]> git.ipfire.org Git - thirdparty/qemu.git/commit
osdep: Fix ROUND_UP(64-bit, 32-bit)
authorEric Blake <eblake@redhat.com>
Thu, 14 Sep 2017 13:49:23 +0000 (08:49 -0500)
committerMichael Roth <mdroth@linux.vnet.ibm.com>
Thu, 28 Sep 2017 21:37:36 +0000 (16:37 -0500)
commit33a599667a9e70588483a31286dfff8cfc27d513
treefb0de0f241295811cf81b721d44d469a737a9f04
parenta432f419abce2d63907978c3d48dfc9e3f740e49
osdep: Fix ROUND_UP(64-bit, 32-bit)

When using bit-wise operations that exploit the power-of-two
nature of the second argument of ROUND_UP(), we still need to
ensure that the mask is as wide as the first argument (done
by using a ternary to force proper arithmetic promotion).
Unpatched, ROUND_UP(2ULL*1024*1024*1024*1024, 512U) produces 0,
instead of the intended 2TiB, because negation of an unsigned
32-bit quantity followed by widening to 64-bits does not
sign-extend the mask.

Broken since its introduction in commit 292c8e50 (v1.5.0).
Callers that passed the same width type to both macro parameters,
or that had other code to ensure the first parameter's maximum
runtime value did not exceed the second parameter's width, are
unaffected, but I did not audit to see which (if any) existing
clients of the macro could trigger incorrect behavior (I found
the bug while adding a new use of the macro).

While preparing the patch, checkpatch complained about poor
spacing, so I also fixed that here and in the nearby DIV_ROUND_UP.

CC: qemu-trivial@nongnu.org
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
(cherry picked from commit 2098b073f398cd628c09c5a78537a6854e85830d)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
include/qemu/osdep.h