fuse: break infinite loop in fuse_fill_write_pages()
authorRoman Gushchin <klamm@yandex-team.ru>
Mon, 12 Oct 2015 13:33:44 +0000 (16:33 +0300)
committerZefan Li <lizefan@huawei.com>
Wed, 26 Oct 2016 15:15:34 +0000 (23:15 +0800)
commit0dfa2ec52321e4d6fb0ef02a58e86b3fb734a284
tree6fa8634f622e50940e62ce094ed44839d0803a75
parentb411ae9afe4ddc0d7f31c023b09d50cd9559ffbd
fuse: break infinite loop in fuse_fill_write_pages()

commit 3ca8138f014a913f98e6ef40e939868e1e9ea876 upstream.

I got a report about unkillable task eating CPU. Further
investigation shows, that the problem is in the fuse_fill_write_pages()
function. If iov's first segment has zero length, we get an infinite
loop, because we never reach iov_iter_advance() call.

Fix this by calling iov_iter_advance() before repeating an attempt to
copy data from userspace.

A similar problem is described in 124d3b7041f ("fix writev regression:
pan hanging unkillable and un-straceable"). If zero-length segmend
is followed by segment with invalid address,
iov_iter_fault_in_readable() checks only first segment (zero-length),
iov_iter_copy_from_user_atomic() skips it, fails at second and
returns zero -> goto again without skipping zero-length segment.

Patch calls iov_iter_advance() before goto again: we'll skip zero-length
segment at second iteraction and iov_iter_fault_in_readable() will detect
invalid address.

Special thanks to Konstantin Khlebnikov, who helped a lot with the commit
description.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Maxim Patlasov <mpatlasov@parallels.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Roman Gushchin <klamm@yandex-team.ru>
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Fixes: ea9b9907b82a ("fuse: implement perform_write")
Signed-off-by: Zefan Li <lizefan@huawei.com>
fs/fuse/file.c