Your message dated Sat, 04 Oct 2008 21:02:10 +0000
with message-id <[EMAIL PROTECTED]>
and subject line Bug#499569: fixed in gnupg2 2.0.9-3.1
has caused the Debian Bug report #499569,
regarding gnupg-agent: modifies SigBlk mask of all processes spawned in the X
session breaking unrelated software [i386 only?]
to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact [EMAIL PROTECTED]
immediately.)
--
499569: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=499569
Debian Bug Tracking System
Contact [EMAIL PROTECTED] with problems
--- Begin Message ---
Package: gnupg-agent
Version: 2.0.9-3
Severity: critical
gnupg-agent somehow manages to change the SigBlk mask causing all child
processes that are spawned part of the X session to have the SigBlk
mask set causing real-time signals to be blocked. This breaks all
applications spawned in that X session that rely on
real-time signals.
As this breaks unrelated software I set the severity to critical.
Here the summary of my full investigation:
About 2 months ago I started to suffer from an issue that first looked
like a bug in Mono's threading implementation. The test-case to
reproduce it looks like this:
Attachment:
test-thread-abort.cs
Compile:
mcs test-thread-abort.cs
Run:
mono test-thread-abort.exe
My output:
thread.Start()
Worker(): thread started
Worker(): working... 0
Worker(): working... 1
Worker(): working... 2
Worker(): working... 3
Worker(): working... 4
thread.Abort()
press enter to quit!
Worker(): working... 5
Worker(): working... 6
Worker(): working... 7
Worker(): working... 8
Worker(): working... 9
Worker(): working... 10
Worker(): working... 11
Worker(): working... 12
(continuing forever -> killed using ^C)
The expected output:
thread.Start()
Worker(): thread started
Worker(): working... 0
Worker(): working... 1
Worker(): working... 2
Worker(): working... 3
Worker(): working... 4
thread.Abort()
press enter to quit!
Worker(): thread aborted!
Worker(): thread ended
The thread was never aborted and thus forever continued on my system.
After more investigation in Mono's threading implementation I found
that it uses real-time signals to control the threads. A strace then
showed that the signal (SIGRT1) was send but never received.... This
made me really curious.
So I wrote a C test-case for this issue, which looks like this:
Attachment:
test-signal.c
Compile:
gcc test-signal.c -o test-signal
Run:
./test-signal
My Output:
SIGRTMIN: 34
sigismember(): 1
SIGRTMIN is blocked, trying to unblock...done.
sigismember(): 0
SIGRT_1 received
SIGRT_2 received
SIGQUIT received
The expected output:
SIGRTMIN: 34
sigismember(): 0
SIGRT_1 received
SIGRT_2 received
SIGQUIT received
And that proved the signal (SIGRT) is indeed already blocked when the
application is started.
Then I found out / learned that blocking signals is a feature and the
SigBlk mask is inherited to child processes. So something was setting
this mask, and I could not find out what it was till today.
I reinstalled my system (1 week ago) after I could not find the exact
cause of this issue (after 3 continuous days of investigation)
and today it started to happen again, after I was sponsoring a package
upload, for which I had to install pscd and gnupg-agent to get my
crypto card working again (after the reinstall).
Then I disabled the gpg-agent in my ~/.gnupg/gpg.conf, restarted my X
session and viola SigBlk mask is back to normal and all my lovely
applications work again like they should.
Enabling the gpg-agent in ~/.gnupg/gpg.conf again brings the issue back.
I noticed the gpg-agent is started in /etc/X11/Xsession.d/90gpg-agent,
but I no have no idea how it manages to change (or rather inherits) the
SigBlk mask of other processes spawned in the X session....
Here also a little detection script I wrote to identify processes with
that set SigBlk mask (it also finds some "false-positives" as some apps
really need that mask, but not _all_ processes launched under the X
session!)
