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 <[email protected]>
Reported-by: Oleg Nesterov <[email protected]>
Reported-by: Ivan Delalande <[email protected]>
Cc: [email protected]
Fixes: 35634ffa17 ("signal: Always notice exiting tasks")
Signed-off-by: "Eric W. Biederman" <[email protected]>
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: [email protected]
Fixes: a27341cd5f ("Prioritize synchronous signals over 'normal' signals")
Tested-by: Dmitry Vyukov <[email protected]>
Reported-by: Dmitry Vyukov <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
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: [email protected]
Tested-by: Dmitry Vyukov <[email protected]>
Reported-by: Dmitry Vyukov <[email protected]>
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" <[email protected]>
Since 2.5.34 the code has had the potential to not allocate siginfo
for SIGSTOP signals. Except for ptrace this is perfectly fine as only
ptrace can use PTRACE_PEEK_SIGINFO and see what the contents of
the delivered siginfo are.
Users of PTRACE_PEEK_SIGINFO that care about the contents siginfo
for SIGSTOP are rare, but they do exist. A seccomp self test
has cared and lldb cares.
Jack Andersen <[email protected]> writes:
> The patch titled
> `signal: Never allocate siginfo for SIGKILL or SIGSTOP`
> created a regression for users of PTRACE_GETSIGINFO needing to
> discern signals that were raised via the tgkill syscall.
>
> A notable user of this tgkill+ptrace combination is lldb while
> debugging a multithreaded program. Without the ability to detect a
> SIGSTOP originating from tgkill, lldb does not have a way to
> synchronize on a per-thread basis and falls back to SIGSTOP-ing the
> entire process.
Everyone affected by this please note. The kernel can still fail to
allocate a siginfo structure. The allocation is with GFP_KERNEL and
is best effort only. If memory is tight when the signal allocation
comes in this will fail to allocate a siginfo.
So I strongly recommend looking at more robust solutions for
synchronizing with a single thread such as PTRACE_INTERRUPT. Or if
that does not work persuading your friendly local kernel developer to
build the interface you need.
Reported-by: Tycho Andersen <[email protected]>
Reported-by: Kees Cook <[email protected]>
Reported-by: Jack Andersen <[email protected]>
Acked-by: Linus Torvalds <[email protected]>
Reviewed-by: Christian Brauner <[email protected]>
Cc: [email protected]
Fixes: f149b31557 ("signal: Never allocate siginfo for SIGKILL or SIGSTOP")
Fixes: 6dfc88977e42 ("[PATCH] shared thread signals")
History Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Signed-off-by: "Eric W. Biederman" <[email protected]>
The subvol_name is allocated in btrfs_parse_subvol_options and is
consumed and freed in mount_subvol. Add a free to the error paths that
don't call mount_subvol so that it is guaranteed that subvol_name is
freed when an error happens.
Fixes: 312c89fbca ("btrfs: cleanup btrfs_mount() using btrfs_mount_root()")
Cc: [email protected] # v4.19+
Reviewed-by: Nikolay Borisov <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Reviewed-by: David Sterba <[email protected]>
Signed-off-by: David Sterba <[email protected]>
Eric Biggers reported:
> The following commit, which went into v4.20, introduced undefined behavior when
> sys_rt_sigqueueinfo() is called with sig=0:
>
> commit 4ce5f9c9e7
> Author: Eric W. Biederman <[email protected]>
> Date: Tue Sep 25 12:59:31 2018 +0200
>
> signal: Use a smaller struct siginfo in the kernel
>
> In sig_specific_sicodes(), used from known_siginfo_layout(), the expression
> '1ULL << ((sig)-1)' is undefined as it evaluates to 1ULL << 4294967295.
>
> Reproducer:
>
> #include <signal.h>
> #include <sys/syscall.h>
> #include <unistd.h>
>
> int main(void)
> {
> siginfo_t si = { .si_code = 1 };
> syscall(__NR_rt_sigqueueinfo, 0, 0, &si);
> }
>
> UBSAN report for v5.0-rc1:
>
> UBSAN: Undefined behaviour in kernel/signal.c:2946:7
> shift exponent 4294967295 is too large for 64-bit type 'long unsigned int'
> CPU: 2 PID: 346 Comm: syz_signal Not tainted 5.0.0-rc1 #25
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
> Call Trace:
> __dump_stack lib/dump_stack.c:77 [inline]
> dump_stack+0x70/0xa5 lib/dump_stack.c:113
> ubsan_epilogue+0xd/0x40 lib/ubsan.c:159
> __ubsan_handle_shift_out_of_bounds+0x12c/0x170 lib/ubsan.c:425
> known_siginfo_layout+0xae/0xe0 kernel/signal.c:2946
> post_copy_siginfo_from_user kernel/signal.c:3009 [inline]
> __copy_siginfo_from_user+0x35/0x60 kernel/signal.c:3035
> __do_sys_rt_sigqueueinfo kernel/signal.c:3553 [inline]
> __se_sys_rt_sigqueueinfo kernel/signal.c:3549 [inline]
> __x64_sys_rt_sigqueueinfo+0x31/0x70 kernel/signal.c:3549
> do_syscall_64+0x4c/0x1b0 arch/x86/entry/common.c:290
> entry_SYSCALL_64_after_hwframe+0x49/0xbe
> RIP: 0033:0x433639
> Code: c4 18 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 0f 83 7b 27 00 00 c3 66 2e 0f 1f 84 00 00 00 00
> RSP: 002b:00007fffcb289fc8 EFLAGS: 00000246 ORIG_RAX: 0000000000000081
> RAX: ffffffffffffffda RBX: 00000000004002e0 RCX: 0000000000433639
> RDX: 00007fffcb289fd0 RSI: 0000000000000000 RDI: 0000000000000000
> RBP: 00000000006b2018 R08: 000000000000004d R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000401560
> R13: 00000000004015f0 R14: 0000000000000000 R15: 0000000000000000
I have looked at the other callers of siginmask and they all appear to
in locations where sig can not be zero.
