linux-stable.git/arch/x86/entry, branch linux-6.8.y

x86/mm: Remove broken vsyscall emulation code from the page fault code

2024-05-30T07:48:54+00:00

[ Upstream commit 02b670c1f88e78f42a6c5aee155c7b26960ca054 ]

The syzbot-reported stack trace from hell in this discussion thread
actually has three nested page faults:

https://lore.kernel.org/r/000000000000d5f4fc0616e816d4@google.com

... and I think that's actually the important thing here:

- the first page fault is from user space, and triggers the vsyscall
emulation.

- the second page fault is from __do_sys_gettimeofday(), and that should
just have caused the exception that then sets the return value to
-EFAULT

- the third nested page fault is due to _raw_spin_unlock_irqrestore() ->
preempt_schedule() -> trace_sched_switch(), which then causes a BPF
trace program to run, which does that bpf_probe_read_compat(), which
causes that page fault under pagefault_disable().

It's quite the nasty backtrace, and there's a lot going on.

The problem is literally the vsyscall emulation, which sets

current->thread.sig_on_uaccess_err = 1;

and that causes the fixup_exception() code to send the signal *despite* the
exception being caught.

And I think that is in fact completely bogus. It's completely bogus
exactly because it sends that signal even when it *shouldn't* be sent -
like for the BPF user mode trace gathering.

In other words, I think the whole "sig_on_uaccess_err" thing is entirely
broken, because it makes any nested page-faults do all the wrong things.

Now, arguably, I don't think anybody should enable vsyscall emulation any
more, but this test case clearly does.

I think we should just make the "send SIGSEGV" be something that the
vsyscall emulation does on its own, not this broken per-thread state for
something that isn't actually per thread.

The x86 page fault code actually tried to deal with the "incorrect nesting"
by having that:

if (in_interrupt())
return;

which ignores the sig_on_uaccess_err case when it happens in interrupts,
but as shown by this example, these nested page faults do not need to be
about interrupts at all.

IOW, I think the only right thing is to remove that horrendously broken
code.

The attached patch looks like the ObviouslyCorrect(tm) thing to do.

NOTE! This broken code goes back to this commit in 2011:

4fc3490114bb ("x86-64: Set siginfo and context on vsyscall emulation faults")

... and back then the reason was to get all the siginfo details right.
Honestly, I do not for a moment believe that it's worth getting the siginfo
details right here, but part of the commit says:

This fixes issues with UML when vsyscall=emulate.

... and so my patch to remove this garbage will probably break UML in this
situation.

I do not believe that anybody should be running with vsyscall=emulate in
2024 in the first place, much less if you are doing things like UML. But
let's see if somebody screams.

Reported-and-tested-by: syzbot+83e7f982ca045ab4405c@syzkaller.appspotmail.com
Signed-off-by: Linus Torvalds
Signed-off-by: Ingo Molnar
Tested-by: Jiri Olsa
Acked-by: Andy Lutomirski
Link: https://lore.kernel.org/r/CAHk-=wh9D6f7HUkDgZHKmDCHUQmp+Co89GP+b8+z+G56BKeyNg@mail.gmail.com
Signed-off-by: Sasha Levin

x86/vdso: Fix rethunk patching for vdso-image-x32.o too

2024-04-13T11:10:12+00:00

commit 4969d75dd9077e19e175e60f3c5a6c7653252e63 upstream.

In a similar fashion to

  b388e57d4628 ("x86/vdso: Fix rethunk patching for vdso-image-{32,64}.o")

annotate vdso-image-x32.o too for objtool so that it gets annotated
properly and the unused return thunk warning doesn't fire.

Reported-by: kernel test robot 
Closes: https://lore.kernel.org/oe-lkp/202403251454.23df6278-lkp@intel.com
Signed-off-by: Borislav Petkov (AMD) 
Link: https://lore.kernel.org/r/202403251454.23df6278-lkp@intel.com
Signed-off-by: Greg Kroah-Hartman

x86/vdso: Fix rethunk patching for vdso-image-{32,64}.o

2024-04-13T11:10:05+00:00

[ Upstream commit b388e57d4628eb22782bdad4cd5b83ca87a1b7c9 ]

For CONFIG_RETHUNK kernels, objtool annotates all the function return
sites so they can be patched during boot.  By design, after
apply_returns() is called, all tail-calls to the compiler-generated
default return thunk (__x86_return_thunk) should be patched out and
replaced with whatever's needed for any mitigations (or lack thereof).

The commit

  4461438a8405 ("x86/retpoline: Ensure default return thunk isn't used at runtime")

adds a runtime check and a WARN_ONCE() if the default return thunk ever
gets executed after alternatives have been applied.  This warning is
a sanity check to make sure objtool and apply_returns() are doing their
job.

