On Thu, 14 May 2015, Jan Kara wrote:
> Date: Thu, 14 May 2015 12:35:00 +0200
> From: Jan Kara <jack@xxxxxxx>
> To: Lukáš Czerner <lczerner@xxxxxxxxxx>
> Cc: Jan Kara <jack@xxxxxxx>, linux-ext4@xxxxxxxxxxxxxxx,
> Theodore Ts'o <tytso@xxxxxxx>, stable@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v2] ext4: fix NULL pointer dereference when journal
> restart fails
>
> On Thu 14-05-15 11:37:46, Lukáš Czerner wrote:
> > On Thu, 14 May 2015, Jan Kara wrote:
> >
> > > Date: Thu, 14 May 2015 11:24:57 +0200
> > > From: Jan Kara <jack@xxxxxxx>
> > > To: Lukas Czerner <lczerner@xxxxxxxxxx>
> > > Cc: linux-ext4@xxxxxxxxxxxxxxx, Theodore Ts'o <tytso@xxxxxxx>,
> > > Jan Kara <jack@xxxxxxx>, stable@xxxxxxxxxxxxxxx
> > > Subject: Re: [PATCH v2] ext4: fix NULL pointer dereference when journal
> > > restart fails
> > >
> > > On Thu 14-05-15 11:03:06, Lukas Czerner wrote:
> > > > Currently when journal restart fails, we'll have the h_transaction of
> > > > the handle set to NULL to indicate that the handle has been effectively
> > > > aborted. We handle this situation quietly in the jbd2_journal_stop() and just
> > > > free the handle and exit because everything else has been done before we
> > > > attempted (and failed) to restart the journal.
> > > >
> > > > Unfortunately there are a number of problems with that approach
> > > > introduced with commit
> > > >
> > > > 41a5b913197c "jbd2: invalidate handle if jbd2_journal_restart()
> > > > fails"
> > > >
> > > > First of all in ext4 jbd2_journal_stop() will be called through
> > > > __ext4_journal_stop() where we would try to get a hold of the superblock
> > > > by dereferencing h_transaction which in this case would lead to NULL
> > > > pointer dereference and crash.
> > > >
> > > > In addition we're going to free the handle regardless of the refcount
> > > > which is bad as well, because others up the call chain will still
> > > > reference the handle so we might potentially reference already freed
> > > > memory.
> > > >
> > > > Moreover it's expected that we'll get aborted handle as well as detached
> > > > handle in some of the journalling function as the error propagates up
> > > > the stack, so it's unnecessary to call WARN_ON every time we get
> > > > detached handle.
> > > >
> > > > And finally we might leak some memory by forgetting to free reserved
> > > > handle in jbd2_journal_stop() in the case where handle was detached from
> > > > the transaction (h_transaction is NULL).
> > > >
> > > > Fix the NULL pointer dereference in __ext4_journal_stop() by just
> > > > calling jbd2_journal_stop() quietly as suggested by Jan Kara. Also fix
> > > > the potential memory leak in jbd2_journal_stop() and use proper
> > > > handle refcounting before we attempt to free it to avoid use-after-free
> > > > issues.
> > > >
> > > > And finally remove all WARN_ON(!transaction) from the code so that we do
> > > > not get random traces when something goes wrong because when journal
> > > > restart fails we will get to some of those functions.
> > > >
> > > > Signed-off-by: Lukas Czerner <lczerner@xxxxxxxxxx>
> > > > Cc: stable@xxxxxxxxxxxxxxx
> > > > ---
> > > > v2: As Jan Kara pointed out setting h_transaction to NULL is actually
> > > > desirable because the handle was detached from that transaction.
> > >
> > > The patch looks good, you can add:
> > >
> > > Reviewed-by: Jan Kara <jack@xxxxxxx>
> > >
> > > Just one small nit below:
> > >
> > > ...
> > > > @@ -1530,8 +1524,22 @@ int jbd2_journal_stop(handle_t *handle)
> > > > tid_t tid;
> > > > pid_t pid;
> > > >
> > > > - if (!transaction)
> > > > - goto free_and_exit;
> > > > + if (!transaction) {
> > > > + /*
> > > > + * Handle is already detached from the transaction so
> > > > + * there is nothing to do other than decrease a refcount,
> > > > + * or free the handle if refcount drops to zero
> > > > + */
> > > > + if (--handle->h_ref > 0) {
> > > > + jbd_debug(4, "h_ref %d -> %d\n", handle->h_ref + 1,
> > > > + handle->h_ref);
> > > > + return err;
> > > > + } else {
> > > > + if (handle->h_rsv_handle)
> > > > + jbd2_free_handle(handle->h_rsv_handle);
> > > > + goto free_and_exit;
> > >
> > > It would we nicer if you just moved free_and_exit label before freeing of
> > > the reserved handle instead of duplicating the code here. Also it is more
> > > future proof if we add more places jumping to the label...
> >
> > The problem is that this a special case when we have detached handle
> > so we only need to free the structure. In normal case we're calling
> > jbd2_journal_free_reserved() instead and I did not wanted to add
> > (!transaction) condition to multiple places.
> >
> > jbd2_journal_free_reserved() is different in that it does
> > sub_reserved_credits() for the journal, but in this case it has
> > already been done in jbd2__journal_restart().
> Hum, actually you must call jbd2_journal_free_reserved() so that you
> don't leak credits assigned for the reserved handle, I missed that you call
> only jbd2_free_handle(). And note that although the current handle is
> detached, its reserved handle is still properly attached to the journal
> (not any particular transaction) so it is safe to call
> jbd2_journal_free_reserved().
Sure, but this is already done in jbd2__journal_restart()
if (handle->h_rsv_handle) {
sub_reserved_credits(journal, handle->h_rsv_handle->h_buffer_credits);
}
and AFAICT if start_this_handle() fails then we would not have
increased the reserved credits, or am I missing something ?
-Lukas
>
> Honza
>
