RE: [patch V3 00/20] x86/fpu: Clean up exception fixups and error handling in sigframe related code

[Luck, Tony]

> Much better. This version boots.  Machine check recovery works as well as it did before
> (user triggered machine checks are fine, kernel triggered ones are ok where they were
> before and still trigger infinite loops where they did before).

With these three patches:

https://lore.kernel.org/all/20210818002942.1607544-1-tony.luck@xxxxxxxxx/

on top of your series the infinite loops are fixed (first patch).

Other two patches fix my write(2) test which does:

	ret = write(fd, buf, 512);

Where I have injected poison at some offset within "buf".

If the poison is at the start of buf I get ret == -1 with errno == -EFAULT.
If the poison is at a non-zero offset, the the bytes up to the poison
are written to the file and "ret" tells you how many were written.

-Tony