== Background ==

Support for large, "dynamic" fpstates was recently merged.  This
included code to ensure that sigaltstacks are sufficiently sized for
these large states.  A new lock was added to remove races between
enabling large features and setting up sigaltstacks.

== Problem ==

The new lock (sigaltstack_lock()) is acquired in the sigreturn path
before restoring the old sigaltstack.  Unfortunately, contention on the
new lock causes a measurable signal handling performance regression [1].
However, the common case is that no *changes* are made to the
sigaltstack state at sigreturn.

== Solution ==

do_sigaltstack() acquires sigaltstack_lock() and is used for both
sys_sigaltstack() and restoring the sigaltstack in sys_sigreturn().
Check for changes to the sigaltstack before taking the lock.  If no
changes were made, return before acquiring the lock.

This removes lock contention from the common-case sigreturn path.

[1] https://lore.kernel.org/lkml/20211207012128.GA16074@xsang-OptiPlex-9020/

Fixes: 3aac3ebea08f ("x86/signal: Implement sigaltstack size validation")
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20211210225503.12734-1-chang.seok.bae@intel.com

Recently to prevent issues with SECCOMP_RET_KILL and similar signals
being changed before they are delivered SA_IMMUTABLE was added.

Unfortunately this broke debuggers[1][2] which reasonably expect
to be able to trap synchronous SIGTRAP and SIGSEGV even when
the target process is not configured to handle those signals.

Add force_exit_sig and use it instead of force_fatal_sig where
historically the code has directly called do_exit.  This has the
implementation benefits of going through the signal exit path
(including generating core dumps) without the danger of allowing
userspace to ignore or change these signals.

This avoids userspace regressions as older kernels exited with do_exit
which debuggers also can not intercept.

In the future is should be possible to improve the quality of
implementation of the kernel by changing some of these force_exit_sig
calls to force_fatal_sig.  That can be done where it matters on
a case-by-case basis with careful analysis.

Reported-by: Kyle Huey <me@kylehuey.com>
Reported-by: kernel test robot <oliver.sang@intel.com>
[1] https://lkml.kernel.org/r/CAP045AoMY4xf8aC_4QU_-j7obuEPYgTcnQQP3Yxk=2X90jtpjw@mail.gmail.com
[2] https://lkml.kernel.org/r/20211117150258.GB5403@xsang-OptiPlex-9020
Fixes: 00b06da29cf9 ("signal: Add SA_IMMUTABLE to ensure forced siganls do not get changed")
Fixes: a3616a3c0272 ("signal/m68k: Use force_sigsegv(SIGSEGV) in fpsp040_die")
Fixes: 83a1f27ad773 ("signal/powerpc: On swapcontext failure force SIGSEGV")
Fixes: 9bc508cf0791 ("signal/s390: Use force_sigsegv in default_trap_handler")
Fixes: 086ec444f866 ("signal/sparc32: In setup_rt_frame and setup_fram use force_fatal_sig")
Fixes: c317d306d550 ("signal/sparc32: Exit with a fatal signal when try_to_clear_window_buffer fails")
Fixes: 695dd0d634df ("signal/x86: In emulate_vsyscall force a signal instead of calling do_exit")
Fixes: 1fbd60df8a85 ("signal/vm86_32: Properly send SIGSEGV when the vm86 state cannot be saved.")
Fixes: 941edc5bf174 ("exit/syscall_user_dispatch: Send ordinary signals on failure")
Link: https://lkml.kernel.org/r/871r3dqfv8.fsf_-_@email.froward.int.ebiederm.org
Reviewed-by: Kees Cook <keescook@chromium.org>
Tested-by: Kees Cook <keescook@chromium.org>
Tested-by: Kyle Huey <khuey@kylehuey.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Recently to prevent issues with SECCOMP_RET_KILL and similar signals
being changed before they are delivered SA_IMMUTABLE was added.

Unfortunately this broke debuggers[1][2] which reasonably expect to be
able to trap synchronous SIGTRAP and SIGSEGV even when the target
process is not configured to handle those signals.

