linux-stable.git/kernel/signal.c, branch v4.9.232

signal: Extend exec_id to 64bits

2020-04-24T05:58:54+00:00

commit d1e7fd6462ca9fc76650fbe6ca800e35b24267da upstream.

Replace the 32bit exec_id with a 64bit exec_id to make it impossible
to wrap the exec_id counter.  With care an attacker can cause exec_id
wrap and send arbitrary signals to a newly exec'd parent.  This
bypasses the signal sending checks if the parent changes their
credentials during exec.

The severity of this problem can been seen that in my limited testing
of a 32bit exec_id it can take as little as 19s to exec 65536 times.
Which means that it can take as little as 14 days to wrap a 32bit
exec_id.  Adam Zabrocki has succeeded wrapping the self_exe_id in 7
days.  Even my slower timing is in the uptime of a typical server.
Which means self_exec_id is simply a speed bump today, and if exec
gets noticably faster self_exec_id won't even be a speed bump.

Extending self_exec_id to 64bits introduces a problem on 32bit
architectures where reading self_exec_id is no longer atomic and can
take two read instructions.  Which means that is is possible to hit
a window where the read value of exec_id does not match the written
value.  So with very lucky timing after this change this still
remains expoiltable.

I have updated the update of exec_id on exec to use WRITE_ONCE
and the read of exec_id in do_notify_parent to use READ_ONCE
to make it clear that there is no locking between these two
locations.

Link: https://lore.kernel.org/kernel-hardening/20200324215049.GA3710@pi3.com.pl
Fixes: 2.3.23pre2
Cc: stable@vger.kernel.org
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Greg Kroah-Hartman

signal: avoid double atomic counter increments for user accounting

2020-03-20T08:07:57+00:00

[ Upstream commit fda31c50292a5062332fa0343c084bd9f46604d9 ]

When queueing a signal, we increment both the users count of pending
signals (for RLIMIT_SIGPENDING tracking) and we increment the refcount
of the user struct itself (because we keep a reference to the user in
the signal structure in order to correctly account for it when freeing).

That turns out to be fairly expensive, because both of them are atomic
updates, and particularly under extreme signal handling pressure on big
machines, you can get a lot of cache contention on the user struct.
That can then cause horrid cacheline ping-pong when you do these
multiple accesses.

So change the reference counting to only pin the user for the _first_
pending signal, and to unpin it when the last pending signal is
dequeued.  That means that when a user sees a lot of concurrent signal
queuing - which is the only situation when this matters - the only
atomic access needed is generally the 'sigpending' count update.

This was noticed because of a particularly odd timing artifact on a
dual-socket 96C/192T Cascade Lake platform: when you get into bad
contention, on that machine for some reason seems to be much worse when
the contention happens in the upper 32-byte half of the cacheline.

As a result, the kernel test robot will-it-scale 'signal1' benchmark had
an odd performance regression simply due to random alignment of the
'struct user_struct' (and pointed to a completely unrelated and
apparently nonsensical commit for the regression).

Avoiding the double increments (and decrements on the dequeueing side,
of course) makes for much less contention and hugely improved
performance on that will-it-scale microbenchmark.

Quoting Feng Tang:

 "It makes a big difference, that the performance score is tripled! bump
  from original 17000 to 54000. Also the gap between 5.0-rc6 and
  5.0-rc6+Jiri's patch is reduced to around 2%"

[ The "2% gap" is the odd cacheline placement difference on that
  platform: under the extreme contention case, the effect of which half
  of the cacheline was hot was 5%, so with the reduced contention the
  odd timing artifact is reduced too ]

It does help in the non-contended case too, but is not nearly as
noticeable.

Reported-and-tested-by: Feng Tang 
Cc: Eric W. Biederman 
Cc: Huang, Ying 
Cc: Philip Li 
Cc: Andi Kleen 
Cc: Jiri Olsa 
Cc: Peter Zijlstra 
Signed-off-by: Linus Torvalds 
Signed-off-by: Sasha Levin

signal: Allow cifs and drbd to receive their terminating signals

2020-01-29T09:24:30+00:00

[ Upstream commit 33da8e7c814f77310250bb54a9db36a44c5de784 ]

My recent to change to only use force_sig for a synchronous events
wound up breaking signal reception cifs and drbd.  I had overlooked
the fact that by default kthreads start out with all signals set to
SIG_IGN.  So a change I thought was safe turned out to have made it
impossible for those kernel thread to catch their signals.

Reverting the work on force_sig is a bad idea because what the code
was doing was very much a misuse of force_sig.  As the way force_sig
ultimately allowed the signal to happen was to change the signal
handler to SIG_DFL.  Which after the first signal will allow userspace
to send signals to these kernel threads.  At least for
wake_ack_receiver in drbd that does not appear actively wrong.