I have looked at the code generation of adding an extra test against
zero and gcc was able with a simple decrement instruction to combine
the two tests together. So the at most adding this test cost a single
cpu cycle. In practice that decrement instruction was already present
as part of the mask comparison, so the only change was when the
instruction was executed.
So given that it is cheap, and obviously correct to update siginmask
to verify the signal is not zero. Fix this issue there to avoid any
future problems.
Reported-by: Eric Biggers <[email protected]>
Fixes: 4ce5f9c9e7 ("signal: Use a smaller struct siginfo in the kernel")
Signed-off-by: "Eric W. Biederman" <[email protected]>
Timothy Baldwin <[email protected]> wrote:
> As per mount_namespaces(7) unprivileged users should not be able to look under mount points:
>
> Mounts that come as a single unit from more privileged mount are locked
> together and may not be separated in a less privileged mount namespace.
>
> However they can:
>
> 1. Create a mount namespace.
> 2. In the mount namespace open a file descriptor to the parent of a mount point.
> 3. Destroy the mount namespace.
> 4. Use the file descriptor to look under the mount point.
>
> I have reproduced this with Linux 4.16.18 and Linux 4.18-rc8.
>
> The setup:
>
> $ sudo sysctl kernel.unprivileged_userns_clone=1
> kernel.unprivileged_userns_clone = 1
> $ mkdir -p A/B/Secret
> $ sudo mount -t tmpfs hide A/B
>
>
> "Secret" is indeed hidden as expected:
>
> $ ls -lR A
> A:
> total 0
> drwxrwxrwt 2 root root 40 Feb 12 21:08 B
>
> A/B:
> total 0
>
>
> The attack revealing "Secret":
>
> $ unshare -Umr sh -c "exec unshare -m ls -lR /proc/self/fd/4/ 4<A"
> /proc/self/fd/4/:
> total 0
> drwxr-xr-x 3 root root 60 Feb 12 21:08 B
>
> /proc/self/fd/4/B:
> total 0
> drwxr-xr-x 2 root root 40 Feb 12 21:08 Secret
>
> /proc/self/fd/4/B/Secret:
> total 0
I tracked this down to put_mnt_ns running passing UMOUNT_SYNC and
disconnecting all of the mounts in a mount namespace. Fix this by
factoring drop_mounts out of drop_collected_mounts and passing
0 instead of UMOUNT_SYNC.
There are two possible behavior differences that result from this.
- No longer setting UMOUNT_SYNC will no longer set MNT_SYNC_UMOUNT on
the vfsmounts being unmounted. This effects the lazy rcu walk by
kicking the walk out of rcu mode and forcing it to be a non-lazy
walk.
- No longer disconnecting locked mounts will keep some mounts around
longer as they stay because the are locked to other mounts.
There are only two users of drop_collected mounts: audit_tree.c and
put_mnt_ns.
In audit_tree.c the mounts are private and there are no rcu lazy walks
only calls to iterate_mounts. So the changes should have no effect
except for a small timing effect as the connected mounts are disconnected.
In put_mnt_ns there may be references from process outside the mount
namespace to the mounts. So the mounts remaining connected will
be the bug fix that is needed. That rcu walks are allowed to continue
appears not to be a problem especially as the rcu walk change was about
an implementation detail not about semantics.
Cc: [email protected]
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Reported-by: Timothy Baldwin <[email protected]>
Tested-by: Timothy Baldwin <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Jonathan Calmels from NVIDIA reported that he's able to bypass the
mount visibility security check in place in the Linux kernel by using
a combination of the unbindable property along with the private mount
propagation option to allow a unprivileged user to see a path which
was purposefully hidden by the root user.
Reproducer:
# Hide a path to all users using a tmpfs
root@castiana:~# mount -t tmpfs tmpfs /sys/devices/
root@castiana:~#
# As an unprivileged user, unshare user namespace and mount namespace
stgraber@castiana:~$ unshare -U -m -r
# Confirm the path is still not accessible
root@castiana:~# ls /sys/devices/
# Make /sys recursively unbindable and private
root@castiana:~# mount --make-runbindable /sys
root@castiana:~# mount --make-private /sys
# Recursively bind-mount the rest of /sys over to /mnnt
root@castiana:~# mount --rbind /sys/ /mnt
# Access our hidden /sys/device as an unprivileged user
root@castiana:~# ls /mnt/devices/
breakpoint cpu cstate_core cstate_pkg i915 intel_pt isa kprobe
LNXSYSTM:00 msr pci0000:00 platform pnp0 power software system
tracepoint uncore_arb uncore_cbox_0 uncore_cbox_1 uprobe virtual
Solve this by teaching copy_tree to fail if a mount turns out to be
both unbindable and locked.
Cc: [email protected]
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Reported-by: Jonathan Calmels <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
It was recently pointed out that the one instance of testing MNT_LOCKED
outside of the namespace_sem is in ksys_umount.
Fix that by adding a test inside of do_umount with namespace_sem and
the mount_lock held. As it helps to fail fails the existing test is
maintained with an additional comment pointing out that it may be racy
because the locks are not held.
Cc: [email protected]
Reported-by: Al Viro <[email protected]>
Fixes: 5ff9d8a65c ("vfs: Lock in place mounts from more privileged users")
Signed-off-by: "Eric W. Biederman" <[email protected]>
While fixing an out of bounds array access in known_siginfo_layout
reported by the kernel test robot it became apparent that the same bug
exists in siginfo_layout and affects copy_siginfo_from_user32.
The straight forward fix that makes guards against making this mistake
in the future and should keep the code size small is to just take an
unsigned signal number instead of a signed signal number, as I did to
fix known_siginfo_layout.
Cc: [email protected]
Fixes: cc731525f2 ("signal: Remove kernel interal si_code magic")
Signed-off-by: "Eric W. Biederman" <[email protected]>
The bounds checks in known_siginfo_layout only guards against positive
numbers that are too large, large negative can slip through and
can cause out of bounds accesses.
