Bug 9550 - sigprocmask does not work on FreeBSD to stop further signals in a signal handler
Summary: sigprocmask does not work on FreeBSD to stop further signals in a signal handler
Status: RESOLVED FIXED
Alias: None
Product: Samba 3.6
Classification: Unclassified
Component: File services (show other bugs)
Version: unspecified
Hardware: All All
: P5 normal
Target Milestone: ---
Assignee: Karolin Seeger
QA Contact: Samba QA Contact
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-01-08 19:54 UTC by Richard Sharpe
Modified: 2020-12-11 08:28 UTC (History)
1 user (show)

See Also:

Attachments
git-am patchset for 4.0.x and master. (6.97 KB, patch)
2013-01-14 23:25 UTC, Jeremy Allison 		rsharpe: review+
Details
git-am patchset for 3.6.next. (5.36 KB, patch)
2013-01-15 18:38 UTC, Jeremy Allison 		rsharpe: review+
Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Richard Sharpe 2013-01-08 19:54:00 UTC 
The code in lib/tevent/tevent_signal.c:tevent_common_signal_handler_info does not correctly switch off further signals when TEVENT_SA_INFO_QUEUE_COUNT is reached.

This is at least the case for FreeBSD 8.0. I am not sure yet about subsequent releases.

The code is:

        if (count+1 == TEVENT_SA_INFO_QUEUE_COUNT) {
                /* we've filled the info array - block this signal until
                   these ones are delivered */
                sigset_t set;
                sigemptyset(&set);
                sigaddset(&set, signum);
                sigprocmask(SIG_BLOCK, &set, NULL);
                TEVENT_SIG_INCREMENT(sig_state->sig_blocked[signum]);
        }

Under FreeBSD 8.0 we have to use:

                sigaddset(&((ucontext_t *)uctx)->uc_sigmask, signum);

and include <sys/ucontext.h>.

It seems that we need a configure test or something for this, but I am not sure how to detect it.
Comment 1 Richard Sharpe 2013-01-08 20:51:48 UTC 
Perhaps we need a symbol like NEED_FREEBSD_STYLE_SIGBLOCK in config.h.
Comment 2 Richard Sharpe 2013-01-08 21:03:00 UTC 
Perhaps something like:

dnl We need to use a fugly hack to stop further signals on FreeBSD
AC_MSG_CHECKING([whether to use FreeBSD Style signal stopping])
case "$host" in
*-freebsd*)
	AC_DEFINE(NEED_FREEBSD_SIGSTOP, "yes",
			[need FreeBSD style signal stopping])
	;;
esac
Comment 3 Jeremy Allison 2013-01-14 23:25:28 UTC 
Created attachment 8429 [details]
git-am patchset for 4.0.x and master.

Richard, please review and push to master if you're happy with this.

I'll also create a 3.6.x version.

Jeremy.
Comment 4 Jeremy Allison 2013-01-15 18:38:36 UTC 
Created attachment 8434 [details]
git-am patchset for 3.6.next.
Comment 5 Richard Sharpe 2013-01-15 19:08:42 UTC 
Comment on attachment 8429 [details]
git-am patchset for 4.0.x and master.

Looks good to me
Comment 6 Richard Sharpe 2013-01-15 19:09:49 UTC 
(In reply to comment #3)
> Created attachment 8429 [details]
> git-am patchset for 4.0.x and master.
> 
> Richard, please review and push to master if you're happy with this.

I should be able to push this tonight.
Comment 7 Jeremy Allison 2013-01-15 19:38:10 UTC 
Richard,

Metze already pushed this (and Karolin already pushed to 4.0.next).

Just give the 3.6 version a +1 and I'll re-assign to Karolin for inclusion.

Cheers,

Jeremy.
Comment 8 Richard Sharpe 2013-01-15 19:43:37 UTC 
Comment on attachment 8434 [details]
git-am patchset for 3.6.next.

Looks good.
Comment 9 Jeremy Allison 2013-01-15 19:47:40 UTC 
Re-assigning to Karolin for inclusion in 3.6.next.

Jeremy.
Comment 10 Karolin Seeger 2013-01-16 08:53:30 UTC 
Pushed to v3-6-test.
Closing out bug report.

Thanks!
Comment 11 Stefan Metzmacher 2013-01-16 11:06:36 UTC 
This is needed for 4.0 too
Comment 12 Karolin Seeger 2013-01-17 08:04:48 UTC 
(In reply to comment #11)
> This is needed for 4.0 too

Pushed to autobuild-v4-0-test.
Comment 13 Karolin Seeger 2013-01-18 08:39:05 UTC 
Pushed to v4-0-test.
Closing out bug report.

Thanks!