Skip to content
  1. Dec 24, 2022
  3. Dec 17, 2022
  5. Dec 15, 2022
  7. Dec 03, 2022
  9. Dec 02, 2022
    • Kees Cook's avatar
      exit: Expose "oops_count" to sysfs · 9db89b41
      Kees Cook authored 
      

Since Oops count is now tracked and is a fairly interesting signal, add
the entry /sys/kernel/oops_count to expose it to userspace.

Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Jann Horn <jannh@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: default avatarLuis Chamberlain <mcgrof@kernel.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221117234328.594699-3-keescook@chromium.org
      9db89b41
    • Jann Horn's avatar
      exit: Put an upper limit on how often we can oops · d4ccd54d
      Jann Horn authored 
      

Many Linux systems are configured to not panic on oops; but allowing an
attacker to oops the system **really** often can make even bugs that look
completely unexploitable exploitable (like NULL dereferences and such) if
each crash elevates a refcount by one or a lock is taken in read mode, and
this causes a counter to eventually overflow.

The most interesting counters for this are 32 bits wide (like open-coded
refcounts that don't use refcount_t). (The ldsem reader count on 32-bit
platforms is just 16 bits, but probably nobody cares about 32-bit platforms
that much nowadays.)

So let's panic the system if the kernel is constantly oopsing.

The speed of oopsing 2^32 times probably depends on several factors, like
how long the stack trace is and which unwinder you're using; an empirically
important one is whether your console is showing a graphical environment or
a text console that oopses will be printed to.
In a quick single-threaded benchmark, it looks like oopsing in a vfork()
child with a very short stack trace only takes ~510 microseconds per run
when a graphical console is active; but switching to a text console that
oopses are printed to slows it down around 87x, to ~45 milliseconds per
run.
(Adding more threads makes this faster, but the actual oops printing
happens under &die_lock on x86, so you can maybe speed this up by a factor
of around 2 and then any further improvement gets eaten up by lock
contention.)

It looks like it would take around 8-12 days to overflow a 32-bit counter
with repeated oopsing on a multi-core X86 system running a graphical
environment; both me (in an X86 VM) and Seth (with a distro kernel on
normal hardware in a standard configuration) got numbers in that ballpark.

12 days aren't *that* short on a desktop system, and you'd likely need much
longer on a typical server system (assuming that people don't run graphical
desktop environments on their servers), and this is a *very* noisy and
violent approach to exploiting the kernel; and it also seems to take orders
of magnitude longer on some machines, probably because stuff like EFI
pstore will slow it down a ton if that's active.

Signed-off-by: default avatarJann Horn <jannh@google.com>
Link: https://lore.kernel.org/r/20221107201317.324457-1-jannh@google.com


Reviewed-by: default avatarLuis Chamberlain <mcgrof@kernel.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221117234328.594699-2-keescook@chromium.org
      d4ccd54d
    • Kees Cook's avatar
      panic: Separate sysctl logic from CONFIG_SMP · 9360d035
      Kees Cook authored 
      

In preparation for adding more sysctls directly in kernel/panic.c, split
CONFIG_SMP from the logic that adds sysctls.

Cc: Petr Mladek <pmladek@suse.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: tangmeng <tangmeng@uniontech.com>
Cc: "Guilherme G. Piccoli" <gpiccoli@igalia.com>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: default avatarLuis Chamberlain <mcgrof@kernel.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221117234328.594699-1-keescook@chromium.org
      9360d035
    • Gustavo A. R. Silva's avatar
      mm/pgtable: Fix multiple -Wstringop-overflow warnings · 25226df4
      Gustavo A. R. Silva authored 
      The actual size of the following arrays at run-time depends on
CONFIG_X86_PAE.

427         pmd_t *u_pmds[MAX_PREALLOCATED_USER_PMDS];
428         pmd_t *pmds[MAX_PREALLOCATED_PMDS];

If CONFIG_X86_PAE is not enabled, their final size will be zero (which
is technically not a legal storage size in C, but remains "valid" via
the GNU extension). In that case, the compiler complains about trying to
access objects of size zero when calling functions where these objects
are passed as arguments.