Ordinarily this is not a concern because early in signal processing
the signal number is filtered with valid_signal which ensures it
is a small positive signal number, but copy_siginfo_from_user
is called before this check is performed.
[ 73.031126] BUG: unable to handle kernel paging request at ffffffff6281bcb6
[ 73.032038] PGD 3014067 P4D 3014067 PUD 0
[ 73.032565] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC PTI
[ 73.033287] CPU: 0 PID: 732 Comm: trinity-c3 Tainted: G W T 4.19.0-rc1-00077-g4ce5f9c #1
[ 73.034423] RIP: 0010:copy_siginfo_from_user+0x4d/0xd0
[ 73.034908] Code: 00 8b 53 08 81 fa 80 00 00 00 0f 84 90 00 00 00 85 d2 7e 2d 48 63 0b 83 f9 1f 7f 1c 8d 71 ff bf d8 04 01 50 48 0f a3 f7 73 0e <0f> b6 8c 09 20 bb 81 82 39 ca 7f 15 eb 68 31 c0 83 fa 06 7f 0c eb
[ 73.036665] RSP: 0018:ffff88001b8f7e20 EFLAGS: 00010297
[ 73.037160] RAX: 0000000000000000 RBX: ffff88001b8f7e90 RCX: fffffffff00000cb
[ 73.037865] RDX: 0000000000000001 RSI: 00000000f00000ca RDI: 00000000500104d8
[ 73.038546] RBP: ffff88001b8f7e80 R08: 0000000000000000 R09: 0000000000000000
[ 73.039201] R10: 0000000000000001 R11: 0000000000000000 R12: 0000000000000008
[ 73.039874] R13: 00000000000002dc R14: 0000000000000000 R15: 0000000000000000
[ 73.040613] FS: 000000000104a880(0000) GS:ffff88001f000000(0000) knlGS:0000000000000000
[ 73.041649] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 73.042405] CR2: ffffffff6281bcb6 CR3: 000000001cb52003 CR4: 00000000001606b0
[ 73.043351] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 73.044286] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600
[ 73.045221] Call Trace:
[ 73.045556] __x64_sys_rt_tgsigqueueinfo+0x34/0xa0
[ 73.046199] do_syscall_64+0x1a4/0x390
[ 73.046708] ? vtime_user_enter+0x61/0x80
[ 73.047242] ? __context_tracking_enter+0x4e/0x60
[ 73.047714] ? __context_tracking_enter+0x4e/0x60
[ 73.048278] entry_SYSCALL_64_after_hwframe+0x44/0xa9
Therefore fix known_siginfo_layout to take an unsigned signal number
instead of a signed signal number. All valid signal numbers are small
positive numbers so they will not be affected, but invalid negative
signal numbers will now become large positive signal numbers and will
not be used as indices into the sig_sicodes array.
Making the signal number unsigned makes it difficult for similar mistakes to
happen in the future.
Fixes: 4ce5f9c9e7 ("signal: Use a smaller struct siginfo in the kernel")
Inspired-by: Sean Christopherson <[email protected]>
Reported-by: kernel test robot <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Andrei Vagin <[email protected]> reported:
> Accoding to the man page, the user should not set si_signo, it has to be set
> by kernel.
>
> $ man 2 rt_sigqueueinfo
>
> The uinfo argument specifies the data to accompany the signal. This
> argument is a pointer to a structure of type siginfo_t, described in
> sigaction(2) (and defined by including <sigaction.h>). The caller
> should set the following fields in this structure:
>
> si_code
> This must be one of the SI_* codes in the Linux kernel source
> file include/asm-generic/siginfo.h, with the restriction that
> the code must be negative (i.e., cannot be SI_USER, which is
> used by the kernel to indicate a signal sent by kill(2)) and
> cannot (since Linux 2.6.39) be SI_TKILL (which is used by the
> kernel to indicate a signal sent using tgkill(2)).
>
> si_pid This should be set to a process ID, typically the process ID of
> the sender.
>
> si_uid This should be set to a user ID, typically the real user ID of
> the sender.
>
> si_value
> This field contains the user data to accompany the signal. For
> more information, see the description of the last (union sigval)
> argument of sigqueue(3).
>
> Internally, the kernel sets the si_signo field to the value specified
> in sig, so that the receiver of the signal can also obtain the signal
> number via that field.
>
> On Tue, Sep 25, 2018 at 07:19:02PM +0200, Eric W. Biederman wrote:
>>
>> If there is some application that calls sigqueueinfo directly that has
>> a problem with this added sanity check we can revisit this when we see
>> what kind of crazy that application is doing.
>
>
> I already know two "applications" ;)
>
> https://github.com/torvalds/linux/blob/master/tools/testing/selftests/ptrace/peeksiginfo.c
> https://github.com/checkpoint-restore/criu/blob/master/test/zdtm/static/sigpending.c
>
> Disclaimer: I'm the author of both of them.
Looking at the kernel code the historical behavior has alwasy been to prefer
the signal number passed in by the kernel.
So sigh. Implmenet __copy_siginfo_from_user and __copy_siginfo_from_user32 to
take that signal number and prefer it. The user of ptrace will still
use copy_siginfo_from_user and copy_siginfo_from_user32 as they do not and
never have had a signal number there.
Luckily this change has never made it farther than linux-next.
Fixes: e75dc036c4 ("signal: Fail sigqueueinfo if si_signo != sig")
Reported-by: Andrei Vagin <[email protected]>
Tested-by: Andrei Vagin <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
We reserve 128 bytes for struct siginfo but only use about 48 bytes on
64bit and 32 bytes on 32bit. Someday we might use more but it is unlikely
to be anytime soon.
Userspace seems content with just enough bytes of siginfo to implement
sigqueue. Or in the case of checkpoint/restart reinjecting signals
the kernel has sent.
Reducing the stack footprint and the work to copy siginfo around from
2 cachelines to 1 cachelines seems worth doing even if I don't have
benchmarks to show a performance difference.
Suggested-by: Linus Torvalds <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Linus recently observed that if we did not worry about the padding
member in struct siginfo it is only about 48 bytes, and 48 bytes is
much nicer than 128 bytes for allocating on the stack and copying
around in the kernel.
