syscalls.) In "none" mode, there are > no vsyscalls. > > For all practical purposes, "native" was really just a chicken bit > in case something went wrong with the emulation. It's been over six > years, and nothing has gone wrong. Delete it. > > Signed-off-by: Andy Lutomirski <luto@kernel.org> Acked-by: Kees Cook <keescook@chromium.org> I related news, I wonder how long before we can switch from EMULATE to NONE as the default? glibc 2.15 (which stopped using vsyscall) is (not coincidentally) 6 years old too... -Kees > --- > arch/x86/Kconfig | 11 +---------- > arch/x86/entry/vsyscall/vsyscall_64.c | 16 +++------------- > arch/x86/include/asm/pgtable_types.h | 2 -- > tools/testing/selftests/x86/test_vsyscall.c | 11 ++++++----- > 4 files changed, 10 insertions(+), 30 deletions(-) > > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig > index eb7f43f23521..0fa71a78ec99 100644 > --- a/arch/x86/Kconfig > +++ b/arch/x86/Kconfig > @@ -2307,7 +2307,7 @@ choice > it can be used to assist security vulnerability exploitation. > > This setting can be changed at boot time via the kernel command > - line parameter vsyscall=[native|emulate|none]. > + line parameter vsyscall=[emulate|none]. > > On a system with recent enough glibc (2.14 or newer) and no > static binaries, you can say None without a performance penalty > @@ -2315,15 +2315,6 @@ choice > > If unsure, select "Emulate". > > - config LEGACY_VSYSCALL_NATIVE > - bool "Native" > - help > - Actual executable code is located in the fixed vsyscall > - address mapping, implementing time() efficiently. Since > - this makes the mapping executable, it can be used during > - security vulnerability exploitation (traditionally as > - ROP gadgets). This configuration is not recommended. > - > config LEGACY_VSYSCALL_EMULATE > bool "Emulate" > help > diff --git a/arch/x86/entry/vsyscall/vsyscall_64.c b/arch/x86/entry/vsyscall/vsyscall_64.c > index 577fa8adb785..8560ef68a9d6 100644 > --- a/arch/x86/entry/vsyscall/vsyscall_64.c > +++ b/arch/x86/entry/vsyscall/vsyscall_64.c > @@ -42,10 +42,8 @@ > #define CREATE_TRACE_POINTS > #include "vsyscall_trace.h" > > -static enum { EMULATE, NATIVE, NONE } vsyscall_mode = > -#if defined(CONFIG_LEGACY_VSYSCALL_NATIVE) > - NATIVE; > -#elif defined(CONFIG_LEGACY_VSYSCALL_NONE) > +static enum { EMULATE, NONE } vsyscall_mode = > +#ifdef CONFIG_LEGACY_VSYSCALL_NONE > NONE; > #else > EMULATE; > @@ -56,8 +54,6 @@ static int __init vsyscall_setup(char *str) > if (str) { > if (!strcmp("emulate", str)) > vsyscall_mode = EMULATE; > - else if (!strcmp("native", str)) > - vsyscall_mode = NATIVE; > else if (!strcmp("none", str)) > vsyscall_mode = NONE; > else > @@ -139,10 +135,6 @@ bool emulate_vsyscall(struct pt_regs *regs, unsigned long address) > > WARN_ON_ONCE(address != regs->ip); > > - /* This should be unreachable in NATIVE mode. */ > - if (WARN_ON(vsyscall_mode == NATIVE)) > - return false; > - > if (vsyscall_mode == NONE) { > warn_bad_vsyscall(KERN_INFO, regs, > "vsyscall attempted with vsyscall=none"); > @@ -370,9 +362,7 @@ void __init map_vsyscall(void) > > if (vsyscall_mode != NONE) { > __set_fixmap(VSYSCALL_PAGE, physaddr_vsyscall, > - vsyscall_mode == NATIVE > - ? PAGE_KERNEL_VSYSCALL > - : PAGE_KERNEL_VVAR); > + PAGE_KERNEL_VVAR); > set_vsyscall_pgtable_user_bits(swapper_pg_dir); > } > > diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h > index 246f15b4e64c..acfe755562a6 100644 > --- a/arch/x86/include/asm/pgtable_types.h > +++ b/arch/x86/include/asm/pgtable_types.h > @@ -174,7 +174,6 @@ enum page_cache_mode { > #define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW) > #define __PAGE_KERNEL_RX (__PAGE_KERNEL_EXEC & ~_PAGE_RW) > #define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_NOCACHE) > -#define __PAGE_KERNEL_VSYSCALL (__PAGE_KERNEL_RX | _PAGE_USER) > #define __PAGE_KERNEL_VVAR (__PAGE_KERNEL_RO | _PAGE_USER) > #define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE) > #define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE) > @@ -206,7 +205,6 @@ enum page_cache_mode { > #define PAGE_KERNEL_NOCACHE __pgprot(__PAGE_KERNEL_NOCACHE | _PAGE_ENC) > #define PAGE_KERNEL_LARGE __pgprot(__PAGE_KERNEL_LARGE | _PAGE_ENC) > #define PAGE_KERNEL_LARGE_EXEC __pgprot(__PAGE_KERNEL_LARGE_EXEC | _PAGE_ENC) > -#define PAGE_KERNEL_VSYSCALL __pgprot(__PAGE_KERNEL_VSYSCALL | _PAGE_ENC) > #define PAGE_KERNEL_VVAR __pgprot(__PAGE_KERNEL_VVAR | _PAGE_ENC) > > #define PAGE_KERNEL_IO __pgprot(__PAGE_KERNEL_IO) > diff --git a/tools/testing/selftests/x86/test_vsyscall.c b/tools/testing/selftests/x86/test_vsyscall.c > index be81621446f0..0b4f1cc2291c 100644 > --- a/tools/testing/selftests/x86/test_vsyscall.c > +++ b/tools/testing/selftests/x86/test_vsyscall.c > @@ -450,7 +450,7 @@ static void sigtrap(int sig, siginfo_t *info, void *ctx_void) > num_vsyscall_traps++; > } > > -static int test_native_vsyscall(void) > +static int test_emulation(void) > { > time_t tmp; > bool is_native; > @@ -458,7 +458,7 @@ static int test_native_vsyscall(void) > if (!vtime) > return 0; > > - printf("[RUN]\tchecking for native vsyscall\n"); > + printf("[RUN]\tchecking that vsyscalls are emulated\n"); > sethandler(SIGTRAP, sigtrap, 0); > set_eflags(get_eflags() | X86_EFLAGS_TF); > vtime(&tmp); > @@ -474,11 +474,12 @@ static int test_native_vsyscall(void) > */ > is_native = (num_vsyscall_traps > 1); > > - printf("\tvsyscalls are %s (%d instructions in vsyscall page)\n", > + printf("[%s]\tvsyscalls are %s (%d instructions in vsyscall page)\n", > + (is_native ? "FAIL" : "OK"), > (is_native ? "native" : "emulated"), > (int)num_vsyscall_traps); > > - return 0; > + return is_native; > } > #endif > > @@ -498,7 +499,7 @@ int main(int argc, char **argv) > nerrs += test_vsys_r(); > > #ifdef __x86_64__ > - nerrs += test_native_vsyscall(); > + nerrs += test_emulation(); > #endif > > return nerrs ? 1 : 0; > -- > 2.14.3 >
On Wed, Mar 7, 2018 at 11:12 AM, Andy Lutomirski <luto@kernel.org> wrote: > > For all practical purposes, "native" was really just a chicken bit > in case something went wrong with the emulation. It's been over six > years, and nothing has gone wrong. Delete it. Ack. I guess we';ll hear if anybody might still use it and cares, but.. Linus
On Wed, Mar 07, 2018 at 11:21:46AM -0800, Kees Cook wrote: > On Wed, Mar 7, 2018 at 11:12 AM, Andy Lutomirski <luto@kernel.org> wrote: > > Since Linux 3.2, vsyscalls have been deprecated and slow. From 3.2 > > on, Linux had three vsyscall modes: "native", "emulate", and "none". > > > > "emulate" is the default. All known user programs work correctly in > > emulate mode, but vsyscalls turn into page faults and are emulated. > > This is very slow. In "native" mode, the vsyscall page is easily > > usable as an exploit gadget, but vsyscalls are a bit faster -- they > > turn into normal syscalls. (This is in contrast to vDSO functions, > > which can be much faster than syscalls.) In "none" mode, there are > > no vsyscalls. > > > > For all practical purposes, "native" was really just a chicken bit > > in case something went wrong with the emulation. It's been over six > > years, and nothing has gone wrong. Delete it. > > > > Signed-off-by: Andy Lutomirski <luto@kernel.org> > > Acked-by: Kees Cook <keescook@chromium.org> > > I related news, I wonder how long before we can switch from EMULATE to > NONE as the default? glibc 2.15 (which stopped using vsyscall) is (not > coincidentally) 6 years old too... Or, in the meantime, add a warning if (vsyscall_mode == EMULATE) { warn_bad_vsyscall(KERN_INFO, regs, "vsyscalls are deprecated -- use vDSO instead"); } Otherwise, the patch looks good. Thanks, Dominik