Fix this by sanity-checking the size of those arrays just before the
function calls. Also, the following warnings are fixed by these changes
when building with GCC 11+ and -Wstringop-overflow enabled:

arch/x86/mm/pgtable.c:437:13: warning: ‘preallocate_pmds.constprop’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]
arch/x86/mm/pgtable.c:440:13: warning: ‘preallocate_pmds.constprop’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]
arch/x86/mm/pgtable.c:462:9: warning: ‘free_pmds.constprop’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]
arch/x86/mm/pgtable.c:455:9: warning: ‘pgd_prepopulate_user_pmd’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]
arch/x86/mm/pgtable.c:464:9: warning: ‘free_pmds.constprop’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]

This is one of the last cases in the ongoing effort to globally enable
-Wstringop-overflow.

The alternative to this is to make the originally suggested change:
make the pmds argument from an array pointer to a pointer pointer. That
situation is considered "legal" for C in the sense that it does not have
a way to reason about the storage. i.e.:

-static void pgd_prepopulate_pmd(struct mm_struct *mm, pgd_t *pgd, pmd_t *pmds[])
+static void pgd_prepopulate_pmd(struct mm_struct *mm, pgd_t *pgd, pmd_t **pmds)

With the above change, there's no difference in binary output, and the
compiler warning is silenced.

However, with this patch, the compiler can actually figure out that it
isn't using the code at all, and it gets dropped:

   text    data     bss     dec     hex filename
   8218     718      32    8968    2308 arch/x86/mm/pgtable.o.before
   7765     694      32    8491    212b arch/x86/mm/pgtable.o.after

So this case (fixing a warning and reducing image size) is a clear win.

Additionally drops an old work-around for GCC in the same code.

Link: https://github.com/KSPP/linux/issues/203
Link: https://github.com/KSPP/linux/issues/181


Signed-off-by: default avatarGustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: default avatarKees Cook <keescook@chromium.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/Yytb67xvrnctxnEe@work
      25226df4
    • Kees Cook's avatar
      mm: Make ksize() a reporting-only function · 38931d89
      Kees Cook authored 
      

With all "silently resizing" callers of ksize() refactored, remove the
logic in ksize() that would allow it to be used to effectively change
the size of an allocation (bypassing __alloc_size hints, etc). Users
wanting this feature need to either use kmalloc_size_roundup() before an
allocation, or use krealloc() directly.

For kfree_sensitive(), move the unpoisoning logic inline. Replace the
some of the partially open-coded ksize() in __do_krealloc with ksize()
now that it doesn't perform unpoisoning.

Adjust the KUnit tests to match the new ksize() behavior. Execution
tested with:

$ ./tools/testing/kunit/kunit.py run \
	--kconfig_add CONFIG_KASAN=y \
	--kconfig_add CONFIG_KASAN_GENERIC=y \
	--arch x86_64 kasan

Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: linux-mm@kvack.org
Cc: kasan-dev@googlegroups.com
Acked-by: default avatarVlastimil Babka <vbabka@suse.cz>
Acked-by: default avatarDavid Rientjes <rientjes@google.com>
Enhanced-by: default avatarAndrey Konovalov <andreyknvl@gmail.com>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
      38931d89
  11. Nov 23, 2022
    • Kees Cook's avatar
      kunit/fortify: Validate __alloc_size attribute results · 9124a264
      Kees Cook authored 
      

Validate the effect of the __alloc_size attribute on allocators. If the
compiler doesn't support __builtin_dynamic_object_size(), skip the
associated tests.

(For GCC, just remove the "--make_options" line below...)

$ ./tools/testing/kunit/kunit.py run --arch x86_64 \
        --kconfig_add CONFIG_FORTIFY_SOURCE=y \
	--make_options LLVM=1
        fortify
...
[15:16:30] ================== fortify (10 subtests) ===================
[15:16:30] [PASSED] known_sizes_test
[15:16:30] [PASSED] control_flow_split_test
[15:16:30] [PASSED] alloc_size_kmalloc_const_test
[15:16:30] [PASSED] alloc_size_kmalloc_dynamic_test
[15:16:30] [PASSED] alloc_size_vmalloc_const_test
[15:16:30] [PASSED] alloc_size_vmalloc_dynamic_test
[15:16:30] [PASSED] alloc_size_kvmalloc_const_test
[15:16:30] [PASSED] alloc_size_kvmalloc_dynamic_test
[15:16:30] [PASSED] alloc_size_devm_kmalloc_const_test
[15:16:30] [PASSED] alloc_size_devm_kmalloc_dynamic_test
[15:16:30] ===================== [PASSED] fortify =====================
[15:16:30] ============================================================
[15:16:30] Testing complete. Ran 10 tests: passed: 10
[15:16:31] Elapsed time: 8.348s total, 0.002s configuring, 6.923s building, 1.075s running