The obvious thing of only adding the padding when userspace is
including siginfo.h won't work as there are sigframe definitions in
the kernel that embed struct siginfo.
So split siginfo in two; kernel_siginfo and siginfo. Keeping the
traditional name for the userspace definition. While the version that
is used internally to the kernel and ultimately will not be padded to
128 bytes is called kernel_siginfo.
The definition of struct kernel_siginfo I have put in include/signal_types.h
A set of buildtime checks has been added to verify the two structures have
the same field offsets.
To make it easy to verify the change kernel_siginfo retains the same
size as siginfo. The reduction in size comes in a following change.
Signed-off-by: "Eric W. Biederman" <[email protected]>
In preparation for using a smaller version of siginfo in the kernel
introduce copy_siginfo_from_user and use it when siginfo is copied from
userspace.
Make the pattern for using copy_siginfo_from_user and
copy_siginfo_from_user32 to capture the return value and return that
value on error.
This is a necessary prerequisite for using a smaller siginfo
in the kernel than the kernel exports to userspace.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Rework the defintion of struct siginfo so that the array padding
struct siginfo to SI_MAX_SIZE can be placed in a union along side of
the rest of the struct siginfo members. The result is that we no
longer need the __ARCH_SI_PREAMBLE_SIZE or SI_PAD_SIZE definitions.
Signed-off-by: "Eric W. Biederman" <[email protected]>
The kernel needs to validate that the contents of struct siginfo make
sense as siginfo is copied into the kernel, so that the proper union
members can be put in the appropriate locations. The field si_signo
is a fundamental part of that validation. As such changing the
contents of si_signo after the validation make no sense and can result
in nonsense values in the kernel.
As such simply fail if someone is silly enough to set si_signo out of
sync with the signal number passed to sigqueueinfo.
I don't expect a problem as glibc's sigqueue implementation sets
"si_signo = sig" and CRIU just returns to the kernel what the kernel
gave to it.
If there is some application that calls sigqueueinfo directly that has
a problem with this added sanity check we can revisit this when we see
what kind of crazy that application is doing.
Signed-off-by: "Eric W. Biederman" <[email protected]>
When moving all of the architectures specific si_codes into
siginfo.h, I apparently overlooked EMT_TAGOVF. Move it now.
Remove the now redundant test in siginfo_layout for SIGEMT
as now NSIGEMT is always defined.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Pass the signal number, and the signal code, and the faulting
address into uc32_notify_die so the callers do not need
to generate a struct siginfo.
In ucs32_ntoify_die use the newly passed in information to
call force_sig_fault to generate the siginfo and send the error.
This simplifies the code making the chances of bugs much less likely.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Pass signr, sicode, and address into unhandled_exception as explicit
parameters instead of members of struct siginfo. Then in unhandled
exception generate and send the siginfo using force_sig_fault.
This keeps the code simpler and less error prone.
Acked-by: Vineet Gupta <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The ia64 handling of failure to return from a signal frame has been trying
to set overlapping fields in struct siginfo since 2.3.43. The si_code
corresponds to the fields that were stomped (not the field that is
actually written), so I can not imagine a piece of userspace code
making sense of the signal frame if it looks closely.
In practice failure to return from a signal frame is a rare event that
almost never happens. Someone using an alternate signal stack to
recover and looking in detail is even more rare. So I presume no one
has ever noticed and reported this ia64 nonsense.
Sort this out by causing ia64 to use force_sig(SIGSEGV) like other architectures.
Fixes: 2.3.43
Cc: Tony Luck <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: [email protected]
Acked-by: Tony Luck <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The ia64 handling of failure to setup a signal frame has been trying
to set overlapping fields in struct siginfo since 2.3.43. The si_pid
and si_uid fields are stomped when the si_addr field is set. The
si_code of SI_KERNEL indicates that si_pid and si_uid should be valid,
and that si_addr does not exist.
Being at odds with the definition of SI_KERNEL and with nothing to
indicate that this was a signal frame setup failure there is no way
for userspace to know that si_addr was filled out instead.
In practice failure to setup a signal frame is rare, and si_pid and
si_uid are always set to 0 when si_code is SI_KERNEL so I expect no
one has looked closely enough before to see this weirdness. Further
the only difference between force_sigsegv_info and the generic
force_sigsegv other than the return code is that force_sigsegv_info
stomps the si_uid and si_pid fields.
Remove the bug and simplify the code by using force_sigsegv in this
case just like other architectures.
Fixes: 2.3.43
Cc: Tony Luck <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: [email protected]
Acked-by: Tony Luck <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
This simplifies the code making it clearer what is going on, and
making the siginfo generation easier to maintain.
Signed-off-by: "Eric W. Biederman" <[email protected]>
In arm_notify_die call force_sig_fault to let the generic
code handle siginfo generation.
This removes some boiler plate making the code easier to
maintain in the long run.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Add arm64_force_sig_ptrace_errno_trap for consistency with
arm64_force_sig_fault and use it where appropriate.
This adds the show_signal logic to the force_sig_errno_trap case,
where it was apparently overlooked earlier.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Add arm64_force_sig_mceerr for consistency with arm64_force_sig_fault,
and use it in the one location that can take advantage of it.
