x64: fix FPU corruption with signals and preemption
PATCH 1/1 x64: fix FPU corruption with signals and preemption
From: Suresh Siddha <[email protected]>
Subject: x64: fix FPU corruption with signals and preemption
Impact: fix FPU state corruption
In 64bit signal delivery path, clear_used_math() was happening before saving the current active FPU state on to the user stack for signal handling. Between clear_used_math() and the state store on to the user stack, potentially we can get a page fault for the user address and can block. Infact, while testing we were hitting the might_fault() in __clear_user() which can do a schedule().
At a later point in time, we will schedule back into this process and resume the save state (using "xsave/fxsave" instruction) which can lead to DNA fault. And as used_math was cleared before, we will reinit the FP state in the DNA fault and continue. This reinit will result in loosing the FPU state of the process.
Move clear_used_math() to a point after the FPU state has been stored onto the user stack.
This issue is present from a long time (even before the xsave changes and the x86 merge). But it can easily be exposed in 2.6.28.x and 2.6.29.x series because of the __clear_user() in this path, which has an explicit __cond_resched() leading to a context switch with CONFIG_PREEMPT_VOLUNTARY.
---
arch/x86/kernel/xsave.c +3 -1
@@ -89,7 +89,7 @@ int save_i387_xstate(void __user *buf)if (!used_math ())return 0 ; -clear_used_math ();/* trigger finit */ +if (task_thread_info (tsk)->status &TS_USEDFPU ) {/* * Start with clearing the user buffer. This will present a
@@ -114,6 +114,8 @@ int save_i387_xstate(void __user *buf)return -1 ; } +clear_used_math ();/* trigger finit */ +if (task_thread_info (tsk)->status &TS_XSAVE ) {struct _fpstate __user *fx = buf;struct _xstate __user *x = buf;
Quoted 20 lines · * Suresh Siddha ([email protected]) wrote:
From: Suresh Siddha <[email protected]>
Subject: x64: fix FPU corruption with signals and preemptionImpact: fix FPU state corruption
In 64bit signal delivery path,
clear_used_math()was happening before saving the current active FPU state on to the user stack for signal handling. Betweenclear_used_math()and the state store on to the user stack, potentially we can get a page fault for the user address and can block. Infact, while testing we were hitting themight_fault()in__clear_user()which can do aschedule().At a later point in time, we will schedule back into this process and resume the save state (using "xsave/fxsave" instruction) which can lead to DNA fault. And as
used_mathwas cleared before, we will reinit the FP state in the DNA fault and continue. This reinit will result in loosing the FPU state of the process.Move
clear_used_math()to a point after the FPU state has been stored onto the user stack.
Quoted 7 lines
This issue is present from a long time (even before the xsave changes and the x86 merge). But it can easily be exposed in 2.6.28.x and 2.6.29.x series because of the
__clear_user()in this path, which has an explicit__cond_resched()leading to a context switch withCONFIG_PREEMPT_VOLUNTARY.Signed-off-by Suresh Siddha <[email protected]>Cc [email protected] [2.6.28.x, 2.6.29.x]
This one get lost?
thanks,
-chris
Quoted 6 lines · Chris Wright
Quoted 3 lines
Signed-off-by Suresh Siddha <[email protected]>Cc [email protected] [2.6.28.x, 2.6.29.x]This one get lost?
No, it's queued up in tip:x86/urgent and will be pushed upstream with the next push of x86 fixes to Linus.
-hpa
Quoted 9 lines · * H. Peter Anvin ([email protected]) wrote:
Quoted 5 lines · Chris Wright
Quoted 3 lines
Signed-off-by Suresh Siddha <[email protected]>Cc [email protected] [2.6.28.x, 2.6.29.x]This one get lost?
No, it's queued up in tip:x86/urgent and will be pushed upstream with the next push of x86 fixes to Linus.
OK, I had it marked as not upstream and not in tip, thanks for the correction.
thanks,
-chris
Quoted 6 lines · Chris Wright
Quoted 5 lines · * H. Peter Anvin ([email protected]) wrote:
Quoted 4 lines · Chris Wright
Quoted 3 lines
Signed-off-by Suresh Siddha <[email protected]>Cc [email protected] [2.6.28.x, 2.6.29.x]
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
This one get lost?
No, it's queued up in tip:x86/urgent and will be pushed upstream with the next push of x86 fixes to Linus.
OK, I had it marked as not upstream and not in tip, thanks for the correction.
there was a bit of a patch delay with it. Since this is an old race leading back to ancient times, and because this is touching critical code, we kept it around some more before pushing it to Linus. The plan is for it to show up in -git in the next few days and then in -rc4.
Ingo