From: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
To: Arjun Shankar <arjun@redhat.com>
Cc: libc-alpha@sourceware.org
Subject: Re: [PATCH v2] linux: Avoid shifting a negative signed on POSIX timer interface
Date: Thu, 20 Oct 2022 10:39:33 -0300 [thread overview]
Message-ID: <ee974e11-179d-b701-fbf6-b48392a42a4c@linaro.org> (raw)
In-Reply-To: <CAG_osabTiNXs3cYSSNJEB-yhixLsDA1=sJt7NO8y7N2y4Sr0FQ@mail.gmail.com>
On 17/10/22 19:33, Arjun Shankar wrote:
> Hi Adhemerval,
>
>> The current macros uses pid as signed value, which triggers a compiler
>> warning for process and thread timers. Replace MAKE_PROCESS_CPUCLOCK
>> with static inline function that expects the pid as unsigned. These
>> are similar to what Linux does internally.
>>
>> Checked on x86_64-linux-gnu.
>
> I tried building master with -Wshift-negative-value and ran into a
> couple of other left-shifts on negative values in
> sysdeps/x86/dl-cacheinfo.h:
That's a good experiment, I see that clang enables it as default.
>
>> count_mask = ~(-1 << (count_mask + 1));
>
> Perhaps a cast of the -1 to unsigned here would make the intent even
> more clear, and get glibc building cleanly with
> -Wshift-negative-value. What do you think?
I think we can also make the mask unsigned to avoid a potential UB
and remove the inline assembly since we have a proper builtin for it:
unsigned int count_mask = __builtin_clz (threads_l2);
count_mask = ~(-1U << (count_mask + 1));
>
> Anyway, this patch itself looks good to me.
>
> Reviewed-by: Arjun Shankar <arjun@redhat.com>
Thanks.
>
>> nptl/pthread_getcpuclockid.c | 2 +-
>> sysdeps/unix/sysv/linux/clock_getcpuclockid.c | 2 +-
>> sysdeps/unix/sysv/linux/clock_nanosleep.c | 2 +-
>> .../unix/sysv/linux/kernel-posix-cpu-timers.h | 28 +++++++++++++++----
>> sysdeps/unix/sysv/linux/timer_create.c | 4 +--
>> 5 files changed, 28 insertions(+), 10 deletions(-)
>>
>> diff --git a/nptl/pthread_getcpuclockid.c b/nptl/pthread_getcpuclockid.c
>> index 344bd6560e..b8bf09f550 100644
>> --- a/nptl/pthread_getcpuclockid.c
>> +++ b/nptl/pthread_getcpuclockid.c
>> @@ -35,7 +35,7 @@ __pthread_getcpuclockid (pthread_t threadid, clockid_t *clockid)
>>
>> /* The clockid_t value is a simple computation from the TID. */
>>
>> - const clockid_t tidclock = MAKE_THREAD_CPUCLOCK (pd->tid, CPUCLOCK_SCHED);
>> + const clockid_t tidclock = make_thread_cpuclock (pd->tid, CPUCLOCK_SCHED);
>>
>> *clockid = tidclock;
>> return 0;
>
> OK. Use the new function instead of the old macro.
>
>> diff --git a/sysdeps/unix/sysv/linux/clock_getcpuclockid.c b/sysdeps/unix/sysv/linux/clock_getcpuclockid.c
>> index 5534127ed7..355d3c86af 100644
>> --- a/sysdeps/unix/sysv/linux/clock_getcpuclockid.c
>> +++ b/sysdeps/unix/sysv/linux/clock_getcpuclockid.c
>> @@ -29,7 +29,7 @@ __clock_getcpuclockid (pid_t pid, clockid_t *clock_id)
>> /* The clockid_t value is a simple computation from the PID.
>> But we do a clock_getres call to validate it. */
>>
>> - const clockid_t pidclock = MAKE_PROCESS_CPUCLOCK (pid, CPUCLOCK_SCHED);
>> + const clockid_t pidclock = make_process_cpuclock (pid, CPUCLOCK_SCHED);
>>
>> #ifndef __NR_clock_getres_time64
>> # define __NR_clock_getres_time64 __NR_clock_getres
>
> OK. Same.
>
>> diff --git a/sysdeps/unix/sysv/linux/clock_nanosleep.c b/sysdeps/unix/sysv/linux/clock_nanosleep.c
>> index befe6ecb8c..e610fd4e8d 100644
>> --- a/sysdeps/unix/sysv/linux/clock_nanosleep.c
>> +++ b/sysdeps/unix/sysv/linux/clock_nanosleep.c
>> @@ -34,7 +34,7 @@ __clock_nanosleep_time64 (clockid_t clock_id, int flags,
>> if (clock_id == CLOCK_THREAD_CPUTIME_ID)
>> return EINVAL;
>> if (clock_id == CLOCK_PROCESS_CPUTIME_ID)
>> - clock_id = MAKE_PROCESS_CPUCLOCK (0, CPUCLOCK_SCHED);
>> + clock_id = PROCESS_CLOCK;
>>
>> /* If the call is interrupted by a signal handler or encounters an error,
>> it returns a positive value similar to errno. */
>
> OK. PROCESS_CLOCK expands to the appropriate call to make_process_cpuclock.