Update force_sig_to_task to support both the case when we can allow
the debugger to intercept and possibly ignore the signal and the case
when it is not safe to let userspace know about the signal until the
process has exited.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Kyle Huey <me@kylehuey.com>
Reported-by: kernel test robot <oliver.sang@intel.com>
Cc: stable@vger.kernel.org
[1] https://lkml.kernel.org/r/CAP045AoMY4xf8aC_4QU_-j7obuEPYgTcnQQP3Yxk=2X90jtpjw@mail.gmail.com
[2] https://lkml.kernel.org/r/20211117150258.GB5403@xsang-OptiPlex-9020
Fixes: 00b06da29cf9 ("signal: Add SA_IMMUTABLE to ensure forced siganls do not get changed")
Link: https://lkml.kernel.org/r/877dd5qfw5.fsf_-_@email.froward.int.ebiederm.org
Reviewed-by: Kees Cook <keescook@chromium.org>
Tested-by: Kees Cook <keescook@chromium.org>
Tested-by: Kyle Huey <khuey@kylehuey.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Pull exit cleanups from Eric Biederman:
 "While looking at some issues related to the exit path in the kernel I
  found several instances where the code is not using the existing
  abstractions properly.

  This set of changes introduces force_fatal_sig a way of sending a
  signal and not allowing it to be caught, and corrects the misuse of
  the existing abstractions that I found.

  A lot of the misuse of the existing abstractions are silly things such
  as doing something after calling a no return function, rolling BUG by
  hand, doing more work than necessary to terminate a kernel thread, or
  calling do_exit(SIGKILL) instead of calling force_sig(SIGKILL).

  In the review a deficiency in force_fatal_sig and force_sig_seccomp
  where ptrace or sigaction could prevent the delivery of the signal was
  found. I have added a change that adds SA_IMMUTABLE to change that
  makes it impossible to interrupt the delivery of those signals, and
  allows backporting to fix force_sig_seccomp

  And Arnd found an issue where a function passed to kthread_run had the
  wrong prototype, and after my cleanup was failing to build."

* 'exit-cleanups-for-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (23 commits)
  soc: ti: fix wkup_m3_rproc_boot_thread return type
  signal: Add SA_IMMUTABLE to ensure forced siganls do not get changed
  signal: Replace force_sigsegv(SIGSEGV) with force_fatal_sig(SIGSEGV)
  exit/r8188eu: Replace the macro thread_exit with a simple return 0
  exit/rtl8712: Replace the macro thread_exit with a simple return 0
  exit/rtl8723bs: Replace the macro thread_exit with a simple return 0
  signal/x86: In emulate_vsyscall force a signal instead of calling do_exit
  signal/sparc32: In setup_rt_frame and setup_fram use force_fatal_sig
  signal/sparc32: Exit with a fatal signal when try_to_clear_window_buffer fails
  exit/syscall_user_dispatch: Send ordinary signals on failure
  signal: Implement force_fatal_sig
  exit/kthread: Have kernel threads return instead of calling do_exit
  signal/s390: Use force_sigsegv in default_trap_handler
  signal/vm86_32: Properly send SIGSEGV when the vm86 state cannot be saved.
  signal/vm86_32: Replace open coded BUG_ON with an actual BUG_ON
  signal/sparc: In setup_tsb_params convert open coded BUG into BUG
  signal/powerpc: On swapcontext failure force SIGSEGV
  signal/sh: Use force_sig(SIGKILL) instead of do_group_exit(SIGKILL)
  signal/mips: Update (_save|_restore)_fp_context to fail with -EFAULT
  signal/sparc32: Remove unreachable do_exit in do_sparc_fault
  ...

Pull per signal_struct coredumps from Eric Biederman:
 "Current coredumps are mixed up with the exit code, the signal handling
  code, and the ptrace code making coredumps much more complicated than
  necessary and difficult to follow.

  This series of changes starts with ptrace_stop and cleans it up,
  making it easier to follow what is happening in ptrace_stop. Then
  cleans up the exec interactions with coredumps. Then cleans up the
  coredump interactions with exit. Finally the coredump interactions
  with the signal handling code is cleaned up.

  The first and last changes are bug fixes for minor bugs.

  I believe the fact that vfork followed by execve can kill the process
  the called vfork if exec fails is sufficient justification to change
  the userspace visible behavior.

  In previous discussions some of these changes were organized
  differently and individually appeared to make the code base worse. As
  currently written I believe they all stand on their own as cleanups
  and bug fixes.