So correct this problem by adding allow_kernel_signal that will allow
signals whose siginfo reports they were sent by the kernel through,
but will not allow userspace generated signals, and update cifs and
drbd to call allow_kernel_signal in an appropriate place so that their
thread can receive this signal.

Fixing things this way ensures that userspace won't be able to send
signals and cause problems, that it is clear which signals the
threads are expecting to receive, and it guarantees that nothing
else in the system will be affected.

This change was partly inspired by similar cifs and drbd patches that
added allow_signal.

Reported-by: ronnie sahlberg 
Reported-by: Christoph Böhmwalder 
Tested-by: Christoph Böhmwalder 
Cc: Steve French 
Cc: Philipp Reisner 
Cc: David Laight 
Fixes: 247bc9470b1e ("cifs: fix rmmod regression in cifs.ko caused by force_sig changes")
Fixes: 72abe3bcf091 ("signal/cifs: Fix cifs_put_tcp_session to call send_sig instead of force_sig")
Fixes: fee109901f39 ("signal/drbd: Use send_sig not force_sig")
Fixes: 3cf5d076fb4d ("signal: Remove task parameter from force_sig")
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Sasha Levin

signal: Always ignore SIGKILL and SIGSTOP sent to the global init

2019-11-25T08:52:12+00:00

[ Upstream commit 86989c41b5ea08776c450cb759592532314a4ed6 ]

If the first process started (aka /sbin/init) receives a SIGKILL it
will panic the system if it is delivered.  Making the system unusable
and undebugable.  It isn't much better if the first process started
receives SIGSTOP.

So always ignore SIGSTOP and SIGKILL sent to init.

This is done in a separate clause in sig_task_ignored as force_sig_info
can clear SIG_UNKILLABLE and this protection should work even then.

Reviewed-by: Thomas Gleixner 
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Sasha Levin

kernel/signal.c: trace_signal_deliver when signal_group_exit

2019-06-11T10:22:43+00:00

commit 98af37d624ed8c83f1953b1b6b2f6866011fc064 upstream.

In the fixes commit, removing SIGKILL from each thread signal mask and
executing "goto fatal" directly will skip the call to
"trace_signal_deliver".  At this point, the delivery tracking of the
SIGKILL signal will be inaccurate.

Therefore, we need to add trace_signal_deliver before "goto fatal" after
executing sigdelset.

Note: SEND_SIG_NOINFO matches the fact that SIGKILL doesn't have any info.

Link: http://lkml.kernel.org/r/20190425025812.91424-1-weizhenliang@huawei.com
Fixes: cf43a757fd4944 ("signal: Restore the stop PTRACE_EVENT_EXIT")
Signed-off-by: Zhenliang Wei 
Reviewed-by: Christian Brauner 
Reviewed-by: Oleg Nesterov 
Cc: Eric W. Biederman 
Cc: Ivan Delalande 
Cc: Arnd Bergmann 
Cc: Thomas Gleixner 
Cc: Deepa Dinamani 
Cc: Greg Kroah-Hartman 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Greg Kroah-Hartman

signal: Restore the stop PTRACE_EVENT_EXIT

2019-02-20T09:18:33+00:00

commit cf43a757fd49442bc38f76088b70c2299eed2c2f upstream.

In the middle of do_exit() there is there is a call
"ptrace_event(PTRACE_EVENT_EXIT, code);" That call places the process
in TACKED_TRACED aka "(TASK_WAKEKILL | __TASK_TRACED)" and waits for
for the debugger to release the task or SIGKILL to be delivered.

Skipping past dequeue_signal when we know a fatal signal has already
been delivered resulted in SIGKILL remaining pending and
TIF_SIGPENDING remaining set.  This in turn caused the
scheduler to not sleep in PTACE_EVENT_EXIT as it figured
a fatal signal was pending.  This also caused ptrace_freeze_traced
in ptrace_check_attach to fail because it left a per thread
SIGKILL pending which is what fatal_signal_pending tests for.

This difference in signal state caused strace to report
strace: Exit of unknown pid NNNNN ignored

Therefore update the signal handling state like dequeue_signal
would when removing a per thread SIGKILL, by removing SIGKILL
from the per thread signal mask and clearing TIF_SIGPENDING.

Acked-by: Oleg Nesterov 
Reported-by: Oleg Nesterov 
Reported-by: Ivan Delalande 
Cc: stable@vger.kernel.org
Fixes: 35634ffa1751 ("signal: Always notice exiting tasks")
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Greg Kroah-Hartman

signal: Better detection of synchronous signals

2019-02-15T07:07:37+00:00

commit 7146db3317c67b517258cb5e1b08af387da0618b upstream.

Recently syzkaller was able to create unkillablle processes by
creating a timer that is delivered as a thread local signal on SIGHUP,
and receiving SIGHUP SA_NODEFERER.  Ultimately causing a loop failing
to deliver SIGHUP but always trying.