Attachment:
proc-sigblk.sh:
My output:
Name: x-session-manag
Pid: 21699
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: dbus-launch
Pid: 21757
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: dbus-daemon
Pid: 21758
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: seahorse-agent
Pid: 21764
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-settings-
Pid: 21769
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: metacity
Pid: 21781
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-panel
Pid: 21782
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: nautilus
Pid: 21783
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-screensav
Pid: 21786
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-vfs-daemo
Pid: 21795
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: bluetooth-apple
Pid: 21798
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-do
Pid: 21799
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: update-notifier
Pid: 21803
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-do
Pid: 21804
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: system-config-p
Pid: 21811
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: kerneloops-appl
Pid: 21815
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-volume-ma
Pid: 21817
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: nm-applet
Pid: 21819
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-power-man
Pid: 21827
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: multiload-apple
Pid: 21849
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: mixer_applet2
Pid: 21852
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gweather-applet
Pid: 21855
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: locate
Pid: 21862
Uid: 1000 1000 1000 1000
Gid: 1000 105 105 105
Name: locate
Pid: 21863
Uid: 1000 1000 1000 1000
Gid: 1000 105 105 105
Name: mapping-daemon
Pid: 21866
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-terminal
Pid: 21868
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: gnome-pty-helpe
Pid: 21871
Uid: 1000 1000 1000 1000
Gid: 1000 43 43 43
Name: bonobo-activati
Pid: 3878
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
Name: egrep
Pid: 22046
Uid: 1000 1000 1000 1000
Gid: 1000 1000 1000 1000
The expected output:
(none or a very small set of applications like gconfd-2 or some
gnome applets)
--
Regards,
Mirco 'meebey' Bauer
PGP-Key ID: 0xEEF946C8
FOSS Developer [EMAIL PROTECTED] http://www.meebey.net/
PEAR Developer [EMAIL PROTECTED] http://pear.php.net/
Debian Developer [EMAIL PROTECTED] http://www.debian.org/
#include <stdio.h>
#include <signal.h>
void sigint_handler(void);
void sigquit_handler(void);
void sigrt1_handler(void);
void sigrt2_handler(void);
void sigrt3_handler(void);
void sigpwr_handler(void);
main()
{
printf("SIGRTMIN: %i\n", SIGRTMIN);
sigset_t set, old_set, new_set;
int ismember;
if (sigprocmask(SIG_SETMASK, NULL, &old_set) != 0)
printf("sigprocmask() failed!\n");
ismember = sigismember(&old_set, SIGRTMIN);
printf("sigismember(): %i\n", ismember);
if (ismember) {
printf("SIGRTMIN is blocked, trying to unblock...");
sigemptyset(&new_set);
sigaddset(&new_set, SIGRTMIN);
sigaddset(&new_set, SIGRTMIN+1);
if (sigprocmask(SIG_UNBLOCK, &new_set, NULL) != 0) {
printf("failed!\n");
exit(1);
}
printf("done.\n");
sigemptyset(&set);
ismember = sigismember(&set, SIGRTMIN);
printf("sigismember(): %i\n", ismember);
}
signal(SIGINT, sigint_handler);
signal(SIGQUIT, sigquit_handler);
signal(SIGRTMIN, sigrt1_handler);
signal(SIGRTMIN+1, sigrt2_handler);
signal(SIGRTMIN+2, sigrt3_handler);
signal(SIGPWR, sigpwr_handler);
kill(getpid(), SIGRTMIN);
kill(getpid(), SIGRTMIN+1);
kill(getpid(), SIGQUIT);
for(;;); /* infinite loop */
}
void sigint_handler(void)
{
signal(SIGINT, sigint_handler);
/* NOTE some versions of UNIX will reset signal to default
after each call. So for portability reset signal each time */
printf("SIGINT received\n");
}
void sigquit_handler(void)
{
printf("SIGQUIT received\n");
exit(0); /* normal exit status */
}
void sigrt1_handler(void)
{
printf("SIGRT_1 received\n");
}
void sigrt2_handler(void)
{
printf("SIGRT_2 received\n");
}
void sigrt3_handler(void)
{
printf("SIGRT_3 received\n");
}
void sigpwr_handler(void)
{
printf("SIGPWR received\n");
}
using System;
using System.Threading;
using System.Runtime.InteropServices;
class MainClass
{
static int loopCount;
static void Main()
{
Thread thread = new Thread(new ThreadStart(Worker));
Console.WriteLine("thread.Start()");
thread.IsBackground = true;
thread.Start();
// give the thread some time to spawn
Thread.Sleep(5000);
Console.WriteLine("thread.Abort()");
thread.Abort();
Console.WriteLine("press enter to quit!");
Console.ReadLine();
}
static void Worker()
{
Console.WriteLine("Worker(): thread started");
try {
while (true) {
Console.WriteLine("Worker(): working... " + loopCount++); Thread.Sleep(1000);
}
} catch (ThreadAbortException) {
Thread.ResetAbort();
Console.WriteLine("Worker(): thread aborted!");
}
Console.WriteLine("Worker(): thread ended");
}
}
proc-sigblk.sh
Description: application/shellscript
--- End Message ---
--- Begin Message ---
Source: gnupg2
Source-Version: 2.0.9-3.1
We believe that the bug you reported is fixed in the latest version of
gnupg2, which is due to be installed in the Debian FTP archive:
gnupg-agent_2.0.9-3.1_amd64.deb
to pool/main/g/gnupg2/gnupg-agent_2.0.9-3.1_amd64.deb
gnupg2_2.0.9-3.1.diff.gz
to pool/main/g/gnupg2/gnupg2_2.0.9-3.1.diff.gz
gnupg2_2.0.9-3.1.dsc
to pool/main/g/gnupg2/gnupg2_2.0.9-3.1.dsc
gnupg2_2.0.9-3.1_amd64.deb
to pool/main/g/gnupg2/gnupg2_2.0.9-3.1_amd64.deb
gpgsm_2.0.9-3.1_amd64.deb
to pool/main/g/gnupg2/gpgsm_2.0.9-3.1_amd64.deb
A summary of the changes between this version and the previous one is
attached.
