diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2017-07-07 23:48:18 +0300 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2017-07-07 23:48:18 +0300 |
commit | b59eea554f57befa2aa3172fcb63e521bdd850dd (patch) | |
tree | 7b68bd0bd6666dfd5e377b0999d99d74c1396736 /tools/testing | |
parent | 6481352082e8c373a005387c4fc7258b197f0bc9 (diff) | |
download | linux-b59eea554f57befa2aa3172fcb63e521bdd850dd.tar.xz |
vfs: fix flock compat thinko
Michael Ellerman reported that commit 8c6657cb50cb ("Switch flock
copyin/copyout primitives to copy_{from,to}_user()") broke his
networking on a bunch of PPC machines (64-bit kernel, 32-bit userspace).
The reason is a brown-paper bug by that commit, which had the arguments
to "copy_flock_fields()" in the wrong order, breaking the compat
handling for file locking. Apparently very few people run 32-bit user
space on x86 any more, so the PPC people got the honor of noticing this
"feature".
Michael also sent a minimal diff that just changed the order of the
arguments in that macro.
This is not that minimal diff.
This not only changes the order of the arguments in the macro, it also
changes them to be pointers (to be consistent with all the other uses of
those pointers), and makes the functions that do all of this also have
the proper "const" attribution on the source pointers in order to make
issues like that (using the source as a destination) be really obvious.
Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'tools/testing')
0 files changed, 0 insertions, 0 deletions