When the stack overflows delivery of SIGHUP fails and force_sigsegv is
called.  Unfortunately because SIGSEGV is numerically higher than
SIGHUP next_signal tries again to deliver a SIGHUP.

From a quality of implementation standpoint attempting to deliver the
timer SIGHUP signal is wrong.  We should attempt to deliver the
synchronous SIGSEGV signal we just forced.

We can make that happening in a fairly straight forward manner by
instead of just looking at the signal number we also look at the
si_code.  In particular for exceptions (aka synchronous signals) the
si_code is always greater than 0.

That still has the potential to pick up a number of asynchronous
signals as in a few cases the same si_codes that are used
for synchronous signals are also used for asynchronous signals,
and SI_KERNEL is also included in the list of possible si_codes.

Still the heuristic is much better and timer signals are definitely
excluded.  Which is enough to prevent all known ways for someone
sending a process signals fast enough to cause unexpected and
arguably incorrect behavior.

Cc: stable@vger.kernel.org
Fixes: a27341cd5fcb ("Prioritize synchronous signals over 'normal' signals")
Tested-by: Dmitry Vyukov 
Reported-by: Dmitry Vyukov 
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Greg Kroah-Hartman

signal: Always notice exiting tasks

2019-02-15T07:07:37+00:00

commit 35634ffa1751b6efd8cf75010b509dcb0263e29b upstream.

Recently syzkaller was able to create unkillablle processes by
creating a timer that is delivered as a thread local signal on SIGHUP,
and receiving SIGHUP SA_NODEFERER.  Ultimately causing a loop
failing to deliver SIGHUP but always trying.

Upon examination it turns out part of the problem is actually most of
the solution.  Since 2.5 signal delivery has found all fatal signals,
marked the signal group for death, and queued SIGKILL in every threads
thread queue relying on signal->group_exit_code to preserve the
information of which was the actual fatal signal.

The conversion of all fatal signals to SIGKILL results in the
synchronous signal heuristic in next_signal kicking in and preferring
SIGHUP to SIGKILL.  Which is especially problematic as all
fatal signals have already been transformed into SIGKILL.

Instead of dequeueing signals and depending upon SIGKILL to
be the first signal dequeued, first test if the signal group
has already been marked for death.  This guarantees that
nothing in the signal queue can prevent a process that needs
to exit from exiting.

Cc: stable@vger.kernel.org
Tested-by: Dmitry Vyukov 
Reported-by: Dmitry Vyukov 
Ref: ebf5ebe31d2c ("[PATCH] signal-fixes-2.5.59-A4")
History Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Greg Kroah-Hartman

signal: Introduce COMPAT_SIGMINSTKSZ for use in compat_sys_sigaltstack

2018-12-21T13:11:29+00:00

commit 22839869f21ab3850fbbac9b425ccc4c0023926f upstream.

The sigaltstack(2) system call fails with -ENOMEM if the new alternative
signal stack is found to be smaller than SIGMINSTKSZ. On architectures
such as arm64, where the native value for SIGMINSTKSZ is larger than
the compat value, this can result in an unexpected error being reported
to a compat task. See, for example:

  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=904385

This patch fixes the problem by extending do_sigaltstack to take the
minimum signal stack size as an additional parameter, allowing the
native and compat system call entry code to pass in their respective
values. COMPAT_SIGMINSTKSZ is just defined as SIGMINSTKSZ if it has not
been defined by the architecture.

Cc: Arnd Bergmann 
Cc: Dominik Brodowski 
Cc: "Eric W. Biederman" 
Cc: Andrew Morton 
Cc: Al Viro 
Cc: Oleg Nesterov 
Reported-by: Steve McIntyre 
Tested-by: Steve McIntyre <93sam@debian.org>
Signed-off-by: Will Deacon 
Signed-off-by: Catalin Marinas 
[signal: Fix up cherry-pick conflicts for 22839869f21a]
Signed-off-by: Steve McIntyre <93sam@debian.org>
Signed-off-by: Greg Kroah-Hartman

signal: Always deliver the kernel's SIGKILL and SIGSTOP to a pid namespace init

2018-11-13T19:16:53+00:00

[ Upstream commit 3597dfe01d12f570bc739da67f857fd222a3ea66 ]

Instead of playing whack-a-mole and changing SEND_SIG_PRIV to
SEND_SIG_FORCED throughout the kernel to ensure a pid namespace init
gets signals sent by the kernel, stop allowing a pid namespace init to
ignore SIGKILL or SIGSTOP sent by the kernel.  A pid namespace init is
only supposed to be able to ignore signals sent from itself and
children with SIG_DFL.

Fixes: 921cf9f63089 ("signals: protect cinit from unblocked SIG_DFL signals")
Reviewed-by: Thomas Gleixner 
Signed-off-by: "Eric W. Biederman" 
Signed-off-by: Sasha Levin 
Signed-off-by: Greg Kroah-Hartman