>
>> diff --git a/sysdeps/unix/sysv/linux/kernel-posix-cpu-timers.h b/sysdeps/unix/sysv/linux/kernel-posix-cpu-timers.h
>> index 164a90ddeb..bea1e0e62d 100644
>> --- a/sysdeps/unix/sysv/linux/kernel-posix-cpu-timers.h
>> +++ b/sysdeps/unix/sysv/linux/kernel-posix-cpu-timers.h
>> @@ -1,4 +1,12 @@
>> -/* Parameters for the Linux kernel ABI for CPU clocks. */
>> +/*
>> + Parameters for the Linux kernel ABI for CPU clocks, the bit fields within
>> + a clockid:
>> +
>> + - The most significant 29 bits hold either a pid or a file descriptor.
>> + - Bit 2 indicates whether a cpu clock refers to a thread or a process.
>> + - Bits 1 and 0 give the type: PROF=0, VIRT=1, SCHED=2, or FD=3.
>> + - A clockid is invalid if bits 2, 1, and 0 are all set.
>> + */
>>
>> #define CPUCLOCK_PID(clock) ((pid_t) ~((clock) >> 3))
>> #define CPUCLOCK_PERTHREAD(clock) \
>
> OK. Comment lines up with the one in Linux's "include/linux/posix-timers.h".
>
>> @@ -12,7 +20,17 @@
>> #define CPUCLOCK_SCHED 2
>> #define CPUCLOCK_MAX 3
>>
>> -#define MAKE_PROCESS_CPUCLOCK(pid, clock) \
>> - ((~(clockid_t) (pid) << 3) | (clockid_t) (clock))
>
> OK. clockid_t is a signed type, thus expanding with pid = 0 leads to a
> left-shift on a negative value.
>
>> -#define MAKE_THREAD_CPUCLOCK(tid, clock) \
>> - MAKE_PROCESS_CPUCLOCK((tid), (clock) | CPUCLOCK_PERTHREAD_MASK)
>
> OK. Same macro used. Same issue.
>
>> +static inline clockid_t
>> +make_process_cpuclock (unsigned int pid, clockid_t clock)
>> +{
>> + return ((~pid) << 3) | clock;
>> +}
>> +
>
> OK. Replacement function that uses unsigned. Lines up with the version
> in Linux's sources.
>
>> +static inline clockid_t
>> +make_thread_cpuclock (unsigned int tid, clockid_t clock)
>> +{
>> + return make_process_cpuclock (tid, clock | CPUCLOCK_PERTHREAD_MASK);
>> +}
>> +
>
> OK. Replaces the second macro for tid.
>
>> +#define PROCESS_CLOCK make_process_cpuclock (0, CPUCLOCK_SCHED)
>> +#define THREAD_CLOCK make_thread_cpuclock (0, CPUCLOCK_SCHED)
>
> OK. Macros for convenience.
>
>> diff --git a/sysdeps/unix/sysv/linux/timer_create.c b/sysdeps/unix/sysv/linux/timer_create.c
>> index a8b2a41d9e..290324a7ea 100644
>> --- a/sysdeps/unix/sysv/linux/timer_create.c
>> +++ b/sysdeps/unix/sysv/linux/timer_create.c
>> @@ -33,9 +33,9 @@ ___timer_create (clockid_t clock_id, struct sigevent *evp, timer_t *timerid)
>> {
>> {
>> clockid_t syscall_clockid = (clock_id == CLOCK_PROCESS_CPUTIME_ID
>> - ? MAKE_PROCESS_CPUCLOCK (0, CPUCLOCK_SCHED)
>> + ? PROCESS_CLOCK
>> : clock_id == CLOCK_THREAD_CPUTIME_ID
>> - ? MAKE_THREAD_CPUCLOCK (0, CPUCLOCK_SCHED)
>> + ? THREAD_CLOCK
>> : clock_id);
>>
>> /* If the user wants notification via a thread we need to handle
>
> OK. Use the new macros.
>
prev parent reply other threads:[~2022-10-20 13:39 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-30 12:08 Adhemerval Zanella
2022-10-17 22:33 ` Arjun Shankar
2022-10-20 13:39 ` Adhemerval Zanella Netto [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ee974e11-179d-b701-fbf6-b48392a42a4c@linaro.org \
--to=adhemerval.zanella@linaro.org \
--cc=arjun@redhat.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).