On Wed, Mar 7, 2018 at 7:31 PM, Dominik Brodowski <linux@dominikbrodowski.net> wrote: > On Wed, Mar 07, 2018 at 11:21:46AM -0800, Kees Cook wrote: >> On Wed, Mar 7, 2018 at 11:12 AM, Andy Lutomirski <luto@kernel.org> wrote: >> > Since Linux 3.2, vsyscalls have been deprecated and slow. From 3.2 >> > on, Linux had three vsyscall modes: "native", "emulate", and "none". >> > >> > "emulate" is the default. All known user programs work correctly in >> > emulate mode, but vsyscalls turn into page faults and are emulated. >> > This is very slow. In "native" mode, the vsyscall page is easily >> > usable as an exploit gadget, but vsyscalls are a bit faster -- they >> > turn into normal syscalls. (This is in contrast to vDSO functions, >> > which can be much faster than syscalls.) In "none" mode, there are >> > no vsyscalls. >> > >> > For all practical purposes, "native" was really just a chicken bit >> > in case something went wrong with the emulation. It's been over six >> > years, and nothing has gone wrong. Delete it. >> > >> > Signed-off-by: Andy Lutomirski <luto@kernel.org> >> >> Acked-by: Kees Cook <keescook@chromium.org> >> >> I related news, I wonder how long before we can switch from EMULATE to >> NONE as the default? glibc 2.15 (which stopped using vsyscall) is (not >> coincidentally) 6 years old too... > > Or, in the meantime, add a warning > > if (vsyscall_mode == EMULATE) { > warn_bad_vsyscall(KERN_INFO, regs, > "vsyscalls are deprecated -- use vDSO instead"); > } > > Otherwise, the patch looks good. > I'm not sure the warning buys us anything. AFAICT there are no modern toolchains in any language that will use vsyscalls. The problem is old binaries.
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index eb7f43f23521..0fa71a78ec99 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -2307,7 +2307,7 @@ choice it can be used to assist security vulnerability exploitation. This setting can be changed at boot time via the kernel command - line parameter vsyscall=[native|emulate|none]. + line parameter vsyscall=[emulate|none]. On a system with recent enough glibc (2.14 or newer) and no static binaries, you can say None without a performance penalty @@ -2315,15 +2315,6 @@ choice If unsure, select "Emulate". - config LEGACY_VSYSCALL_NATIVE - bool "Native" - help - Actual executable code is located in the fixed vsyscall - address mapping, implementing time() efficiently. Since - this makes the mapping executable, it can be used during - security vulnerability exploitation (traditionally as - ROP gadgets). This configuration is not recommended. - config LEGACY_VSYSCALL_EMULATE bool "Emulate" help diff --git a/arch/x86/entry/vsyscall/vsyscall_64.c b/arch/x86/entry/vsyscall/vsyscall_64.c index 577fa8adb785..8560ef68a9d6 100644 --- a/arch/x86/entry/vsyscall/vsyscall_64.c +++ b/arch/x86/entry/vsyscall/vsyscall_64.c @@ -42,10 +42,8 @@ #define CREATE_TRACE_POINTS #include "vsyscall_trace.h" -static enum { EMULATE, NATIVE, NONE } vsyscall_mode = -#if defined(CONFIG_LEGACY_VSYSCALL_NATIVE) - NATIVE; -#elif defined(CONFIG_LEGACY_VSYSCALL_NONE) +static enum { EMULATE, NONE } vsyscall_mode = +#ifdef CONFIG_LEGACY_VSYSCALL_NONE NONE; #else EMULATE; @@ -56,8 +54,6 @@ static int __init vsyscall_setup(char *str) if (str) { if (!strcmp("emulate", str)) vsyscall_mode = EMULATE; - else if (!strcmp("native", str)) - vsyscall_mode = NATIVE; else if (!strcmp("none", str)) vsyscall_mode = NONE; else @@ -139,10 +135,6 @@ bool emulate_vsyscall(struct pt_regs *regs, unsigned long address) WARN_ON_ONCE(address != regs->ip); - /* This should be unreachable in NATIVE mode. */ - if (WARN_ON(vsyscall_mode == NATIVE)) - return false; - if (vsyscall_mode == NONE) { warn_bad_vsyscall(KERN_INFO, regs, "vsyscall attempted with vsyscall=none"); @@ -370,9 +362,7 @@ void __init map_vsyscall(void) if (vsyscall_mode != NONE) { __set_fixmap(VSYSCALL_PAGE, physaddr_vsyscall, - vsyscall_mode == NATIVE - ? PAGE_KERNEL_VSYSCALL - : PAGE_KERNEL_VVAR); + PAGE_KERNEL_VVAR); set_vsyscall_pgtable_user_bits(swapper_pg_dir); } diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h index 246f15b4e64c..acfe755562a6 100644 --- a/arch/x86/include/asm/pgtable_types.h +++ b/arch/x86/include/asm/pgtable_types.h @@ -174,7 +174,6 @@ enum page_cache_mode { #define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW) #define __PAGE_KERNEL_RX (__PAGE_KERNEL_EXEC & ~_PAGE_RW) #define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_NOCACHE) -#define __PAGE_KERNEL_VSYSCALL (__PAGE_KERNEL_RX | _PAGE_USER) #define __PAGE_KERNEL_VVAR (__PAGE_KERNEL_RO | _PAGE_USER) #define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE) #define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE) @@ -206,7 +205,6 @@ enum page_cache_mode { #define PAGE_KERNEL_NOCACHE __pgprot(__PAGE_KERNEL_NOCACHE | _PAGE_ENC) #define PAGE_KERNEL_LARGE __pgprot(__PAGE_KERNEL_LARGE | _PAGE_ENC) #define PAGE_KERNEL_LARGE_EXEC __pgprot(__PAGE_KERNEL_LARGE_EXEC | _PAGE_ENC) -#define PAGE_KERNEL_VSYSCALL __pgprot(__PAGE_KERNEL_VSYSCALL | _PAGE_ENC) #define PAGE_KERNEL_VVAR __pgprot(__PAGE_KERNEL_VVAR | _PAGE_ENC) #define PAGE_KERNEL_IO __pgprot(__PAGE_KERNEL_IO) diff --git a/tools/testing/selftests/x86/test_vsyscall.c b/tools/testing/selftests/x86/test_vsyscall.c index be81621446f0..0b4f1cc2291c 100644 --- a/tools/testing/selftests/x86/test_vsyscall.c +++ b/tools/testing/selftests/x86/test_vsyscall.c @@ -450,7 +450,7 @@ static void sigtrap(int sig, siginfo_t *info, void *ctx_void) num_vsyscall_traps++; } -static int test_native_vsyscall(void) +static int test_emulation(void) { time_t tmp; bool is_native; @@ -458,7 +458,7 @@ static int test_native_vsyscall(void) if (!vtime) return 0; - printf("[RUN]\tchecking for native vsyscall\n"); + printf("[RUN]\tchecking that vsyscalls are emulated\n"); sethandler(SIGTRAP, sigtrap, 0); set_eflags(get_eflags() | X86_EFLAGS_TF); vtime(&tmp); @@ -474,11 +474,12 @@ static int test_native_vsyscall(void) */ is_native = (num_vsyscall_traps > 1); - printf("\tvsyscalls are %s (%d instructions in vsyscall page)\n", + printf("[%s]\tvsyscalls are %s (%d instructions in vsyscall page)\n", + (is_native ? "FAIL" : "OK"), (is_native ? "native" : "emulated"), (int)num_vsyscall_traps); - return 0; + return is_native; } #endif @@ -498,7 +499,7 @@ int main(int argc, char **argv) nerrs += test_vsys_r(); #ifdef __x86_64__ - nerrs += test_native_vsyscall(); + nerrs += test_emulation(); #endif return nerrs ? 1 : 0;
Since Linux 3.2, vsyscalls have been deprecated and slow. From 3.2 on, Linux had three vsyscall modes: "native", "emulate", and "none". "emulate" is the default. All known user programs work correctly in emulate mode, but vsyscalls turn into page faults and are emulated. This is very slow. In "native" mode, the vsyscall page is easily usable as an exploit gadget, but vsyscalls are a bit faster -- they turn into normal syscalls. (This is in contrast to vDSO functions, which can be much faster than syscalls.) In "none" mode, there are no vsyscalls. For all practical purposes, "native" was really just a chicken bit in case something went wrong with the emulation. It's been over six years, and nothing has gone wrong. Delete it. Signed-off-by: Andy Lutomirski <luto@kernel.org> --- arch/x86/Kconfig | 11 +---------- arch/x86/entry/vsyscall/vsyscall_64.c | 16 +++------------- arch/x86/include/asm/pgtable_types.h | 2 -- tools/testing/selftests/x86/test_vsyscall.c | 11 ++++++----- 4 files changed, 10 insertions(+), 30 deletions(-)