  Which means that even if the worst should happen and the last change
  needs to be reverted for some unimaginable reason, the code base will
  still be improved.

  If the worst does not happen there are a more cleanups that can be
  made. Signals that generate coredumps can easily become eligible for
  short circuit delivery in complete_signal. The entire rendezvous for
  generating a coredump can move into get_signal. The function
  force_sig_info_to_task be written in a way that does not modify the
  signal handling state of the target task (because coredumps are
  eligible for short circuit delivery). Many of these future cleanups
  can be done another way but nothing so cleanly as if coredumps become
  per signal_struct"

* 'per_signal_struct_coredumps-for-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
  coredump: Limit coredumps to a single thread group
  coredump:  Don't perform any cleanups before dumping core
  exit: Factor coredump_exit_mm out of exit_mm
  exec: Check for a pending fatal signal instead of core_state
  ptrace: Remove the unnecessary arguments from arch_ptrace_stop
  signal: Remove the bogus sigkill_pending in ptrace_stop

As Andy pointed out that there are races between
force_sig_info_to_task and sigaction[1] when force_sig_info_task.  As
Kees discovered[2] ptrace is also able to change these signals.

In the case of seeccomp killing a process with a signal it is a
security violation to allow the signal to be caught or manipulated.

Solve this problem by introducing a new flag SA_IMMUTABLE that
prevents sigaction and ptrace from modifying these forced signals.
This flag is carefully made kernel internal so that no new ABI is
introduced.

Longer term I think this can be solved by guaranteeing short circuit
delivery of signals in this case.  Unfortunately reliable and
guaranteed short circuit delivery of these signals is still a ways off
from being implemented, tested, and merged.  So I have implemented a much
simpler alternative for now.

[1] https://lkml.kernel.org/r/b5d52d25-7bde-4030-a7b1-7c6f8ab90660@www.fastmail.com
[2] https://lkml.kernel.org/r/202110281136.5CE65399A7@keescook
Cc: stable@vger.kernel.org
Fixes: 307d522f5eb8 ("signal/seccomp: Refactor seccomp signal and coredump generation")
Tested-by: Andrea Righi <andrea.righi@canonical.com>
Tested-by: Kees Cook <keescook@chromium.org>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Pull x86 fpu updates from Thomas Gleixner:

 - Cleanup of extable fixup handling to be more robust, which in turn
   allows to make the FPU exception fixups more robust as well.

 - Change the return code for signal frame related failures from
   explicit error codes to a boolean fail/success as that's all what the
   calling code evaluates.

 - A large refactoring of the FPU code to prepare for adding AMX
   support:

      - Distangle the public header maze and remove especially the
        misnomed kitchen sink internal.h which is despite it's name
        included all over the place.

      - Add a proper abstraction for the register buffer storage (struct
        fpstate) which allows to dynamically size the buffer at runtime
        by flipping the pointer to the buffer container from the default
        container which is embedded in task_struct::tread::fpu to a
        dynamically allocated container with a larger register buffer.

      - Convert the code over to the new fpstate mechanism.

      - Consolidate the KVM FPU handling by moving the FPU related code
        into the FPU core which removes the number of exports and avoids
        adding even more export when AMX has to be supported in KVM.
        This also removes duplicated code which was of course
        unnecessary different and incomplete in the KVM copy.

      - Simplify the KVM FPU buffer handling by utilizing the new
        fpstate container and just switching the buffer pointer from the
        user space buffer to the KVM guest buffer when entering
        vcpu_run() and flipping it back when leaving the function. This
        cuts the memory requirements of a vCPU for FPU buffers in half
        and avoids pointless memory copy operations.

        This also solves the so far unresolved problem of adding AMX
        support because the current FPU buffer handling of KVM inflicted
        a circular dependency between adding AMX support to the core and
        to KVM. With the new scheme of switching fpstate AMX support can
        be added to the core code without affecting KVM.

