Re: [PATCH] Fix DIO EIO error caused by race betweenjbd_commit_transaction() and journal_try_to_drop_buffers()

From: Badari Pulavarty
Date: Fri May 16 2008 - 13:17:33 EST



On Fri, 2008-05-16 at 10:11 -0700, Mingming Cao wrote:
> On Fri, 2008-05-16 at 11:01 -0400, Josef Bacik wrote:
>
> >
> > Got a couple of whitespace problems above it looks like. Thanks,
> >
>
> Thanks for catching this, below is updated patch, fixed the whitespace
> and comments.
>
>
> ---------------------------------------------------
> JBD: fix journal_try_to_free_buffers race with
> journal_commit_transaction
>
> From: Mingming Cao <cmm@xxxxxxxxxx>
>
> This patch fixed a few races between direct IO and kjournld commit
> transaction.
> An unexpected EIO error gets returned to direct IO caller when it failed
> to
> free those data buffers. This could be reproduced easily with parallel
> direct write and buffered write to the same file
>
> More specificly, those races could cause journal_try_to_free_buffers()
> fail to free the data buffers, when jbd is committing the transaction
> that has
> those data buffers on its t_syncdata_list or t_locked_list.
> journal_commit_transaction() still holds the reference to those buffers
> before data reach to disk and buffers are removed from the
> t_syncdata_list of t_locked_list. This prevent the concurrent
> journal_try_to_free_buffers() to free those buffers at the same time,
> but cause
> EIO error returns back to direct IO.
>
> With this patch, in case of direct IO and when try_to_free_buffers()
> failed,
> let's waiting for journal_commit_transaction() to finish
> flushing the current committing transaction's data buffers to disk,
> then try to free those buffers again.
>
> Signed-off-by: Mingming Cao <cmm@xxxxxxxxxx>
> Reviewed-by: Badari Pulavarty <pbadari@xxxxxxxxxx>
> ---
> fs/jbd/commit.c | 1 +
> fs/jbd/journal.c | 1 +
> fs/jbd/transaction.c | 46
> ++++++++++++++++++++++++++++++++++++++++++++++
> include/linux/jbd.h | 3 +++
> 4 files changed, 51 insertions(+)
>
> Index: linux-2.6.26-rc1/include/linux/jbd.h
> ===================================================================
> --- linux-2.6.26-rc1.orig/include/linux/jbd.h 2008-05-14
> 16:36:41.000000000 -0700
> +++ linux-2.6.26-rc1/include/linux/jbd.h 2008-05-15 14:12:10.000000000
> -0700
> @@ -667,6 +667,9 @@ struct journal_s
> */
> wait_queue_head_t j_wait_transaction_locked;
>
> + /* Wait queu for waiting for data buffers to flushed to disk*/
> + wait_queue_head_t j_wait_data_flushed;
> +
> /* Wait queue for waiting for checkpointing to complete */
> wait_queue_head_t j_wait_logspace;
>
> Index: linux-2.6.26-rc1/fs/jbd/commit.c
> ===================================================================
> --- linux-2.6.26-rc1.orig/fs/jbd/commit.c 2008-05-03 11:59:44.000000000
> -0700
> +++ linux-2.6.26-rc1/fs/jbd/commit.c 2008-05-15 14:12:46.000000000 -0700
> @@ -462,6 +462,7 @@ void journal_commit_transaction(journal_
> * clean by now, so check that it is in fact empty.
> */
> J_ASSERT (commit_transaction->t_sync_datalist == NULL);
> + wake_up(&journal->j_wait_data_flushed)
>
> jbd_debug (3, "JBD: commit phase 3\n");
>
> Index: linux-2.6.26-rc1/fs/jbd/journal.c
> ===================================================================
> --- linux-2.6.26-rc1.orig/fs/jbd/journal.c 2008-05-14 16:36:41.000000000
> -0700
> +++ linux-2.6.26-rc1/fs/jbd/journal.c 2008-05-15 14:13:02.000000000
> -0700
> @@ -660,6 +660,7 @@ static journal_t * journal_init_common (
> goto fail;
>
> init_waitqueue_head(&journal->j_wait_transaction_locked);
> + init_waitqueue_head(&journal->j_wait_data_flushed);
> init_waitqueue_head(&journal->j_wait_logspace);
> init_waitqueue_head(&journal->j_wait_done_commit);
> init_waitqueue_head(&journal->j_wait_checkpoint);
> Index: linux-2.6.26-rc1/fs/jbd/transaction.c
> ===================================================================
> --- linux-2.6.26-rc1.orig/fs/jbd/transaction.c 2008-05-03
> 11:59:44.000000000 -0700
> +++ linux-2.6.26-rc1/fs/jbd/transaction.c 2008-05-16 09:27:21.000000000
> -0700
> @@ -1648,12 +1648,49 @@ out:
> return;
> }
>
> +/*
> + * journal_try_to_free_buffers() could race with
> journal_commit_transaction()
> + * The later might still hold the reference count to the buffers when
> inspecting
> + * them on t_syncdata_list or t_locked_list.
> + *
> + * Journal_try_to_free_buffers() will call this function to
> + * wait for the current transaction finishing syncing data buffers,
> before
> + * try to free that buffer.
> + *
> + * Called with journal->j_state_lock hold.