Thank you for reporting the bug, which will now be closed. If you
have further comments please address them to [EMAIL PROTECTED],
and the maintainer will reopen the bug report if appropriate.
Debian distribution maintenance software
pp.
Thomas Viehmann <[EMAIL PROTECTED]> (supplier of updated gnupg2 package)
(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing [EMAIL PROTECTED])
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Format: 1.8
Date: Sat, 04 Oct 2008 10:25:53 +0200
Source: gnupg2
Binary: gnupg-agent gpgsm gnupg2
Architecture: source amd64
Version: 2.0.9-3.1
Distribution: unstable
Urgency: low
Maintainer: Eric Dorland <[EMAIL PROTECTED]>
Changed-By: Thomas Viehmann <[EMAIL PROTECTED]>
Description:
gnupg-agent - GNU privacy guard - password agent
gnupg2 - GNU privacy guard - a free PGP replacement
gpgsm - GNU privacy guard - S/MIME version
Closes: 499569
Changes:
gnupg2 (2.0.9-3.1) unstable; urgency=low
.
* Non-maintainer upload.
* agent/gpg-agent.c: Deinit the threading library before exec'ing
the command to run in --daemon mode. And because that still doesn't
restore the sigprocmask, do that manually. Closes: #499569
Checksums-Sha1:
7ea431c20af542c390549d3c47ca167d9a25a6ec 1362 gnupg2_2.0.9-3.1.dsc
76449de5f5a4b57bb3e3207cfd9fe26926041295 40261 gnupg2_2.0.9-3.1.diff.gz
57560a2160cd63949467c4ef143f5303bd7312ff 316324 gnupg-agent_2.0.9-3.1_amd64.deb
a67be8bc8a5e6b7fc36416499427c7c53514a620 463176 gpgsm_2.0.9-3.1_amd64.deb
3094faebcc1ff5bc3e0a95df24e1c218e196a4d5 2183752 gnupg2_2.0.9-3.1_amd64.deb
Checksums-Sha256:
531ea81e97b88f275880a08075360bf6ec3343544958696eddf8138dcfb25cfa 1362
gnupg2_2.0.9-3.1.dsc
82aa02d999376c2c19695e408a4ed68f707c2246d8207c576edfa3e06a261272 40261
gnupg2_2.0.9-3.1.diff.gz
9e5ba276ddfcff41810f92df2adc4f8b6376762a132e458c1c7fbc28bfbbda48 316324
gnupg-agent_2.0.9-3.1_amd64.deb
f8771d612bd630c36b243a0a4be8886c7c382bac146c865e5be4ea58d84c89e6 463176
gpgsm_2.0.9-3.1_amd64.deb
046287105a95b3ab02156da035f12b8c1f2b56f2a754c5667c6be388fc9d28b2 2183752
gnupg2_2.0.9-3.1_amd64.deb
Files:
f7a7fa98ed5409bc91f740df4d8e3c55 1362 utils optional gnupg2_2.0.9-3.1.dsc
32bcaadb77e5755d335076dcd6ee49eb 40261 utils optional gnupg2_2.0.9-3.1.diff.gz
978cad7ec48e0188e11e548251a99707 316324 utils optional
gnupg-agent_2.0.9-3.1_amd64.deb
c97f44db2d3084c78391759bc16ecdd0 463176 utils optional
gpgsm_2.0.9-3.1_amd64.deb
0897f52a39b25347be1a92d40d77ca4e 2183752 utils optional
gnupg2_2.0.9-3.1_amd64.deb
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
iEYEARECAAYFAkjnvmwACgkQriZpaaIa1PlsFgCgwLMSQp3EKsdhqy3kII4fEOyV
uWwAn2MrMEzNOtAnoi5tUBWlaTSvFXZ3
=T+U2
-----END PGP SIGNATURE-----
--- End Message ---
