random: restore O_NONBLOCK support
authorJason A. Donenfeld <Jason@zx2c4.com>
Thu, 8 Sep 2022 14:14:00 +0000 (16:14 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 15 Oct 2022 05:59:01 +0000 (07:59 +0200)
commitbe53fa6cf667b9aac5ad64ba209ac9e14c610dee
tree8a292792e29c6efc99fdfcdbd0eb6c451db59c47
parent2f96da3fd18fd7ba21768a3e5e0d2fdd4fde0b17
random: restore O_NONBLOCK support

commit cd4f24ae9404fd31fc461066e57889be3b68641b upstream.

Prior to 5.6, when /dev/random was opened with O_NONBLOCK, it would
return -EAGAIN if there was no entropy. When the pools were unified in
5.6, this was lost. The post 5.6 behavior of blocking until the pool is
initialized, and ignoring O_NONBLOCK in the process, went unnoticed,
with no reports about the regression received for two and a half years.
However, eventually this indeed did break somebody's userspace.

So we restore the old behavior, by returning -EAGAIN if the pool is not
initialized. Unlike the old /dev/random, this can only occur during
early boot, after which it never blocks again.

In order to make this O_NONBLOCK behavior consistent with other
expectations, also respect users reading with preadv2(RWF_NOWAIT) and
similar.

Fixes: 30c08efec888 ("random: make /dev/random be almost like /dev/urandom")
Reported-by: Guozihua <guozihua@huawei.com>
Reported-by: Zhongguohua <zhongguohua1@huawei.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Andrew Lutomirski <luto@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/char/mem.c
drivers/char/random.c