Fix wrapping lines ?

> + */
> +static void journal_wait_for_transaction_sync_data(journal_t *journal)
> +{
> + transaction_t *transaction = NULL;
> +
> + transaction = journal->j_committing_transaction;
> +
> + if (!transaction)
> + return;
> +
> + /*
> + * If the current transaction is flushing and waiting for data buffers
> + * (t_state is T_FLUSH), wait for the j_wait_data_flushed event
> + */
> + if (transaction->t_state == T_FLUSH) {
> + DEFINE_WAIT(wait);
> +
> + prepare_to_wait(&journal->j_wait_data_flushed,
> + &wait, TASK_UNINTERRUPTIBLE);
> + spin_unlock(&journal->j_state_lock);
> + schedule();
> + finish_wait(&journal->j_wait_data_flushed, &wait);
> + spin_lock(&journal->j_state_lock);
> + }
> + return;
> +}
>
> /**
> * int journal_try_to_free_buffers() - try to free page buffers.
> * @journal: journal for operation
> * @page: to try and free
> - * @unused_gfp_mask: unused
> + * @gfp_mask: unused for allocation purpose. Here is used
> + * as a flag to tell if direct IO is attemping to free buffers.
> *
> *
> * For all the buffers on this page,
> @@ -1682,13 +1719,16 @@ out:
> * journal_try_to_free_buffer() is changing its state. But that
> * cannot happen because we never reallocate freed data as metadata
> * while the data is part of a transaction. Yes?
> + *
> + * Return 0 on failure, 1 on success
> */
> int journal_try_to_free_buffers(journal_t *journal,
> - struct page *page, gfp_t unused_gfp_mask)
> + struct page *page, gfp_t gfp_mask)
> {
> struct buffer_head *head;
> struct buffer_head *bh;
> int ret = 0;
> + int dio = gfp_mask & __GFP_REPEAT;
>
> J_ASSERT(PageLocked(page));
>
> @@ -1713,7 +1753,31 @@ int journal_try_to_free_buffers(journal_
> if (buffer_jbd(bh))
> goto busy;
> } while ((bh = bh->b_this_page) != head);
> +
> ret = try_to_free_buffers(page);
> +
> + /*
> + * In the case of concurrent direct IO and buffered IO,
> + * There are a number of places where we
> + * could race with journal_commit_transaction(), the later still
> + * helds the reference to the buffers to free while processing them.
> + * try_to_free_buffers() failed to free those buffers,
> + * resulting in an unexpected EIO error
> + * returns back to the generic_file_direct_IO()
> + *
> + * So let's wait for the current transaction finished flush
> + * dirty data buffers before we try to free those buffers
> + * again. This wait is needed by direct IO code path only,
> + * gfp_mask __GFP_REPEAT is passed from the direct IO code
> + * path to flag if we need to wait and retry free buffers.
> + */
> + if (ret == 0 && dio) {

drop "dio" variable and compare here, like
if (ret == 0 && (gfp_mask & __GFP_REPEAT)

> + spin_lock(&journal->j_state_lock);
> + journal_wait_for_transaction_sync_data(journal);
> + ret = try_to_free_buffers(page);
> + spin_unlock(&journal->j_state_lock);
> + }

Thanks,
Badari

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/