fix: make zfs_strerror really thread-safe and portable

[rkojedzinszky]

Fixes #15793
Fixes #16640
Closes #16916

Motivation and Context

#15793 wanted to make zfs_strerror threadsafe, unfortunately, it turned out that strerror_l() usage was wrong, and also, some libc implementations dont have strerror_l().

Description

zfs_strerror() now simply calls original strerror() and copies the result to a thread-local buffer, then returns that.

How Has This Been Tested?

Code did compile.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[rkojedzinszky]

May it happen that strerror_r() is available in most libc implementations, and the mutex could be dropped?

[amotin]

May it happen that strerror_r() is available in most libc implementations, and the mutex could be dropped?

I can't speak about other OS'es, but in FreeBSD I see it for 20+ years since FreeBSD 4.x.

[rkojedzinszky]

Did a quick check, found that uClibc and Musl do have strerror_r(). Glibc and FreeBSD have it too, so maybe, we could remove that piece of code?

[amotin]

From FreeBSD's man page: "The strerror_r() function conforms to IEEE Std 1003.1-2001 (“POSIX.1”)". Sounds like quite a while ago.

[rkojedzinszky]

uClibc and Musl are known to be small libc implementations, perhaps with missing functions. Howewer, I found that both have strerror_r(), and it seems not to be configurable/optional.

[behlendorf]

There appears to be two versions of strerror_r(), an "XSI-compliant version specified in POSIX.1-2001" and a "GNU-specific version". So that's something to be careful about.

From https://manpages.ubuntu.com/manpages/bionic/man3/strerror.3.html

int strerror_r(int errnum, char *buf, size_t buflen); /* XSI-compliant */
char *strerror_r(int errnum, char *buf, size_t buflen); /* GNU-specific */

       strerror_r():
           The XSI-compliant version is provided if:
           (_POSIX_C_SOURCE >= 200112L) && !  _GNU_SOURCE
           Otherwise, the GNU-specific version is provided.

It's also not clear to me from the documentation if strerror_r() honors the locale. That's something we'd need to test.

[rkojedzinszky]

Good catch, the GNU version may not write to the supplied buffer. Oh.

[rkojedzinszky]

Regarding locale, it should behave as strerror().

[rkojedzinszky]

Unfortunately, openzfs has AM_CPPFLAGS += -D_GNU_SOURCE in config/Rules.am, which definitely makes the GNU version available with glibc. FreeBSD on the other hand definitely provides the POSIX variant and behavior, and there are the other libcs too. So, we must stick back to simple strerror() guarder by the pthread mutex.