This removes the fiddly filling out of siginfo before sending a signal
reporting an memory error to userspace.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Wrap force_sig_fault with a helper that calls arm64_show_signal
and call arm64_force_sig_fault where appropraite.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: Eric W. Biederman <[email protected]>
This code is truly common between the signal sending cases so share it.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: Eric W. Biederman <[email protected]>
As this work is truly common between all of the signal sending cases
there is no need to repeat it between the different cases.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: Eric W. Biederman <[email protected]>
Not all of the signals passed to __do_user_fault can be handled
the same way so expand the now tiny __do_user_fault in it's callers
and remove it.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
It gets easy to confuse what is going on when some code is shared and some not
so stop sharing the trivial bits of signal generation to make future updates
easier to understand.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
These two cases are practically the same and use siginfo differently
from the other signals sent from do_page_fault. So consolidate them
to make future changes easier.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
This pepares for sending signals with something other than
arm64_force_sig_info.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Filling in siginfo is error prone and so it is wise to use more
specialized helpers to do that work. Factor out the arm specific
unhandled signal reporting from the work of delivering a signal so
the code can be modified to use functions that take the information
to fill out siginfo as parameters.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Every caller passes in current for tsk so there is no need to pass
tsk. Instead make tsk a local variable initialized to current.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Instead of generating a struct siginfo before calling arm64_notify_die
pass the signal number, tne sicode and the fault address into
arm64_notify_die and have it call force_sig_fault instead of
force_sig_info to let the generic code generate the struct siginfo.
This keeps code passing just the needed information into
siginfo generating code, making it easier to see what
is happening and harder to get wrong. Further by letting
the generic code handle the generation of struct siginfo
it reduces the number of sites generating struct siginfo
making it possible to review them and verify that all
of the fiddly details for a structure passed to userspace
are handled properly.
Reviewed-by: Catalin Marinas <[email protected]>
Tested-by: Catalin Marinas <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Call force_sig_pkuerr directly instead of rolling it by hand
in _exception_pkey.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that _exception no longer calls _exception_pkey it is no longer
necessary to handle any signal with any si_code. All pkey exceptions
are SIGSEGV with paired with SEGV_PKUERR. So just handle
that case and remove the now unnecessary parameters from _exception_pkey.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The callers of _exception don't need the pkey exception logic because
they are not processing a pkey exception. So just call exception_common
directly and then call force_sig_fault to generate the appropriate siginfo
and deliver the appropriate signal.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
It is brittle and wrong to populate si_pkey when there was not a pkey
exception. The field does not exist for all si_codes and in some
cases another field exists in the same memory location.
So factor out the code that all exceptions handlers must run
into exception_common, leaving the individual exception handlers
to generate the signals themselves.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that bad_key_fault_exception no longer calls __bad_area_nosemaphore
there is no reason for __bad_area_nosemaphore to handle pkeys.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
This removes the need for other code paths to deal with pkey exceptions.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
There are no callers of __bad_area that pass in a pkey parameter so it makes
no sense to take one.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
In do_sigbus isolate the mceerr signaling code and call
force_sig_mceerr instead of falling through to the force_sig_info that
works for all of the other signals.
Reviewed-by: Stephen Rothwell <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that si_code == SEGV_PKUERR is the flag indicating that a pkey
is present there is no longer a need to pass a pointer to a local
pkey value, instead pkey can be passed more efficiently by value.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that the pkey handling has been removed force_sig_info_fault and
force_sig_fault perform identical work. Just the type of the address
paramter is different. So replace calls to force_sig_info_fault with
calls to force_sig_fault, and remove force_sig_info_fault.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
There is only one code path that can generate a pkuerr signal. That
code path calls __bad_area_nosemaphore and can be dectected by testing
if si_code == SEGV_PKUERR. It can be seen from inspection that all of
the other tests in fill_sig_info_pkey are unnecessary.
Therefore call force_sig_pkuerr directly from __bad_area_semaphore and
remove fill_sig_info_pkey.
At the same time move the comment above force_sig_info_pkey into
bad_area_access_error, so that the documentation about pkey generation
races is not lost.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
There is only one caller of __bad_area that passes in PKUERR and thus
will generate a siginfo with si_pkey set. Therefore simplify the
logic and hoist reading of vma_pkey up into that caller, and just
pass *pkey into __bad_area.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
There are no more users of the computed pkey value in __do_page_fault
so stop computing the value.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
After the previous cleanups to do_sigbus and and bad_area_nosemaphore
mm_fault_error no now longer uses it's pkey parameter. Therefore
remove the unused parameter.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The function do_sigbus never sets si_code to PKUERR so it can never
return a pkey to userspace. Therefore remove the unusable pkey
parameter from do_sigbus.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The function bad_area_nosemaphore always sets si_code to SEGV_MAPERR
and as such can never return a pkey parameter. Therefore remove the
unusable pkey parameter from bad_area_nosemaphore.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Update the DO_ERROR macro to take si_code and si_addr values for a siginfo,
removing the need for the fill_trap_info function.
Update do_trap to also take the sicode and si_addr values for a sigininfo
and modify the code to call force_sig when a sicode is not passed in
and to call force_sig_fault when all of the information is present.
Making this a more obvious, simpler and less error prone construction.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The function "force_sig(sig, tsk)" is equivalent to "
force_sig_info(sig, SEND_SIG_PRIV, tsk)". Using the siginfo variants can
be error prone so use the simpler old fashioned force_sig variant,
and with luck the force_sig_info variant can go away.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Instead of generating the siginfo in x86 specific code use the new
helper function force_sig_bnderr to separate the concerns of
collecting the information and generating a proper siginfo.
Making the code easier to understand and maintain.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The function do_trap_no_signal embodies almost all of the work of the
function do_trap. The exceptions are setting of thread.error_code and
thread.trap_nr in the case when the signal will be sent, and reporting
which signal will be sent with show_signal.
Filling in struct siginfo and then calling do_trap is problematic as
filling in struct siginfo is an fiddly process that can through
inattention has resulted in fields not initialized and the wrong
fields being filled in.
To avoid this error prone situation I am replacing force_sig_info with
a set of functions that take as arguments the information needed to
send a specific kind of signal.
The function do_trap is called in the context of several different
kinds of signals today. Having a solid do_trap_no_signal that
can be reused allows call sites that send different kinds of
signals to reuse all of the code in do_trap_no_signal.
Modify do_trap_no_signal to have a single exit there signals
where be sent (aka returning -1) to allow more of the signal
sending path to be moved to from do_trap to do_trap_no_signal.
Move setting thread.trap_nr and thread.error_code into do_trap_no_signal
so the code does not need to be duplicated.
Make the type of the string that is passed into do_trap_no_signal to
const. The only user of that str is die and it already takes a const
string, so this just makes it explicit that the string won't change.