As Nathan reported, that check found something:

  Unpatched return thunk in use. This should not happen!
  WARNING: CPU: 0 PID: 1 at arch/x86/kernel/cpu/bugs.c:2856 __warn_thunk+0x27/0x40
  RIP: 0010:__warn_thunk+0x27/0x40
  Call Trace:
   
   ? show_regs
   ? __warn
   ? __warn_thunk
   ? report_bug
   ? console_unlock
   ? handle_bug
   ? exc_invalid_op
   ? asm_exc_invalid_op
   ? ia32_binfmt_init
   ? __warn_thunk
   warn_thunk_thunk
   do_one_initcall
   kernel_init_freeable
   ? __pfx_kernel_init
   kernel_init
   ret_from_fork
   ? __pfx_kernel_init
   ret_from_fork_asm
   

Boris debugged to find that the unpatched return site was in
init_vdso_image_64(), and its translation unit wasn't being analyzed by
objtool, so it never got annotated.  So it got ignored by
apply_returns().

This is only a minor issue, as this function is only called during boot.
Still, objtool needs full visibility to the kernel.  Fix it by enabling
objtool on vdso-image-{32,64}.o.

Note this problem can only be seen with !CONFIG_X86_KERNEL_IBT, as that
requires objtool to run individually on all translation units rather on
vmlinux.o.

  [ bp: Massage commit message. ]

Reported-by: Nathan Chancellor 
Signed-off-by: Josh Poimboeuf 
Signed-off-by: Borislav Petkov (AMD) 
Link: https://lore.kernel.org/r/20240215032049.GA3944823@dev-arch.thelio-3990X
Signed-off-by: Sasha Levin

x86/bhi: Add support for clearing branch history at syscall entry

2024-04-10T14:38:24+00:00

commit 7390db8aea0d64e9deb28b8e1ce716f5020c7ee5 upstream.

Branch History Injection (BHI) attacks may allow a malicious application to
influence indirect branch prediction in kernel by poisoning the branch
history. eIBRS isolates indirect branch targets in ring0.  The BHB can
still influence the choice of indirect branch predictor entry, and although
branch predictor entries are isolated between modes when eIBRS is enabled,
the BHB itself is not isolated between modes.

Alder Lake and new processors supports a hardware control BHI_DIS_S to
mitigate BHI.  For older processors Intel has released a software sequence
to clear the branch history on parts that don't support BHI_DIS_S. Add
support to execute the software sequence at syscall entry and VMexit to
overwrite the branch history.

For now, branch history is not cleared at interrupt entry, as malicious
applications are not believed to have sufficient control over the
registers, since previous register state is cleared at interrupt
entry. Researchers continue to poke at this area and it may become
necessary to clear at interrupt entry as well in the future.

This mitigation is only defined here. It is enabled later.

Signed-off-by: Pawan Gupta 
Co-developed-by: Daniel Sneddon 
Signed-off-by: Daniel Sneddon 
Signed-off-by: Thomas Gleixner 
Reviewed-by: Alexandre Chartre 
Reviewed-by: Josh Poimboeuf 
Signed-off-by: Daniel Sneddon 
Signed-off-by: Greg Kroah-Hartman

x86/syscall: Don't force use of indirect calls for system calls

2024-04-10T14:38:24+00:00

commit 1e3ad78334a69b36e107232e337f9d693dcc9df2 upstream.

Make  build a switch statement instead, and the compiler can
either decide to generate an indirect jump, or - more likely these days due
to mitigations - just a series of conditional branches.

Yes, the conditional branches also have branch prediction, but the branch
prediction is much more controlled, in that it just causes speculatively
running the wrong system call (harmless), rather than speculatively running
possibly wrong random less controlled code gadgets.

This doesn't mitigate other indirect calls, but the system call indirection
is the first and most easily triggered case.

Signed-off-by: Linus Torvalds 
Signed-off-by: Daniel Sneddon 
Signed-off-by: Thomas Gleixner 
Reviewed-by: Josh Poimboeuf 
Signed-off-by: Daniel Sneddon 
Signed-off-by: Greg Kroah-Hartman

x86/entry_32: Add VERW just before userspace transition

2024-02-20T00:31:46+00:00

As done for entry_64, add support for executing VERW late in exit to
user path for 32-bit mode.

Signed-off-by: Pawan Gupta 
Signed-off-by: Dave Hansen 
Link: https://lore.kernel.org/all/20240213-delay-verw-v8-3-a6216d83edb7%40linux.intel.com

x86/entry_64: Add VERW just before userspace transition

2024-02-20T00:31:42+00:00

Mitigation for MDS is to use VERW instruction to clear any secrets in
CPU Buffers. Any memory accesses after VERW execution can still remain
in CPU buffers. It is safer to execute VERW late in return to user path
to minimize the window in which kernel data can end up in CPU buffers.
There are not many kernel secrets to be had after SWITCH_TO_USER_CR3.

