[GIT PULL] signal fixes for v5.0-rc6

[GIT PULL] signal fixes for v5.0-rc6

[Eric W. Biederman]

Linus,

Please pull the for-linus branch from the git tree:

   git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git for-linus

   HEAD:7146db3317c67b517258cb5e1b08af387da0618b signal: Better detection of synchronous signals

This contains 4 small fixes for signal handling.  A missing range check,
a regression fix, prioritizing signals we have already started a signal
group exit for, and better detection of synchronous signals.

The confused decision of which signals to handle failed spectacularly
when a timer was pointed at SIGBUS and the stack overflowed.  Resulting
in an unkillable process in an infinite loop instead of a SIGSEGV and
core dump.

Eric W. Biederman (4):
      signal: Make siginmask safe when passed a signal of 0
      signal: Always attempt to allocate siginfo for SIGSTOP
      signal: Always notice exiting tasks
      signal: Better detection of synchronous signals

 include/linux/signal.h |  2 +-
 kernel/signal.c        | 63 ++++++++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 60 insertions(+), 5 deletions(-)

Eric

Re: [GIT PULL] signal fixes for v5.0-rc6

[pr-tracker-bot]

The pull request you sent on Fri, 08 Feb 2019 09:52:02 -0600:

> git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace.git for-linus

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/6b2912cedc238c984e6a4039836ed9c911121076

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.wiki.kernel.org/userdoc/prtracker

Re: [GIT PULL] signal fixes for v5.0-rc6

[Oleg Nesterov]

On 02/08, Eric W. Biederman wrote:
>
>       signal: Always notice exiting tasks

Sorry, I could not look at these patches before...

So I think this one should be reverted, it makes PTRACE_EVENT_EXIT even more broken.

Once again, why we can't do something like

	--- x/kernel/signal.c
	+++ x/kernel/signal.c
	@@ -209,6 +209,8 @@ int next_signal(struct sigpending *pendi
		 */
		x = *s &~ *m;
		if (x) {
	+		if (x & sigmask(SIGKILL))
	+			return SIGKILL;
			if (x & SYNCHRONOUS_MASK)
				x &= SYNCHRONOUS_MASK;
			sig = ffz(~x) + 1;

instead? at least for the start.

Oleg.