All of this prepares the way for using do_trap_no_signal outside
of do_trap.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The code for conditionally printing unhanded signals is duplicated twice
in arch/x86/kernel/traps.c. Factor it out into it's own subroutine
called show_signal to make the code clearer and easier to maintain.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
This separates the logic of generating the signal from the logic of
gathering the information about the bounds violation.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The value passed in to addr_referenced is of type void __user *, so update
the addr_referenced parameter in trace_mpx_bounds_register_exception to match.
Also update the addr_referenced paramater in TP_STRUCT__entry as it again
holdes the same value.
I don't know why this was missed earlier but sparse was complaining when
testing test branch so fix this now.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
This simplifies the code making it clearer what is going on.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Only the call from do_sigbus will send SIGBUS due to a memory machine
check error. Consolidate all of the machine check signal generation
code in do_sigbus and remove the now unnecessary fault parameter from
force_sig_info_fault.
Explicitly use the now constant si_code BUS_ADRERR in the call
to force_sig_info_fault from do_sigbus.
This makes the code in arch/x86/mm/fault.c easier to follower and
simpler to maintain.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The function fill_sigtrap_info now only has one caller so remove
it and put it's contents in it's caller.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Replace user_single_step_siginfo with user_single_step_report
that allocates siginfo structure on the stack and sends it.
This allows tracehook_report_syscall_exit to become a simple
if statement that calls user_single_step_report or ptrace_report_syscall
depending on the value of step.
Update the default helper function now called user_single_step_report
to explicitly set si_code to SI_USER and to set si_uid and si_pid to 0.
The default helper has always been doing this (using memset) but it
was far from obvious.
The powerpc helper can now just call force_sig_fault.
The x86 helper can now just call send_sigtrap.
Unfortunately the default implementation of user_single_step_report
can not use force_sig_fault as it does not use a SIGTRAP si_code.
So it has to carefully setup the siginfo and use use force_sig_info.
The net result is code that is easier to understand and simpler
to maintain.
Ref: 85ec7fd9f8 ("ptrace: introduce user_single_step_siginfo() helper")
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
For readability and consistency with the other exports in
kernel/signal.c pair the exports of signal sending functions with
their functions, instead of having the exports in one big clump.
Signed-off-by: "Eric W. Biederman" <[email protected]>
This function is static and it only has two callers. As
specific_send_sig_info is only called twice remembering what
specific_send_sig_info does when reading the code is difficutl and it
makes it hard to see which sending sending functions are equivalent to
which others.
So remove specific_send_sig_info to make the code easier to read.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Replace send_sig_info in zap_pid_ns_processes with
group_send_sig_info. This makes more sense as the entire process
group is being killed. More importantly this allows the kill of those
processes with PIDTYPE_MAX to indicate all of the process in the pid
namespace are being signaled. This is needed for fork to detect when
signals are sent to a group of processes.
Admittedly fork has another case to catch SIGKILL but the principle remains
that it is desirable to know when a group of processes is being signaled.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Replace send_sig and force_sig in __do_SAK with group_send_sig_info
the general helper for sending a signal to a process group. This is
wordier but it allows specifying PIDTYPE_SID so that the signal code
knows the signal went to a session.
Both force_sig() and send_sig(..., 1) specify SEND_SIG_PRIV and the
new call of group_send_sig_info does that explicitly. This is enough
to ensure even a pid namespace init is killed.
The global init remains unkillable. The guarantee that __do_SAK tries
to provide is a clean path to login to a machine. As the global init is
unkillable, if it chooses to hold open a tty it can violate this
guarantee. A technique other than killing processes would be needed
to provide this guarantee to userspace.
The only difference between force_sig and send_sig when sending
SIGKILL is that SIGNAL_UNKILLABLE is cleared. This has no affect on
the processing of a signal sent with SEND_SIG_PRIV by any process, making
it unnecessary, and not behavior that needs to be preserved.
force_sig was used originally because it did not take as many locks as
send_sig. Today send_sig, force_sig and group_send_sig_info take the
same locks when delivering a signal.
group_send_sig_info also contains a permission check that force_sig
and send_sig do not. However the presence of SEND_SIG_PRIV makes the
permission check a noop. So the permission check does not result
in any behavioral differences.
Signed-off-by: "Eric W. Biederman" <[email protected]>
The genweq_add_file and genwqe_del_file by caching current without
using reference counting embed the assumption that a file descriptor
will never be passed from one process to another. It even embeds the
assumption that the the thread that opened the file will be in
existence when the process terminates. Neither of which are
guaranteed to be true.
Therefore replace caching the task_struct of the opener with
pid of the openers thread group id. All the knowledge of the
opener is used for is as the target of SIGKILL and a SIGKILL
will kill the entire process group.
Rename genwqe_force_sig to genwqe_terminate, remove it's unncessary
signal argument, update it's ownly caller, and use kill_pid
instead of force_sig.
The work force_sig does in changing signal handling state is not
relevant to SIGKILL sent as SEND_SIG_PRIV. The exact same processess
will be killed just with less work, and less confusion. The work done
by force_sig is really only needed for handling syncrhonous
exceptions.
It will still be possible to cause genwqe_device_remove to wait
8 seconds by passing a file descriptor to another process but
the possible user after free is fixed.
Fixes: eaf4722d46 ("GenWQE Character device and DDCB queue")
Cc: [email protected]
Cc: Greg Kroah-Hartman <[email protected]>
Cc: Frank Haverkamp <[email protected]>
Cc: Joerg-Stephan Vogt <[email protected]>
Cc: Michael Jung <[email protected]>
Cc: Michael Ruettger <[email protected]>
Cc: Kleber Sacilotto de Souza <[email protected]>
Cc: Sebastian Ott <[email protected]>
Cc: Eberhard S. Amann <[email protected]>
Cc: Gabriel Krisman Bertazi <[email protected]>
Cc: Guilherme G. Piccoli <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
There are no more users of SEND_SIG_FORCED so it may be safely removed.