      - Replace various variables with proper data structures so the
        extra information required for adding dynamically enabled FPU
        features (AMX) can be added in one place

 - Add AMX (Advanced Matrix eXtensions) support (finally):

   AMX is a large XSTATE component which is going to be available with
   Saphire Rapids XEON CPUs. The feature comes with an extra MSR
   (MSR_XFD) which allows to trap the (first) use of an AMX related
   instruction, which has two benefits:

    1) It allows the kernel to control access to the feature

    2) It allows the kernel to dynamically allocate the large register
       state buffer instead of burdening every task with the the extra
       8K or larger state storage.

   It would have been great to gain this kind of control already with
   AVX512.

   The support comes with the following infrastructure components:

    1) arch_prctl() to
        - read the supported features (equivalent to XGETBV(0))
        - read the permitted features for a task
        - request permission for a dynamically enabled feature

       Permission is granted per process, inherited on fork() and
       cleared on exec(). The permission policy of the kernel is
       restricted to sigaltstack size validation, but the syscall
       obviously allows further restrictions via seccomp etc.

    2) A stronger sigaltstack size validation for sys_sigaltstack(2)
       which takes granted permissions and the potentially resulting
       larger signal frame into account. This mechanism can also be used
       to enforce factual sigaltstack validation independent of dynamic
       features to help with finding potential victims of the 2K
       sigaltstack size constant which is broken since AVX512 support
       was added.

    3) Exception handling for #NM traps to catch first use of a extended
       feature via a new cause MSR. If the exception was caused by the
       use of such a feature, the handler checks permission for that
       feature. If permission has not been granted, the handler sends a
       SIGILL like the #UD handler would do if the feature would have
       been disabled in XCR0. If permission has been granted, then a new
       fpstate which fits the larger buffer requirement is allocated.

       In the unlikely case that this allocation fails, the handler
       sends SIGSEGV to the task. That's not elegant, but unavoidable as
       the other discussed options of preallocation or full per task
       permissions come with their own set of horrors for kernel and/or
       userspace. So this is the lesser of the evils and SIGSEGV caused
       by unexpected memory allocation failures is not a fundamentally
       new concept either.

       When allocation succeeds, the fpstate properties are filled in to
       reflect the extended feature set and the resulting sizes, the
       fpu::fpstate pointer is updated accordingly and the trap is
       disarmed for this task permanently.

    4) Enumeration and size calculations

    5) Trap switching via MSR_XFD

       The XFD (eXtended Feature Disable) MSR is context switched with
       the same life time rules as the FPU register state itself. The
       mechanism is keyed off with a static key which is default
       disabled so !AMX equipped CPUs have zero overhead. On AMX enabled
       CPUs the overhead is limited by comparing the tasks XFD value
       with a per CPU shadow variable to avoid redundant MSR writes. In
       case of switching from a AMX using task to a non AMX using task
       or vice versa, the extra MSR write is obviously inevitable.

       All other places which need to be aware of the variable feature
       sets and resulting variable sizes are not affected at all because
       they retrieve the information (feature set, sizes) unconditonally
       from the fpstate properties.

    6) Enable the new AMX states

   Note, this is relatively new code despite the fact that AMX support
   is in the works for more than a year now.

   The big refactoring of the FPU code, which allowed to do a proper
   integration has been started exactly 3 weeks ago. Refactoring of the
   existing FPU code and of the original AMX patches took a week and has
   been subject to extensive review and testing. The only fallout which
   has not been caught in review and testing right away was restricted
   to AMX enabled systems, which is completely irrelevant for anyone
   outside Intel and their early access program. There might be dragons
   lurking as usual, but so far the fine grained refactoring has held up
   and eventual yet undetected fallout is bisectable and should be
   easily addressable before the 5.16 release. Famous last words...

   Many thanks to Chang Bae and Dave Hansen for working hard on this and
   also to the various test teams at Intel who reserved extra capacity
   to follow the rapid development of this closely which provides the
   confidence level required to offer this rather large update for
   inclusion into 5.16-rc1

