This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Refactor Linux raise implementation (BZ#15368)
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: Zack Weinberg <zackw at panix dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 29 Jun 2016 09:57:25 -0300
- Subject: Re: [PATCH] Refactor Linux raise implementation (BZ#15368)
- Authentication-results: sourceware.org; auth=none
- References: <1466188988-19954-1-git-send-email-adhemerval dot zanella at linaro dot org> <CAKCAbMgJvz1QagpX8kAtCjEgJE8D0SPABvcJ6VFZ-Vt+cDMTNA at mail dot gmail dot com> <57658427 dot 2090902 at linaro dot org>
Ping.
On 18/06/2016 14:25, Adhemerval Zanella wrote:
> Thanks for reviews, comment below:
>
> On 18/06/2016 11:09, Zack Weinberg wrote:
>> On Fri, Jun 17, 2016 at 2:43 PM, Adhemerval Zanella
>> <adhemerval.zanella@linaro.org> wrote:
>>> This patch changes both the nptl and libc Linux raise implementation
>>> to avoid the issues described in BZ#15368.
>>
>> It would be nice to have comments in these files which explain why it
>> is necessary to block all signals and why the cached values are not
>> safe to use. The old comment that you deleted ("raise is async-safe
>> and could be called while fork/vfork temporarily invalidated the PID")
>> would be a good start.
>>
>
> Indeed, I did not remove them in this new version below. I also extended
> the comment a bit explaining why to block the application signals.
>
>>> The strategy used is
>>> summarized in bug report first comment:
>>>
>>> 1. Block all signals (including internal NPTL ones);
>>
>> The code appears to block all signals *except* the internal NPTL ones.
>> If I understand Rich's description of the problem correctly, that is
>> wrong.
>
> Right, we need to block only user defined handler that might fork/vfork.
> The call itself is OK, but I will change the comments describing why there
> is no need to block GLIBC internal handlers.
>
>>
>>> +static inline int
>>> +__libc_signal_block_all (const sigset_t *set)
>>> +{
>>
>> Type-safety error here: the 'set' argument is written to and should
>> not be 'const'. It compiles because INTERNAL_SYSCALL() erases types,
>> but the compiler would be entitled to assume that 'set' is unchanged
>> after the call.
>
> Ack.
>
>>
>>> +static inline int
>>> +__libc_signal_block_app (const sigset_t *set)
>>
>> Same type-safety error here.
>
> Ack.
>
>>
>>> + sigset_t set;
>>> + __libc_signal_block_app (&set);
>>> +
>>> + pid_t pid = __getpid ();
>>
>> If I'm reading the code correctly, __getpid does *not* bypass the cache.
>>
>>> + pid_t tid = INLINE_SYSCALL (gettid, 0);
>>
>> INTERNAL_SYSCALL() here too? gettid() can't fail, so there's no point
>> generating the code to potentially set errno.
>>
>
> Right, I will use INTERNAL_SYSCALL.
>
>> zw
>>
>
> I think this version addresses all the issues you raised:
>
> --
>
> * sysdeps/unix/sysv/linux/nptl-signals.h
> (__nptl_clear_internal_signals): New function.
> (__libc_signal_block_all): Likewise.
> (__libc_signal_block_app): Likewise.
> (__libc_signal_restore_set): Likewise.
> * sysdeps/unix/sysv/linux/pt-raise.c (raise): Use Linux raise.c
> implementation.
> * sysdeps/unix/sysv/linux/raise.c (raise): Reimplement to not use
> the cached pid/tid value in pthread structure.
>
> --
>
> diff --git a/sysdeps/unix/sysv/linux/nptl-signals.h b/sysdeps/unix/sysv/linux/nptl-signals.h
> index 01f34c2..6525b6d 100644
> --- a/sysdeps/unix/sysv/linux/nptl-signals.h
> +++ b/sysdeps/unix/sysv/linux/nptl-signals.h
> @@ -39,5 +39,42 @@ __nptl_is_internal_signal (int sig)
> return (sig == SIGCANCEL) || (sig == SIGTIMER) || (sig == SIGSETXID);
> }
>
> +static inline void
> +__nptl_clear_internal_signals (sigset_t *set)
> +{
> + __sigdelset (set, SIGCANCEL);
> + __sigdelset (set, SIGTIMER);
> + __sigdelset (set, SIGSETXID);
> +}
> +
> +#define SIGALL_SET \
> + ((__sigset_t) { .__val = {[0 ... _SIGSET_NWORDS-1 ] = -1 } })
> +
> +static inline int
> +__libc_signal_block_all (sigset_t *set)
> +{
> + INTERNAL_SYSCALL_DECL (err);
> + return INTERNAL_SYSCALL (rt_sigprocmask, err, 4, SIG_BLOCK, &SIGALL_SET,
> + set, _NSIG / 8);
> +}
> +
> +static inline int
> +__libc_signal_block_app (sigset_t *set)
> +{
> + sigset_t allset = SIGALL_SET;
> + __nptl_clear_internal_signals (&allset);
> + INTERNAL_SYSCALL_DECL (err);
> + return INTERNAL_SYSCALL (rt_sigprocmask, err, 4, SIG_BLOCK, &allset, set,
> + _NSIG / 8);
> +}
> +
> +static inline int
> +__libc_signal_restore_set (const sigset_t *set)
> +{
> + INTERNAL_SYSCALL_DECL (err);
> + return INTERNAL_SYSCALL (rt_sigprocmask, err, 4, SIG_SETMASK, set, NULL,
> + _NSIG / 8);
> +}
> +
> /* Used to communicate with signal handler. */
> extern struct xid_command *__xidcmd attribute_hidden;
> diff --git a/sysdeps/unix/sysv/linux/pt-raise.c b/sysdeps/unix/sysv/linux/pt-raise.c
> index 715bbe9..5f6dea1 100644
> --- a/sysdeps/unix/sysv/linux/pt-raise.c
> +++ b/sysdeps/unix/sysv/linux/pt-raise.c
> @@ -1,4 +1,5 @@
> -/* Copyright (C) 2002-2016 Free Software Foundation, Inc.
