Re: [patch 01/10] x86/fpu/signal: Clarify exception handling in restore_fpregs_from_user()

From: Thomas Gleixner
Date: Mon Aug 30 2021 - 16:09:37 EST


On Mon, Aug 30 2021 at 21:33, Borislav Petkov wrote:
> On Mon, Aug 30, 2021 at 06:27:22PM +0200, Thomas Gleixner wrote:
>> FPU restore from a signal frame can trigger various exceptions. The
>> exceptions are caught with an exception table entry. The handler of this
>> entry sets the error return value to the negated exception number.
>>
>> Any other exception than #PF is fatal and recovery is not possible. This
>> relies on the fact that the #PF exception number is the same as EFAULT, but
>> that's not really obvious.
>>
>> Check the error code for -X86_TRAP_PF instead of checking it for -EFAULT to
>> make it clear how that works.
>
> I guess you wanna fixup the comment over XSTATE_OP() too and perhaps
> mention ex_handler_fault() explicitly so that one can make her/his way
> around the code and pinpoint quickly where it sticks that exception
> number into rAX.

Something like the below?

Thanks,

tglx
---
--- a/arch/x86/include/asm/fpu/internal.h
+++ b/arch/x86/include/asm/fpu/internal.h
@@ -88,7 +88,11 @@ static inline void fpstate_init_soft(str
#endif
extern void save_fpregs_to_fpstate(struct fpu *fpu);

-/* Returns 0 or the negated trap number, which results in -EFAULT for #PF */
+/*
+ * Returns 0 on success or the negated trap number when the operation
+ * raises an exception (The exception fixup function ex_handler_fault()
+ * stores the trap number in EAX).
+ */
#define user_insn(insn, output, input...) \
({ \
int err; \
@@ -199,7 +203,8 @@ static inline void fxsave(struct fxregs_

/*
* After this @err contains 0 on success or the negated trap number when
- * the operation raises an exception. For faults this results in -EFAULT.
+ * the operation raises an exception (The exception fixup function
+ * ex_handler_fault() stores the trap number in EAX).
*/
#define XSTATE_OP(op, st, lmask, hmask, err) \
asm volatile("1:" op "\n\t" \