Kernel oops (bug) in fs/buffers.c:create_empty_buffers
Hi,
caught this one on kerneloops.org:
http://www.kerneloops.org/searchweek.php?search=create_empty_buffers
void create_empty_buffers(struct page *page,
unsigned long blocksize, unsigned long b_state)
{
struct buffer_head *bh, *head, *tail;
head = alloc_page_buffers(page, blocksize, 1);
bh = head;
do {
bh->b_state |= b_state;
tail = bh;
bh = bh->b_this_page;
} while (bh);turns out, alloc_page_buffers() can fail and return NULL (for AIO for example)... yet this code blindly dereferences the result, getting a predictable NULL pointer fault.
It's not directly clear what to do about... make this function return the failure to the caller?
Signature · 3 lines
If you want to reach me at my work email, use [email protected] For development, discussion and tips for power savings, visit http://www.lesswatts.org
Quoted 19 lines · On Sun, 6 Jul 2008 13:23:02 -0700 Arjan van de Ven <[email protected]> wrote:
Hi,
caught this one on kerneloops.org:
http://www.kerneloops.org/searchweek.php?search=create_empty_buffersvoid create_empty_buffers(struct page *page,
unsigned long blocksize, unsigned longb_state){ struct buffer_head *bh, *head, *tail; head = alloc_page_buffers(page, blocksize, 1); bh = head; do { bh->b_state |= b_state; tail = bh; bh = bh->b_this_page; } while (bh);
Quoted 6 lines
turns out,
alloc_page_buffers()can fail and return NULL (for AIO for example)... yet this code blindly dereferences the result, getting a predictable NULL pointer fault.It's not directly clear what to do about... make this function return the failure to the caller?
isofs has a habit of returning impossible block sizes and the
while ((offset -= size) >= 0) {consequently loops zero times and alloc_page_buffers() returns null.
Someone was having a look at it - maybe Jan?
I assume that the kernloops.org records link back to the original report somewhere but I can't find it?
Quoted 10 lines · On Sun, 6 Jul 2008 13:31:00 -0700 Andrew Morton <[email protected]>…
isofs has a habit of returning impossible block sizes and the
while ((offset -= size) >= 0) {consequently loops zero times and
alloc_page_buffers()returns null.Someone was having a look at it - maybe Jan?
I assume that the kernloops.org records link back to the original report somewhere but I can't find it?
http://www.kerneloops.org/search.php?search=create_empty_buffers
has two:
http://article.gmane.org/gmane.linux.kernel/696582
http://article.gmane.org/gmane.linux.ide/13658
Signature · 3 lines
If you want to reach me at my work email, use [email protected] For development, discussion and tips for power savings, visit http://www.lesswatts.org
Quoted 27 lines · On Sun, 6 Jul 2008 13:23:02 -0700 Arjan van de Ven <[email protected]> wrote:
Quoted 19 lines
Hi,
caught this one on kerneloops.org:
http://www.kerneloops.org/searchweek.php?search=create_empty_buffersvoid create_empty_buffers(struct page *page,
unsigned long blocksize, unsigned longb_state){ struct buffer_head *bh, *head, *tail; head = alloc_page_buffers(page, blocksize, 1); bh = head; do { bh->b_state |= b_state; tail = bh; bh = bh->b_this_page; } while (bh);Quoted 6 lines
turns out,
alloc_page_buffers()can fail and return NULL (for AIO for example)... yet this code blindly dereferences the result, getting a predictable NULL pointer fault.It's not directly clear what to do about... make this function return the failure to the caller?
Quoted 7 lines
isofs has a habit of returning impossible block sizes and the
while ((offset -= size) >= 0) {consequently loops zero times and
alloc_page_buffers()returns null.Someone was having a look at it - maybe Jan?
Yes, I was looking at that. Actually, I've pinpointed the cause of
the Oopses I've seen to a bug in ide-cd driver. It happily sets blocksize to whatever value the drive reported and some drives have a habit of reporting bogus values when unreadable medium is inserted.
Jens has actually cooked up a fix for this bug
(http://thread.gmane.org/gmane.linux.kernel/696582) and it fixes the problem at least for me.
Quoted 2 lines
I assume that the kernloops.org records link back to the original report somewhere but I can't find it?
Honza
Signature · 2 lines
Jan Kara <[email protected]> SuSE CR Labs