lkml.org 
[lkml]   [2005]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Subject[RFC] ext3/jbd race: releasing in-use journal_heads
From
Date
Hi all,

For the past few months there has been a slow but steady trickle of
reports of oopses in kjournald. Recently I got a couple of reports that
were repeatable enough to rerun with extra debugging code.

It turns out that we're releasing a journal_head while it is still
linked onto the transaction's t_locked_list. The exact location is in
journal_unmap_buffer(). On several exit paths, that does:

spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
journal_put_journal_head(jh);

releasing the jh *after* dropping the j_list_lock and j_state_lock.

kjournald can then be doing journal_commit_transaction():

spin_lock(&journal->j_list_lock);
...
if (buffer_locked(bh)) {
BUFFER_TRACE(bh, "locked");
if (!inverted_lock(journal, bh))
goto write_out_data;
__journal_unfile_buffer(jh);
__journal_file_buffer(jh, commit_transaction,
BJ_Locked);
jbd_unlock_bh_state(bh);

The problem happens if journal_unmap_buffer()'s own put_journal_head()
manages to get in between kjournald's *unfile_buffer and the following
*file_buffer. Because journal_unmap_buffer() has dropped its bh_state
lock by this point, there's nothing to prevent this, leading to a
variety of unpleasant situations. In particular, the jh is unfiled at
this point, so there's nothing to stop the put_journal_head() from
freeing the memory we're just about to link onto the BJ_Locked list.

I _think_ that the attached patch deals with this, but I'm still
awaiting further testing to be sure. I thought I might as well get some
other ext3 eyes on it while I wait for that -- I'll let you know as soon
as I hear back from the other testing.

The patch works by making sure that the various exits from
journal_unmap_buffer() always call journal_put_journal_head() *before*
unlocking the j_list_lock. This is correct according to the documented
lock ranking, and it also matches the order in the existing main exit
path at the end of the function.

Cheers,
Stephen

--- linux-2.6-ext3/fs/jbd/transaction.c.=K0000=.orig
+++ linux-2.6-ext3/fs/jbd/transaction.c
@@ -1775,10 +1775,10 @@ static int journal_unmap_buffer(journal_
JBUFFER_TRACE(jh, "checkpointed: add to BJ_Forget");
ret = __dispose_buffer(jh,
journal->j_running_transaction);
+ journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
- journal_put_journal_head(jh);
return ret;
} else {
/* There is no currently-running transaction. So the
@@ -1789,10 +1789,10 @@ static int journal_unmap_buffer(journal_
JBUFFER_TRACE(jh, "give to committing trans");
ret = __dispose_buffer(jh,
journal->j_committing_transaction);
+ journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
- journal_put_journal_head(jh);
return ret;
} else {
/* The orphan record's transaction has
@@ -1813,10 +1813,10 @@ static int journal_unmap_buffer(journal_
journal->j_running_transaction);
jh->b_next_transaction = NULL;
}
+ journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
- journal_put_journal_head(jh);
return 0;
} else {
/* Good, the buffer belongs to the running transaction.
\
 
 \ /
  Last update: 2005-03-22 14:10    [W:0.132 / U:0.348 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site