Remove the definition of SEND_SIG_FORCED, it's use in is_si_special,
it's use in TP_STORE_SIGINFO, and it's use in __send_signal as without
any users the uses of SEND_SIG_FORCED are now unncessary.
This makes the code simpler, easier to understand and use. Users of
signal sending functions now no longer need to ask themselves do I
need to use SEND_SIG_FORCED.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that siginfo is never allocated for SIGKILL and SIGSTOP there is
no difference between SEND_SIG_PRIV and SEND_SIG_FORCED for SIGKILL
and SIGSTOP. This makes SEND_SIG_FORCED unnecessary and redundant in
the presence of SIGKILL and SIGSTOP. Therefore change users of
SEND_SIG_FORCED that are sending SIGKILL or SIGSTOP to use
SEND_SIG_PRIV instead.
This removes the last users of SEND_SIG_FORCED.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
The SIGKILL and SIGSTOP signals are never delivered to userspace so
queued siginfo for these signals can never be observed. Therefore
remove the chance of failure by never even attempting to allocate
siginfo in those cases.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Today kernel threads never dequeue siginfo so it is pointless to
enqueue siginfo for them. The usb gadget mass storage driver goes
one farther and uses SEND_SIG_FORCED to guarantee that no siginfo is
even enqueued.
Generalize the optimization of the usb mass storage driver and never
perform an unnecessary allocation when delivering signals to kthreads.
Switch the mass storage driver from sending signals with
SEND_SIG_FORCED to SEND_SIG_PRIV. As using SEND_SIG_FORCED is now
unnecessary.
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Now that send_signal always delivers SEND_SIG_PRIV signals to a pid
namespace init it is no longer necessary to use SEND_SIG_FORCED when
calling do_send_sig_info to ensure that pid namespace inits are
signaled and possibly killed. Using SEND_SIG_PRIV is sufficient.
So use SEND_SIG_PRIV so that userspace when it receives a SIGTERM can
tell that the kernel sent the signal and not some random userspace
application.
Fixes: b82c32872d ("sysrq: use SEND_SIG_FORCED instead of force_sig()")
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
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: 921cf9f630 ("signals: protect cinit from unblocked SIG_DFL signals")
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
For userspace to tell the difference between an random signal
and an exception, the exception must include siginfo information.
Using SEND_SIG_FORCED for SIGSEGV is thus wrong, and it will result in
userspace seeing si_code == SI_USER (like a random signal) instead of
si_code == SI_KERNEL or a more specific si_code as all exceptions
deliver.
Therefore replace force_sig_info(SIGSEGV, SEND_SIG_FORCE, current)
with force_sig(SIG_SEGV, current) which gets this right and is shorter
and easier to type.
Fixes: 791eca1010 ("uretprobes/x86: Hijack return address")
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
For userspace to tell the difference between a random signal and an
exception, the exception must include siginfo information.
Using SEND_SIG_FORCED for SIGILL is thus wrong, and it will result
in userspace seeing si_code == SI_USER (like a random signal) instead
of si_code == SI_KERNEL or a more specific si_code as all exceptions
deliver.
Therefore replace force_sig_info(SIGILL, SEND_SIG_FORCE, current)
with force_sig(SIG_ILL, current) which gets this right and is
shorter and easier to type.
Fixes: 014940bad8 ("uprobes/x86: Send SIGILL if arch_uprobe_post_xol() fails")
Fixes: 0b5256c7f1 ("uprobes: Send SIGILL if handle_trampoline() fails")
Reviewed-by: Thomas Gleixner <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
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 <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Recently syzbot reported crashes in send_sigio_to_task and
send_sigurg_to_task in linux-next. Despite finding a reproducer
syzbot apparently did not bisected this or otherwise track down the
offending commit in linux-next.
I happened to see this report and examined the code because I had
recently changed these functions as part of making PIDTYPE_TGID a real
pid type so that fork would does not need to restart when receiving a
signal. By examination I see that I spotted a bug in the code
that could explain the reported crashes.
When I took Oleg's suggestion and optimized send_sigurg and send_sigio
to only send to a single task when type is PIDTYPE_PID or PIDTYPE_TGID
I failed to handle pids that no longer point to tasks. The macro
do_each_pid_task simply iterates for zero iterations. With pid_task
an explicit NULL test is needed.
Update the code to include the missing NULL test.
Fixes: 019191342f ("signal: Use PIDTYPE_TGID to clearly store where file signals will be sent")
Reported-by: [email protected]
Signed-off-by: "Eric W. Biederman" <[email protected]>
Wen Yang <[email protected]> and majiang <[email protected]>
report that a periodic signal received during fork can cause fork to
continually restart preventing an application from making progress.
The code was being overly pessimistic. Fork needs to guarantee that a
signal sent to multiple processes is logically delivered before the
fork and just to the forking process or logically delivered after the
fork to both the forking process and it's newly spawned child. For
signals like periodic timers that are always delivered to a single
process fork can safely complete and let them appear to logically
delivered after the fork().
While examining this issue I also discovered that fork today will miss
signals delivered to multiple processes during the fork and handled by
another thread. Similarly the current code will also miss blocked
signals that are delivered to multiple process, as those signals will
not appear pending during fork.
Add a list of each thread that is currently forking, and keep on that
list a signal set that records all of the signals sent to multiple
processes. When fork completes initialize the new processes
shared_pending signal set with it. The calculate_sigpending function
will see those signals and set TIF_SIGPENDING causing the new task to
take the slow path to userspace to handle those signals. Making it
appear as if those signals were received immediately after the fork.
It is not possible to send real time signals to multiple processes and
exceptions don't go to multiple processes, which means that that are
no signals sent to multiple processes that require siginfo. This
means it is safe to not bother collecting siginfo on signals sent
during fork.
The sigaction of a child of fork is initially the same as the
sigaction of the parent process. So a signal the parent ignores the
child will also initially ignore. Therefore it is safe to ignore
signals sent to multiple processes and ignored by the forking process.
