On Sun, May 30 2021 at 16:41, Andy Lutomirski wrote:
> On 5/30/21 3:02 PM, Thomas Gleixner wrote:
>>> /*
>>> - * Clear the FPU state back to init state.
>>> - *
>>> - * Called by sys_execve(), by the signal handler code and by various
>>> - * error paths.
>>> + * Reset current's user FPU states to the init states. The caller promises
>>> + * that current's supervisor states (in memory or CPU regs as appropriate)
>>> + * as well as the XSAVE header in memory are intact.
>
> ^^^ The caller promises this
Yes, I misread this, but it's more than non-obvious.
> This patch fixes your reproducer and my (to-be-sent) reproducer. I
> tested it on a machine that physically has the XRSTORS instruction but I
> disabled it using virt. Are you still seeing failures with this patch
> applied? I can try to test on a different CPU.
Seems I applied the patch, built it and then failed to actually boot
that kernel. I retested with brain awake and it indeed works.
Sorry for the rant!
Thanks,
tglx
