https://bz.apache.org/bugzilla/show_bug.cgi?id=70299

--- Comment #1 from abernal <[email protected]> ---
Fixed: https://github.com/apache/httpd/pull/965

The read-side bug is exactly as described: confirmed modules/dav/lock/locks.c
still has the 4-byte int cast against the save side's full 8-byte memcpy.

Verified for real this time, not just statically: cross-compiled the exact two
expressions (the current 4-byte-cast read and the fixed full-width memcpy read)
for s390x (a real big-endian architecture) with zig cc, and ran the resulting
binary for real under qemu-s390x-static:

target endianness: big-endian
written key length                  : 20
current code reads (real execution) : 0   WRONG (bug reproduced for real)
fixed code reads   (real execution) : 20  OK (fix confirmed for real)

And natively on x86_64 (little-endian) to confirm no regression:

target endianness: little-endian
current code reads (real execution) : 20  OK
fixed code reads   (real execution) : 20  OK (fix confirmed for real)

This is real compiled machine code executed by QEMU's S/390 CPU emulation, not
a hand-computed simulation. Real physical big-endian hardware was not
available.

mod_dav_lock does not exist on trunk (removed, it was only used as drop-in
locking for pre-1.2.0 Apache Subversion), so this is 2.4.x-only.

-- 
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to