diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2025-09-29 16:31:35 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2025-09-29 16:31:35 -0700 |
commit | 8c1ed3021883f5e49f74dfb41ce0a24e5d07fdf0 (patch) | |
tree | f27d49b017d471fd9931802eed520cc978211bd0 /lib/clz_ctz.c | |
parent | 1896ce8eb6c61824f6c1125d69d8fda1f44a22f8 (diff) | |
parent | 95719dfa323709c06ec34cc96e73e0788e19934f (diff) |
Merge tag 'ffs-const-v6.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux
Pull ffs const-attribute cleanups from Kees Cook:
"While working on various hardening refactoring a while back we
encountered inconsistencies in the application of __attribute_const__
on the ffs() family of functions.
This series fixes this across all archs and adds KUnit tests.
Notably, this found a theoretical underflow in PCI (also fixed here)
and uncovered an inefficiency in ARC (fixed in the ARC arch PR). I
kept the series separate from the general hardening PR since it is a
stand-alone "topic".
- PCI: Fix theoretical underflow in use of ffs().
- Universally apply __attribute_const__ to all architecture's
ffs()-family of functions.
- Add KUnit tests for ffs() behavior and const-ness"
* tag 'ffs-const-v6.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
KUnit: ffs: Validate all the __attribute_const__ annotations
sparc: Add __attribute_const__ to ffs()-family implementations
xtensa: Add __attribute_const__ to ffs()-family implementations
s390: Add __attribute_const__ to ffs()-family implementations
parisc: Add __attribute_const__ to ffs()-family implementations
mips: Add __attribute_const__ to ffs()-family implementations
m68k: Add __attribute_const__ to ffs()-family implementations
openrisc: Add __attribute_const__ to ffs()-family implementations
riscv: Add __attribute_const__ to ffs()-family implementations
hexagon: Add __attribute_const__ to ffs()-family implementations
alpha: Add __attribute_const__ to ffs()-family implementations
sh: Add __attribute_const__ to ffs()-family implementations
powerpc: Add __attribute_const__ to ffs()-family implementations
x86: Add __attribute_const__ to ffs()-family implementations
csky: Add __attribute_const__ to ffs()-family implementations
bitops: Add __attribute_const__ to generic ffs()-family implementations
KUnit: Introduce ffs()-family tests
PCI: Test for bit underflow in pcie_set_readrq()
Diffstat (limited to 'lib/clz_ctz.c')
-rw-r--r-- | lib/clz_ctz.c | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/lib/clz_ctz.c b/lib/clz_ctz.c index fb8c0c5c2bd2..8778ec44bf63 100644 --- a/lib/clz_ctz.c +++ b/lib/clz_ctz.c @@ -15,28 +15,28 @@ #include <linux/kernel.h> int __weak __ctzsi2(int val); -int __weak __ctzsi2(int val) +int __weak __attribute_const__ __ctzsi2(int val) { return __ffs(val); } EXPORT_SYMBOL(__ctzsi2); int __weak __clzsi2(int val); -int __weak __clzsi2(int val) +int __weak __attribute_const__ __clzsi2(int val) { return 32 - fls(val); } EXPORT_SYMBOL(__clzsi2); int __weak __clzdi2(u64 val); -int __weak __clzdi2(u64 val) +int __weak __attribute_const__ __clzdi2(u64 val) { return 64 - fls64(val); } EXPORT_SYMBOL(__clzdi2); int __weak __ctzdi2(u64 val); -int __weak __ctzdi2(u64 val) +int __weak __attribute_const__ __ctzdi2(u64 val) { return __ffs64(val); } |