kexec: bail out upon SIGKILL when allocating memory.
authorTetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Wed, 25 Sep 2019 23:47:33 +0000 (16:47 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 26 Sep 2019 00:51:40 +0000 (17:51 -0700)
syzbot found that a thread can stall for minutes inside kexec_load() after
that thread was killed by SIGKILL [1].  It turned out that the reproducer
was trying to allocate 2408MB of memory using kimage_alloc_page() from
kimage_load_normal_segment().  Let's check for SIGKILL before doing memory
allocation.

[1] https://syzkaller.appspot.com/bug?id=a0e3436829698d5824231251fad9d8e998f94f5e

Link: http://lkml.kernel.org/r/993c9185-d324-2640-d061-bed2dd18b1f7@I-love.SAKURA.ne.jp
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+8ab2d0f39fb79fe6ca40@syzkaller.appspotmail.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/kexec_core.c

index d587072..15d70a9 100644 (file)
@@ -300,6 +300,8 @@ static struct page *kimage_alloc_pages(gfp_t gfp_mask, unsigned int order)
 {
        struct page *pages;
 
+       if (fatal_signal_pending(current))
+               return NULL;
        pages = alloc_pages(gfp_mask & ~__GFP_ZERO, order);
        if (pages) {
                unsigned int count, i;