Add support for deploying VERW mitigation after user register state is
restored. This helps minimize the chances of kernel data ending up into
CPU buffers after executing VERW.

Note that the mitigation at the new location is not yet enabled.

  Corner case not handled
  =======================
  Interrupts returning to kernel don't clear CPUs buffers since the
  exit-to-user path is expected to do that anyways. But, there could be
  a case when an NMI is generated in kernel after the exit-to-user path
  has cleared the buffers. This case is not handled and NMI returning to
  kernel don't clear CPU buffers because:

  1. It is rare to get an NMI after VERW, but before returning to userspace.
  2. For an unprivileged user, there is no known way to make that NMI
     less rare or target it.
  3. It would take a large number of these precisely-timed NMIs to mount
     an actual attack.  There's presumably not enough bandwidth.
  4. The NMI in question occurs after a VERW, i.e. when user state is
     restored and most interesting data is already scrubbed. Whats left
     is only the data that NMI touches, and that may or may not be of
     any interest.

Suggested-by: Dave Hansen 
Signed-off-by: Pawan Gupta 
Signed-off-by: Dave Hansen 
Link: https://lore.kernel.org/all/20240213-delay-verw-v8-2-a6216d83edb7%40linux.intel.com

x86/bugs: Add asm helpers for executing VERW

2024-02-20T00:31:33+00:00

MDS mitigation requires clearing the CPU buffers before returning to
user. This needs to be done late in the exit-to-user path. Current
location of VERW leaves a possibility of kernel data ending up in CPU
buffers for memory accesses done after VERW such as:

  1. Kernel data accessed by an NMI between VERW and return-to-user can
     remain in CPU buffers since NMI returning to kernel does not
     execute VERW to clear CPU buffers.
  2. Alyssa reported that after VERW is executed,
     CONFIG_GCC_PLUGIN_STACKLEAK=y scrubs the stack used by a system
     call. Memory accesses during stack scrubbing can move kernel stack
     contents into CPU buffers.
  3. When caller saved registers are restored after a return from
     function executing VERW, the kernel stack accesses can remain in
     CPU buffers(since they occur after VERW).

To fix this VERW needs to be moved very late in exit-to-user path.

In preparation for moving VERW to entry/exit asm code, create macros
that can be used in asm. Also make VERW patching depend on a new feature
flag X86_FEATURE_CLEAR_CPU_BUF.

Reported-by: Alyssa Milburn 
Suggested-by: Andrew Cooper 
Suggested-by: Peter Zijlstra 
Signed-off-by: Pawan Gupta 
Signed-off-by: Dave Hansen 
Link: https://lore.kernel.org/all/20240213-delay-verw-v8-1-a6216d83edb7%40linux.intel.com

Merge tag 'asm-generic-6.8' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic

2024-01-11T02:13:44+00:00

Pull asm-generic cleanups from Arnd Bergmann:
 "A series from Baoquan He cleans up the asm-generic/io.h to remove the
  ioremap_uc() definition from everything except x86, which still needs
  it for pre-PAT systems. This series notably contains a patch from
  Jiaxun Yang that converts MIPS to use asm-generic/io.h like every
  other architecture does, enabling future cleanups.

  Some of my own patches fix -Wmissing-prototype warnings in
  architecture specific code across several architectures. This is now
  needed as the warning is enabled by default. There are still some
  remaining warnings in minor platforms, but the series should catch
  most of the widely used ones make them more consistent with one
  another.

  David McKay fixes a bug in __generic_cmpxchg_local() when this is used
  on 64-bit architectures. This could currently only affect parisc64 and
  sparc64.

  Additional cleanups address from Linus Walleij, Uwe Kleine-König,
  Thomas Huth, and Kefeng Wang help reduce unnecessary inconsistencies
  between architectures"

* tag 'asm-generic-6.8' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
  asm-generic: Fix 32 bit __generic_cmpxchg_local
  Hexagon: Make pfn accessors statics inlines
  ARC: mm: Make virt_to_pfn() a static inline
  mips: remove extraneous asm-generic/iomap.h include
  sparc: Use $(kecho) to announce kernel images being ready
  arm64: vdso32: Define BUILD_VDSO32_64 to correct prototypes
  csky: fix arch_jump_label_transform_static override
  arch: add do_page_fault prototypes
  arch: add missing prepare_ftrace_return() prototypes
  arch: vdso: consolidate gettime prototypes
  arch: include linux/cpu.h for trap_init() prototype
  arch: fix asm-offsets.c building with -Wmissing-prototypes
  arch: consolidate arch_irq_work_raise prototypes
  hexagon: Remove CONFIG_HEXAGON_ARCH_VERSION from uapi header
  asm/io: remove unnecessary xlate_dev_mem_ptr() and unxlate_dev_mem_ptr()
  mips: io: remove duplicated codes
  arch/*/io.h: remove ioremap_uc in some architectures
  mips: add  including

Merge tag 'lsm-pr-20240105' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/lsm

2024-01-09T20:57:46+00:00

Pull security module updates from Paul Moore:

 - Add three new syscalls: lsm_list_modules(), lsm_get_self_attr(), and
   lsm_set_self_attr().