* tag 'x86-fpu-2021-11-01' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (110 commits)
  Documentation/x86: Add documentation for using dynamic XSTATE features
  x86/fpu: Include vmalloc.h for vzalloc()
  selftests/x86/amx: Add context switch test
  selftests/x86/amx: Add test cases for AMX state management
  x86/fpu/amx: Enable the AMX feature in 64-bit mode
  x86/fpu: Add XFD handling for dynamic states
  x86/fpu: Calculate the default sizes independently
  x86/fpu/amx: Define AMX state components and have it used for boot-time checks
  x86/fpu/xstate: Prepare XSAVE feature table for gaps in state component numbers
  x86/fpu/xstate: Add fpstate_realloc()/free()
  x86/fpu/xstate: Add XFD #NM handler
  x86/fpu: Update XFD state where required
  x86/fpu: Add sanity checks for XFD
  x86/fpu: Add XFD state to fpstate
  x86/msr-index: Add MSRs for XFD
  x86/cpufeatures: Add eXtended Feature Disabling (XFD) feature bit
  x86/fpu: Reset permission and fpstate on exec()
  x86/fpu: Prepare fpu_clone() for dynamically enabled features
  x86/fpu/signal: Prepare for variable sigframe length
  x86/signal: Use fpu::__state_user_size for sigalt stack validation
  ...

Add a simple helper force_fatal_sig that causes a signal to be
delivered to a process as if the signal handler was set to SIG_DFL.

Reimplement force_sigsegv based upon this new helper.  This fixes
force_sigsegv so that when it forces the default signal handler
to be used the code now forces the signal to be unblocked as well.

Reusing the tested logic in force_sig_info_to_task that was built for
force_sig_seccomp this makes the implementation trivial.

This is interesting both because it makes force_sigsegv simpler and
because there are a couple of buggy places in the kernel that call
do_exit(SIGILL) or do_exit(SIGSYS) because there is no straight
forward way today for those places to simply force the exit of a
process with the chosen signal.  Creating force_fatal_sig allows
those places to be implemented with normal signal exits.

Link: https://lkml.kernel.org/r/20211020174406.17889-13-ebiederm@xmission.com
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>

New x86 FPU features will be very large, requiring ~10k of stack in
signal handlers.  These new features require a new approach called
"dynamic features".

The kernel currently tries to ensure that altstacks are reasonably
sized. Right now, on x86, sys_sigaltstack() requires a size of >=2k.
However, that 2k is a constant. Simply raising that 2k requirement
to >10k for the new features would break existing apps which have a
compiled-in size of 2k.

Instead of universally enforcing a larger stack, prohibit a process from
using dynamic features without properly-sized altstacks. This must be
enforced in two places:

 * A dynamic feature can not be enabled without an large-enough altstack
   for each process thread.
 * Once a dynamic feature is enabled, any request to install a too-small
   altstack will be rejected

The dynamic feature enabling code must examine each thread in a
process to ensure that the altstacks are large enough. Add a new lock
(sigaltstack_lock()) to ensure that threads can not race and change
their altstack after being examined.

Add the infrastructure in form of a config option and provide empty
stubs for architectures which do not need dynamic altstack size checks.

This implementation will be fleshed out for x86 in a future patch called

  x86/arch_prctl: Add controls for dynamic XSTATE components

  [dhansen: commit message. ]

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20211021225527.10184-2-chang.seok.bae@intel.com

Pull ucounts fixes from Eric Biederman:
 "There has been one very hard to track down bug in the ucount code that
  we have been tracking since roughly v5.14 was released. Alex managed
  to find a reliable reproducer a few days ago and then I was able to
  instrument the code and figure out what the issue was.

  It turns out the sigqueue_alloc single atomic operation optimization
  did not play nicely with ucounts multiple level rlimits. It turned out
  that either sigqueue_alloc or sigqueue_free could be operating on
  multiple levels and trigger the conditions for the optimization on
  more than one level at the same time.

  To deal with that situation I have introduced inc_rlimit_get_ucounts
  and dec_rlimit_put_ucounts that just focuses on the optimization and
  the rlimit and ucount changes.

  While looking into the big bug I found I couple of other little issues
  so I am including those fixes here as well.

  When I have time I would very much like to dig into process ownership
  of the shared signal queue and see if we could pick a single owner for
  the entire queue so that all of the rlimits can count to that owner.
  That should entirely remove the need to call get_ucounts and
  put_ucounts in sigqueue_alloc and sigqueue_free. It is difficult
  because Linux unlike POSIX supports setuid that works on a single
  thread"

* 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
  ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring
  ucounts: Proper error handling in set_cred_ucounts
  ucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds
  ucounts: Fix signal ucount refcounting