For earlier GCC prior to version 12, the dynamic tests will be skipped:

[15:18:59] ================== fortify (10 subtests) ===================
[15:18:59] [PASSED] known_sizes_test
[15:18:59] [PASSED] control_flow_split_test
[15:18:59] [PASSED] alloc_size_kmalloc_const_test
[15:18:59] [SKIPPED] alloc_size_kmalloc_dynamic_test
[15:18:59] [PASSED] alloc_size_vmalloc_const_test
[15:18:59] [SKIPPED] alloc_size_vmalloc_dynamic_test
[15:18:59] [PASSED] alloc_size_kvmalloc_const_test
[15:18:59] [SKIPPED] alloc_size_kvmalloc_dynamic_test
[15:18:59] [PASSED] alloc_size_devm_kmalloc_const_test
[15:18:59] [SKIPPED] alloc_size_devm_kmalloc_dynamic_test
[15:18:59] ===================== [PASSED] fortify =====================
[15:18:59] ============================================================
[15:18:59] Testing complete. Ran 10 tests: passed: 6, skipped: 4
[15:18:59] Elapsed time: 11.965s total, 0.002s configuring, 10.540s building, 1.068s running

Cc: David Gow <davidgow@google.com>
Cc: linux-hardening@vger.kernel.org
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
      9124a264
  13. Nov 19, 2022
    • Nathan Chancellor's avatar
      drm/sti: Fix return type of sti_{dvo,hda,hdmi}_connector_mode_valid() · 0ad811cc
      Nathan Chancellor authored 
      With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG),
indirect call targets are validated against the expected function
pointer prototype to make sure the call target is valid to help mitigate
ROP attacks. If they are not identical, there is a failure at run time,
which manifests as either a kernel panic or thread getting killed. A
proposed warning in clang aims to catch these at compile time, which
reveals:

  drivers/gpu/drm/sti/sti_hda.c:637:16: error: incompatible function pointer types initializing 'enum drm_mode_status (*)(struct drm_connector *, struct drm_display_mode *)' with an expression of type 'int (struct drm_connector *, struct drm_display_mode *)' [-Werror,-Wincompatible-function-pointer-types-strict]
          .mode_valid = sti_hda_connector_mode_valid,
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
  drivers/gpu/drm/sti/sti_dvo.c:376:16: error: incompatible function pointer types initializing 'enum drm_mode_status (*)(struct drm_connector *, struct drm_display_mode *)' with an expression of type 'int (struct drm_connector *, struct drm_display_mode *)' [-Werror,-Wincompatible-function-pointer-types-strict]
          .mode_valid = sti_dvo_connector_mode_valid,
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
  drivers/gpu/drm/sti/sti_hdmi.c:1035:16: error: incompatible function pointer types initializing 'enum drm_mode_status (*)(struct drm_connector *, struct drm_display_mode *)' with an expression of type 'int (struct drm_connector *, struct drm_display_mode *)' [-Werror,-Wincompatible-function-pointer-types-strict]
          .mode_valid = sti_hdmi_connector_mode_valid,
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

->mode_valid() in 'struct drm_connector_helper_funcs' expects a return
type of 'enum drm_mode_status', not 'int'. Adjust the return type of
sti_{dvo,hda,hdmi}_connector_mode_valid() to match the prototype's to
resolve the warning and CFI failure.

