[RFC PATCH 1/4] nptl: use futex_time64 for lll_futex_syscall() if possible
Adhemerval Zanella Netto
adhemerval.zanella@linaro.org
Wed Nov 11 15:51:29 GMT 2026
On 11/11/26 11:21, Arnd Bergmann wrote:
> On Wed, Nov 11, 2026, at 14:54, Adhemerval Zanella Netto wrote:
>> On 11/11/26 09:56, Arnd Bergmann wrote:
>>> On Wed, Nov 11, 2026, at 13:08, Adhemerval Zanella Netto wrote:
>
>>> // in a new architecture-independent header included from there
>>> #if __WORDSIZE==64 || (defined(__x86_64__) && defined(__ILP32__))
>>> # define __NR_futex __NR_futex_time64
>>> # define __NR_gettimeofday __NR_gettimeofday_time64
>>> ...
>>> #elif !defined(__USE_TIME_BITS64)
>>> # define __NR_futex __NR_futex_time32
>>> # define __NR_gettimeofday __NR_gettimeofday_time32
>>> ...
>>> #else
>>> // leave the old macros undefined when building with __USE_TIME_BITS64
>>> #endif
>>>
>>> This would forcibly break any source code that attempts to
>>> call the time32 syscalls by their traditional name in a source
>>> file built with -D_TIME_BITS=64, which would otherwise be a bug
>>> that may go unnoticed for a long time.
>>
>> The _TIME_BITS is only meant to be used by programs built against glibc, we
>> only use it on tests (and configuring glibc with this should trigger an
>> error). And I am not sure if it would simplify things, internal calls
>> usually only uses 64 bit time_t variants now (either through wrappers like
>> lowlevellock.h or by routing explicitly to internal symbols). The 32 bit
>> time_t variants are only provided for ABI compatibility.
>>
>> Or do you mean by external users, through syscall()? In this case I am
>> not sure if routines the SYS_ macro depending of _TIME_BITS value is a
>> good idea.
>
> I specifically care about external users that call syscall(__NR_...)
> and end up with the wrong ABI when _TIME_BITS=64 is set at build time
> (manually or as the distro default setting). In particular I think
> more than half the users of __NR_futex were at some point converted
> incorrectly, e.g. the first ones I find in debian codesearch:
>
> https://sources.debian.org/src/mesa/26.2.6-1/src/util/futex.c?hl=38#L37
> https://sources.debian.org/src/firefox-esr/153.6.0esr-1/third_party/abseil-cpp/absl/synchronization/internal/futex.h#L81
> https://sources.debian.org/src/jackd2/1.9.22~dfsg-6/linux/JackLinuxFutex.cpp?hl=34#L110
>
> These would be examples where a 'struct timespec' is passed
> into syscall(__NR_futex, ...) regardless of _TIME_BITS.
> On rv32 this works because __NR_futex is not defined, on
> most little-endian 32-bit targets this kind-of works as long
> as the timeout is between 1 and INT_MAX seconds because you
> get the correct value in the first four bytes of the structure.
> For FUTEX_CLOCK_REALTIME timeouts past 2038, for big-endian
> architectures, and for CONFIG_COMPAT_32BIT_TIME=n this is
> of course completely broken.
>
> The best fix here would be to use the libc wrapper for futex
> if available, so I probably wouldn't change the status quo
> before a glibc release that has this.
I don't think glibc can't do much if the callers want to bypass it, and I
agree the best we can do is to offer proper wrappers to make this right.
More information about the Libc-alpha
mailing list