Re: [PATCH v3 2/4] x86/syscalls: Specific usage of verify_pre_usermode_state
From: H. Peter Anvin
Date: Mon Mar 13 2017 - 17:53:31 EST
<skinsbursky@xxxxxxxxxxxxx>,Ingo Molnar <mingo@xxxxxxxxxx>,Paolo Bonzini <pbonzini@xxxxxxxxxx>,Dmitry Safonov <dsafonov@xxxxxxxxxxxxx>,Borislav Petkov <bp@xxxxxxxxx>,Josh Poimboeuf <jpoimboe@xxxxxxxxxx>,Brian Gerst <brgerst@xxxxxxxxx>,Jan Beulich <JBeulich@xxxxxxxx>,Christian Borntraeger <borntraeger@xxxxxxxxxx>,Fenghua Yu <fenghua.yu@xxxxxxxxx>,He Chen <he.chen@xxxxxxxxxxxxxxx>,Russell King <linux@xxxxxxxxxxxxxxx>,Vladimir Murzin <vladimir.murzin@xxxxxxx>,Will Deacon <will.deacon@xxxxxxx>,Catalin Marinas <catalin.marinas@xxxxxxx>,Mark Rutland <mark.rutland@xxxxxxx>,James Morse <james.morse@xxxxxxx>,"David A . Long" <dave.long@xxxxxxxxxx>,Pratyush Anand <panand@xxxxxxxxxx>,Laura Abbott <labbott@xxxxxxxxxx>,Andre Przywara <andre.przywara@xxxxxxx>,Chris Metcalf <cmetcalf@xxxxxxxxxxxx>,linux-s390@xxxxxxxxxxxxxxx,linux-kernel@xxxxxxxxxxxxxxx,linux-api@xxxxxxxxxxxxxxx,x86@xxxxxxxxxx,linux-arm-kernel@xxxxxxxxxxxxxxxxxxx,kernel-hardening@xxxxxxxxxxxxxxxxxx
From: hpa@xxxxxxxxx
Message-ID: <BB78ABF9-382E-43E8-BAC6-1EA6416A30DB@xxxxxxxxx>
On March 11, 2017 1:42:00 AM PST, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
>* Thomas Garnier <thgarnie@xxxxxxxxxx> wrote:
>
>> Implement specific usage of verify_pre_usermode_state for user-mode
>> returns for x86.
>> ---
>> Based on next-20170308
>> ---
>> arch/x86/Kconfig | 1 +
>> arch/x86/entry/common.c | 3 +++
>> arch/x86/entry/entry_64.S | 19 +++++++++++++++++++
>> arch/x86/include/asm/pgtable_64_types.h | 11 +++++++++++
>> arch/x86/include/asm/processor.h | 11 -----------
>> 5 files changed, 34 insertions(+), 11 deletions(-)
>>
>> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
>> index 005df7c825f5..6d48e18e6f09 100644
>> --- a/arch/x86/Kconfig
>> +++ b/arch/x86/Kconfig
>> @@ -63,6 +63,7 @@ config X86
>> select ARCH_MIGHT_HAVE_ACPI_PDC if ACPI
>> select ARCH_MIGHT_HAVE_PC_PARPORT
>> select ARCH_MIGHT_HAVE_PC_SERIO
>> + select ARCH_NO_SYSCALL_VERIFY_PRE_USERMODE_STATE
>> select ARCH_SUPPORTS_ATOMIC_RMW
>> select ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
>> select ARCH_SUPPORTS_NUMA_BALANCING if X86_64
>> diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
>> index 370c42c7f046..525edbb77f03 100644
>> --- a/arch/x86/entry/common.c
>> +++ b/arch/x86/entry/common.c
>> @@ -22,6 +22,7 @@
>> #include <linux/context_tracking.h>
>> #include <linux/user-return-notifier.h>
>> #include <linux/uprobes.h>
>> +#include <linux/syscalls.h>
>>
>> #include <asm/desc.h>
>> #include <asm/traps.h>
>> @@ -180,6 +181,8 @@ __visible inline void
>prepare_exit_to_usermode(struct pt_regs *regs)
>> struct thread_info *ti = current_thread_info();
>> u32 cached_flags;
>>
>> + verify_pre_usermode_state();
>> +
>> if (IS_ENABLED(CONFIG_PROVE_LOCKING) && WARN_ON(!irqs_disabled()))
>> local_irq_disable();
>>
>> diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
>> index d2b2a2948ffe..04db589be466 100644
>> --- a/arch/x86/entry/entry_64.S
>> +++ b/arch/x86/entry/entry_64.S
>> @@ -218,6 +218,25 @@ entry_SYSCALL_64_fastpath:
>> testl $_TIF_ALLWORK_MASK, TASK_TI_flags(%r11)
>> jnz 1f
>>
>> + /*
>> + * Check user-mode state on fast path return, the same check is
>done
>> + * under the slow path through syscall_return_slowpath.
>> + */
>> +#ifdef CONFIG_BUG_ON_DATA_CORRUPTION
>> + call verify_pre_usermode_state
>> +#else
>> + /*
>> + * Similar to set_fs(USER_DS) in verify_pre_usermode_state without
>a
>> + * warning.
>> + */
>> + movq PER_CPU_VAR(current_task), %rax
>> + movq $TASK_SIZE_MAX, %rcx
>> + cmp %rcx, TASK_addr_limit(%rax)
>> + jz 1f
>> + movq %rcx, TASK_addr_limit(%rax)
>> +1:
>> +#endif
>> +
>> LOCKDEP_SYS_EXIT
>> TRACE_IRQS_ON /* user mode is traced as IRQs on */
>> movq RIP(%rsp), %rcx
>
>Ugh, so you call an assembly function just to ... call another
>function.
>
>Plus why is it in assembly to begin with? Is this some older code that
>got
>written when the x86 entry code was in assembly, and never properly
>converted to C?
>
>Thanks,
>
> Ingo
The code does a compare to jump around a store. It would be much cleaner and faster to simply clobber the value unconditionally. If there is a test it should be to avoid the function call, not (only) the assignment.
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.