Link: https://github.com/ClangBuiltLinux/linux/issues/1750


Signed-off-by: default avatarNathan Chancellor <nathan@kernel.org>
Reviewed-by: default avatarKees Cook <keescook@chromium.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221102155623.3042869-1-nathan@kernel.org
      0ad811cc
    • Nathan Chancellor's avatar
      drm/fsl-dcu: Fix return type of fsl_dcu_drm_connector_mode_valid() · 96d845a6
      Nathan Chancellor authored 
      With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG),
indirect call targets are validated against the expected function
pointer prototype to make sure the call target is valid to help mitigate
ROP attacks. If they are not identical, there is a failure at run time,
which manifests as either a kernel panic or thread getting killed. A
proposed warning in clang aims to catch these at compile time, which
reveals:

  drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_rgb.c:74:16: error: incompatible function pointer types initializing 'enum drm_mode_status (*)(struct drm_connector *, struct drm_display_mode *)' with an expression of type 'int (struct drm_connector *, struct drm_display_mode *)' [-Werror,-Wincompatible-function-pointer-types-strict]
          .mode_valid = fsl_dcu_drm_connector_mode_valid,
                        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  1 error generated.

->mode_valid() in 'struct drm_connector_helper_funcs' expects a return
type of 'enum drm_mode_status', not 'int'. Adjust the return type of
fsl_dcu_drm_connector_mode_valid() to match the prototype's to resolve
the warning and CFI failure.

Link: https://github.com/ClangBuiltLinux/linux/issues/1750


Reported-by: default avatarSami Tolvanen <samitolvanen@google.com>
Signed-off-by: default avatarNathan Chancellor <nathan@kernel.org>
Reviewed-by: default avatarKees Cook <keescook@chromium.org>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221102154215.78059-1-nathan@kernel.org
      96d845a6
    • Kees Cook's avatar
      driver core: Add __alloc_size hint to devm allocators · 74c8e6bf
      Kees Cook authored 
      

Mark the devm_*alloc()-family of allocations with appropriate
__alloc_size()/__realloc_size() hints so the compiler can attempt to
reason about buffer lengths from allocations.

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Nishanth Menon <nm@ti.com>
Cc: Michael Kelley <mikelley@microsoft.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Won Chung <wonchung@google.com>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Reviewed-by: default avatarGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/20221029074734.gonna.276-kees@kernel.org
      74c8e6bf
  15. Nov 09, 2022
  17. Nov 03, 2022
    • Kees Cook's avatar
      overflow: Introduce overflows_type() and castable_to_type() · 4b21d25b
      Kees Cook authored 
      Implement a robust overflows_type() macro to test if a variable or
constant value would overflow another variable or type. This can be
used as a constant expression for static_assert() (which requires a
constant expression[1][2]) when used on constant values. This must be
constructed manually, since __builtin_add_overflow() does not produce
a constant expression[3].

Additionally adds castable_to_type(), similar to __same_type(), but for
checking if a constant value would overflow if cast to a given type.

Add unit tests for overflows_type(), __same_type(), and castable_to_type()
to the existing KUnit "overflow" test:

[16:03:33] ================== overflow (21 subtests) ==================
...
[16:03:33] [PASSED] overflows_type_test
[16:03:33] [PASSED] same_type_test
[16:03:33] [PASSED] castable_to_type_test
[16:03:33] ==================== [PASSED] overflow =====================
[16:03:33] ============================================================
[16:03:33] Testing complete. Ran 21 tests: passed: 21
[16:03:33] Elapsed time: 24.022s total, 0.002s configuring, 22.598s building, 0.767s running

[1] https://en.cppreference.com/w/c/language/_Static_assert
[2] C11 standard (ISO/IEC 9899:2011): 6.7.10 Static assertions
[3] https://gcc.gnu.org/onlinedocs/gcc/Integer-Overflow-Builtins.html


    6.56 Built-in Functions to Perform Arithmetic with Overflow Checking
    Built-in Function: bool __builtin_add_overflow (type1 a, type2 b,

Cc: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Tom Rix <trix@redhat.com>
Cc: Daniel Latypov <dlatypov@google.com>
Cc: Vitor Massaru Iha <vitor@massaru.org>
Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: linux-hardening@vger.kernel.org
Cc: llvm@lists.linux.dev
Co-developed-by: default avatarGwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: default avatarGwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Signed-off-by: default avatarKees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221024201125.1416422-1-gwan-gyeong.mun@intel.com
      4b21d25b
  19. Nov 02, 2022
  21. Oct 29, 2022