mm/writeback: fix possible divide-by-zero in wb_dirty_limits(), again
authorZach O'Keefe <zokeefe@google.com>
Thu, 18 Jan 2024 18:19:53 +0000 (10:19 -0800)
committerAndrew Morton <akpm@linux-foundation.org>
Fri, 26 Jan 2024 07:52:20 +0000 (23:52 -0800)
commit9319b647902cbd5cc884ac08a8a6d54ce111fc78
tree1a01168ba0d74fbe600de4cdf86530919d443699
parentbc29036e1da1cf66e5f8312649aeec2d51ea3d86
mm/writeback: fix possible divide-by-zero in wb_dirty_limits(), again

(struct dirty_throttle_control *)->thresh is an unsigned long, but is
passed as the u32 divisor argument to div_u64().  On architectures where
unsigned long is 64 bytes, the argument will be implicitly truncated.

Use div64_u64() instead of div_u64() so that the value used in the "is
this a safe division" check is the same as the divisor.

Also, remove redundant cast of the numerator to u64, as that should happen
implicitly.

This would be difficult to exploit in memcg domain, given the ratio-based
arithmetic domain_drity_limits() uses, but is much easier in global
writeback domain with a BDI_CAP_STRICTLIMIT-backing device, using e.g.
vm.dirty_bytes=(1<<32)*PAGE_SIZE so that dtc->thresh == (1<<32)

Link: https://lkml.kernel.org/r/20240118181954.1415197-1-zokeefe@google.com
Fixes: f6789593d5ce ("mm/page-writeback.c: fix divide by zero in bdi_dirty_limits()")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Maxim Patlasov <MPatlasov@parallels.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/page-writeback.c