Dear Maintainer,

Looking at ci.debian.net, I've seen that binutils tests started to fail
with version 2.37.50. Please refer to:

* https://ci.debian.net/packages/b/binutils/testing/amd64/
* first failure: 2.37.50.20220106-2 at 2022-01-13 10:21:22 UTC, logs at
  <https://ci.debian.net/data/autopkgtest/testing/amd64/b/binutils/18287140/log.gz>

The failure is:

Test results, compared with installed binutils:
W: [ld-elf/linux-x86.exp] REGRESSION (PASS -> FAIL): Run p_align-1b with PIE
W: [ld-elf/linux-x86.exp] REGRESSION (PASS -> FAIL): Run p_align-1d with -Wl,-z,max-page-size=0x1000 with PIE
2 REGRESSIONS (0.04%).

Interestingly, the tests sometimes pass, sometimes fail. I took a look
at the logs, and I can see that successes occur on kernel 5.10 while
failures occur on kernel 4.19.

I did a bit more research, and it might be related to this kernel
commit, that indeed was included in the kernel 5.10-rc1 for the first
time:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ce81bb256a224259ab686742a6284930cbe4f1fa

I wonder if there's something that can be fixed in binutils, so that the
tests can pass again on kernel 4.19. Unless it's something that has to
be backported in the kernel 4.19, no idea.

Thanks!