   The first syscall simply lists the LSMs enabled, while the second and
   third get and set the current process' LSM attributes. Yes, these
   syscalls may provide similar functionality to what can be found under
   /proc or /sys, but they were designed to support multiple,
   simultaneaous (stacked) LSMs from the start as opposed to the current
   /proc based solutions which were created at a time when only one LSM
   was allowed to be active at a given time.

   We have spent considerable time discussing ways to extend the
   existing /proc interfaces to support multiple, simultaneaous LSMs and
   even our best ideas have been far too ugly to support as a kernel
   API; after +20 years in the kernel, I felt the LSM layer had
   established itself enough to justify a handful of syscalls.

   Support amongst the individual LSM developers has been nearly
   unanimous, with a single objection coming from Tetsuo (TOMOYO) as he
   is worried that the LSM_ID_XXX token concept will make it more
   difficult for out-of-tree LSMs to survive. Several members of the LSM
   community have demonstrated the ability for out-of-tree LSMs to
   continue to exist by picking high/unused LSM_ID values as well as
   pointing out that many kernel APIs rely on integer identifiers, e.g.
   syscalls (!), but unfortunately Tetsuo's objections remain.

   My personal opinion is that while I have no interest in penalizing
   out-of-tree LSMs, I'm not going to penalize in-tree development to
   support out-of-tree development, and I view this as a necessary step
   forward to support the push for expanded LSM stacking and reduce our
   reliance on /proc and /sys which has occassionally been problematic
   for some container users. Finally, we have included the linux-api
   folks on (all?) recent revisions of the patchset and addressed all of
   their concerns.

 - Add a new security_file_ioctl_compat() LSM hook to handle the 32-bit
   ioctls on 64-bit systems problem.

   This patch includes support for all of the existing LSMs which
   provide ioctl hooks, although it turns out only SELinux actually
   cares about the individual ioctls. It is worth noting that while
   Casey (Smack) and Tetsuo (TOMOYO) did not give explicit ACKs to this
   patch, they did both indicate they are okay with the changes.

 - Fix a potential memory leak in the CALIPSO code when IPv6 is disabled
   at boot.

   While it's good that we are fixing this, I doubt this is something
   users are seeing in the wild as you need to both disable IPv6 and
   then attempt to configure IPv6 labeled networking via
   NetLabel/CALIPSO; that just doesn't make much sense.

   Normally this would go through netdev, but Jakub asked me to take
   this patch and of all the trees I maintain, the LSM tree seemed like
   the best fit.

 - Update the LSM MAINTAINERS entry with additional information about
   our process docs, patchwork, bug reporting, etc.

   I also noticed that the Lockdown LSM is missing a dedicated
   MAINTAINERS entry so I've added that to the pull request. I've been
   working with one of the major Lockdown authors/contributors to see if
   they are willing to step up and assume a Lockdown maintainer role;
   hopefully that will happen soon, but in the meantime I'll continue to
   look after it.

 - Add a handful of mailmap entries for Serge Hallyn and myself.

* tag 'lsm-pr-20240105' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/lsm: (27 commits)
  lsm: new security_file_ioctl_compat() hook
  lsm: Add a __counted_by() annotation to lsm_ctx.ctx
  calipso: fix memory leak in netlbl_calipso_add_pass()
  selftests: remove the LSM_ID_IMA check in lsm/lsm_list_modules_test
  MAINTAINERS: add an entry for the lockdown LSM
  MAINTAINERS: update the LSM entry
  mailmap: add entries for Serge Hallyn's dead accounts
  mailmap: update/replace my old email addresses
  lsm: mark the lsm_id variables are marked as static
  lsm: convert security_setselfattr() to use memdup_user()
  lsm: align based on pointer length in lsm_fill_user_ctx()
  lsm: consolidate buffer size handling into lsm_fill_user_ctx()
  lsm: correct error codes in security_getselfattr()
  lsm: cleanup the size counters in security_getselfattr()
  lsm: don't yet account for IMA in LSM_CONFIG_COUNT calculation
  lsm: drop LSM_ID_IMA
  LSM: selftests for Linux Security Module syscalls
  SELinux: Add selfattr hooks
  AppArmor: Add selfattr hooks
  Smack: implement setselfattr and getselfattr hooks
  ...