> +/* ISO C raise function for libpthread.
> + Copyright (C) 2002-2016 Free Software Foundation, Inc.
> This file is part of the GNU C Library.
> Contributed by Ulrich Drepper <drepper@redhat.com>, 2002.
>
> @@ -16,22 +17,4 @@
> License along with the GNU C Library; if not, see
> <http://www.gnu.org/licenses/>. */
>
> -#include <errno.h>
> -#include <signal.h>
> -#include <sysdep.h>
> -#include <tls.h>
> -
> -
> -int
> -raise (int sig)
> -{
> - /* raise is an async-safe function. It could be called while the
> - fork function temporarily invalidated the PID field. Adjust for
> - that. */
> - pid_t pid = THREAD_GETMEM (THREAD_SELF, pid);
> - if (__glibc_unlikely (pid < 0))
> - pid = -pid;
> -
> - return INLINE_SYSCALL (tgkill, 3, pid, THREAD_GETMEM (THREAD_SELF, tid),
> - sig);
> -}
> +#include <sysdeps/unix/sysv/linux/raise.c>
> diff --git a/sysdeps/unix/sysv/linux/raise.c b/sysdeps/unix/sysv/linux/raise.c
> index 3795e6e..d386426 100644
> --- a/sysdeps/unix/sysv/linux/raise.c
> +++ b/sysdeps/unix/sysv/linux/raise.c
> @@ -16,42 +16,34 @@
> License along with the GNU C Library; if not, see
> <http://www.gnu.org/licenses/>. */
>
> -#include <errno.h>
> -#include <limits.h>
> #include <signal.h>
> #include <sysdep.h>
> -#include <nptl/pthreadP.h>
> -
> +#include <errno.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <nptl-signals.h>
>
> int
> raise (int sig)
> {
> - struct pthread *pd = THREAD_SELF;
> - pid_t pid = THREAD_GETMEM (pd, pid);
> - pid_t selftid = THREAD_GETMEM (pd, tid);
> - if (selftid == 0)
> - {
> - /* This system call is not supposed to fail. */
> -#ifdef INTERNAL_SYSCALL
> - INTERNAL_SYSCALL_DECL (err);
> - selftid = INTERNAL_SYSCALL (gettid, err, 0);
> -#else
> - selftid = INLINE_SYSCALL (gettid, 0);
> -#endif
> - THREAD_SETMEM (pd, tid, selftid);
> -
> - /* We do not set the PID field in the TID here since we might be
> - called from a signal handler while the thread executes fork. */
> - pid = selftid;
> - }
> - else
> - /* raise is an async-safe function. It could be called while the
> - fork/vfork function temporarily invalidated the PID field. Adjust for
> - that. */
> - if (__glibc_unlikely (pid <= 0))
> - pid = (pid & INT_MAX) == 0 ? selftid : -pid;
> -
> - return INLINE_SYSCALL (tgkill, 3, pid, selftid, sig);
> + /* raise is an async-safe function so it could be called while the
> + fork/vfork function temporarily invalidated the PID field. To avoid
> + relying in the cached value we block all user-defined signal handler
> + (which might call fork/vfork) and issues the getpid and gettid
> + directly. */
> +
> + sigset_t set;
> + __libc_signal_block_app (&set);
> +
> + INTERNAL_SYSCALL_DECL (err);
> + pid_t pid = INTERNAL_SYSCALL (getpid, err, 0);
> + pid_t tid = INTERNAL_SYSCALL (gettid, err, 0);
> +
> + int ret = INLINE_SYSCALL (tgkill, 3, pid, tid, sig);
> +
> + __libc_signal_restore_set (&set);
> +
> + return ret;
> }
> libc_hidden_def (raise)
> weak_alias (raise, gsignal)
>