Signals sent to only a single process or only a single thread and delivered
during fork are treated as if they are received after the fork, and generally
not dealt with. They won't cause any problems.
V2: Added removal from the multiprocess list on failure.
V3: Use -ERESTARTNOINTR directly
V4: - Don't queue both SIGCONT and SIGSTOP
- Initialize signal_struct.multiprocess in init_task
- Move setting of shared_pending to before the new task
is visible to signals. This prevents signals from comming
in before shared_pending.signal is set to delayed.signal
and being lost.
V5: - rework list add and delete to account for idle threads
v6: - Use sigdelsetmask when removing stop signals
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200447
Reported-by: Wen Yang <[email protected]> and
Reported-by: majiang <[email protected]>
Fixes: 4a2c7a7837 ("[PATCH] make fork() atomic wrt pgrp/session signals")
Signed-off-by: "Eric W. Biederman" <[email protected]>
There are only two signals that are delivered to every member of a
signal group: SIGSTOP and SIGKILL. Signal delivery requires every
signal appear to be delivered either before or after a clone syscall.
SIGKILL terminates the clone so does not need to be considered. Which
leaves only SIGSTOP that needs to be considered when creating new
threads.
Today in the event of a group stop TIF_SIGPENDING will get set and the
fork will restart ensuring the fork syscall participates in the group
stop.
A fork (especially of a process with a lot of memory) is one of the
most expensive system so we really only want to restart a fork when
necessary.
It is easy so check to see if a SIGSTOP is ongoing and have the new
thread join it immediate after the clone completes. Making it appear
the clone completed happened just before the SIGSTOP.
The calculate_sigpending function will see the bits set in jobctl and
set TIF_SIGPENDING to ensure the new task takes the slow path to userspace.
V2: The call to task_join_group_stop was moved before the new task is
added to the thread group list. This should not matter as
sighand->siglock is held over both the addition of the threads,
the call to task_join_group_stop and do_signal_stop. But the change
is trivial and it is one less thing to worry about when reading
the code.
Signed-off-by: "Eric W. Biederman" <[email protected]>
The code in calculate_sigpending will now handle this so
it is just redundant and possibly a little confusing
to continue setting TIF_SIGPENDING in ptrace_init_task.
Suggested-by: Oleg Nesterov <[email protected]>
Signed-off-by: "Eric W. Biederman" <[email protected]>
Add a function calculate_sigpending to test to see if any signals are
pending for a new task immediately following fork. Signals have to
happen either before or after fork. Today our practice is to push
all of the signals to before the fork, but that has the downside that
frequent or periodic signals can make fork take much much longer than
normal or prevent fork from completing entirely.
So we need move signals that we can after the fork to prevent that.
This updates the code to set TIF_SIGPENDING on a new task if there
are signals or other activities that have moved so that they appear
to happen after the fork.
As the code today restarts if it sees any such activity this won't
immediately have an effect, as there will be no reason for it
to set TIF_SIGPENDING immediately after the fork.
Adding calculate_sigpending means the code in fork can safely be
changed to not always restart if a signal is pending.
The new calculate_sigpending function sets sigpending if there
are pending bits in jobctl, pending signals, the freezer needs
to freeze the new task or the live kernel patching framework
need the new thread to take the slow path to userspace.
I have verified that setting TIF_SIGPENDING does make a new process
take the slow path to userspace before it executes it's first userspace
instruction.
I have looked at the callers of signal_wake_up and the code paths
setting TIF_SIGPENDING and I don't see anything else that needs to be
handled. The code probably doesn't need to set TIF_SIGPENDING for the
kernel live patching as it uses a separate thread flag as well. But
at this point it seems safer reuse the recalc_sigpending logic and get
the kernel live patching folks to sort out their story later.
V2: I have moved the test into schedule_tail where siglock can
be grabbed and recalc_sigpending can be reused directly.
Further as the last action of setting up a new task this
guarantees that TIF_SIGPENDING will be properly set in the
new process.
The helper calculate_sigpending takes the siglock and
uncontitionally sets TIF_SIGPENDING and let's recalc_sigpending
clear TIF_SIGPENDING if it is unnecessary. This allows reusing
the existing code and keeps maintenance of the conditions simple.
Oleg Nesterov <[email protected]> suggested the movement
and pointed out the need to take siglock if this code
was going to be called while the new task is discoverable.
Signed-off-by: "Eric W. Biederman" <[email protected]>
In practice this does not change anything as testing for fatal_signal_pending
and exiting for with an error code duplicates the work of the next clause
which recalculates pending signals and then exits fork if any are pending.
In both cases the pending signal will trigger the slow path when existing
to userspace, and the fatal signal will cause do_exit to be called.
The advantage of making this a separate test is that it makes it clear
processing the fatal signal will terminate the fork, and it allows the
rest of the signal logic to be updated without fear that this important
case will be lost.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Normally this would be something that would be handled by handling
signals that are sent to a group of processes but in this case the
forking process is not a member of the group being signaled. Thus
special code is needed to prevent a race with pid namespaces exiting,
and fork adding new processes within them.
Move this test up before the signal restart just in case signals are
also pending. Fatal conditions should take presedence over restarts.
Signed-off-by: "Eric W. Biederman" <[email protected]>
This is the bottom and by pushing this down it simplifies the callers
and otherwise leaves things as is. This is in preparation for allowing
fork to implement better handling of signals set to groups of processes.
Signed-off-by: "Eric W. Biederman" <[email protected]>
This information is already available in the callers and by pushing it
down it makes the code a little clearer, and allows implementing
better handling of signales set to a group of processes in fork.
Signed-off-by: "Eric W. Biederman" <[email protected]>
This information is already available in the callers and by pushing it
down it makes the code a little clearer, and allows better group
signal behavior in fork.
Signed-off-by: "Eric W. Biederman" <[email protected]>
This passes the information we already have at the call sight into
do_send_sig_info. Ultimately allowing for better handling of signals
sent to a group of processes during fork.
Signed-off-by: "Eric W. Biederman" <[email protected]>
Eric W. Biederman