From d7961c7fa4d2e3c3f12be67e21ba8799b5a7238a Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Fri, 21 Dec 2012 00:15:51 -0500 Subject: jbd2: fix assertion failure in jbd2_journal_flush() The following race is possible between start_this_handle() and someone calling jbd2_journal_flush(). Process A Process B start_this_handle(). if (journal->j_barrier_count) # false if (!journal->j_running_transaction) { #true read_unlock(&journal->j_state_lock); jbd2_journal_lock_updates() jbd2_journal_flush() write_lock(&journal->j_state_lock); if (journal->j_running_transaction) { # false ... wait for committing trans ... write_unlock(&journal->j_state_lock); ... write_lock(&journal->j_state_lock); if (!journal->j_running_transaction) { # true jbd2_get_transaction(journal, new_transaction); write_unlock(&journal->j_state_lock); goto repeat; # eventually blocks on j_barrier_count > 0 ... J_ASSERT(!journal->j_running_transaction); # fails We fix the race by rechecking j_barrier_count after reacquiring j_state_lock in exclusive mode. Reported-by: yjwsignal@empal.com Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: stable@vger.kernel.org --- fs/jbd2/transaction.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/jbd2') diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c index deffd94..cd4485d 100644 --- a/fs/jbd2/transaction.c +++ b/fs/jbd2/transaction.c @@ -209,7 +209,8 @@ repeat: if (!new_transaction) goto alloc_transaction; write_lock(&journal->j_state_lock); - if (!journal->j_running_transaction) { + if (!journal->j_running_transaction && + !journal->j_barrier_count) { jbd2_get_transaction(journal, new_transaction); new_transaction = NULL; } -- cgit v1.1 From 53e872681fed6a43047e71bf927f77d06f467988 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 25 Dec 2012 13:29:52 -0500 Subject: ext4: fix deadlock in journal_unmap_buffer() We cannot wait for transaction commit in journal_unmap_buffer() because we hold page lock which ranks below transaction start. We solve the issue by bailing out of journal_unmap_buffer() and jbd2_journal_invalidatepage() with -EBUSY. Caller is then responsible for waiting for transaction commit to finish and try invalidation again. Since the issue can happen only for page stradding i_size, it is simple enough to manually call jbd2_journal_invalidatepage() for such page from ext4_setattr(), check the return value and wait if necessary. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" --- fs/jbd2/transaction.c | 27 ++++++++++++++------------- 1 file changed, 14 insertions(+), 13 deletions(-) (limited to 'fs/jbd2') diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c index cd4485d..ddc51a7 100644 --- a/fs/jbd2/transaction.c +++ b/fs/jbd2/transaction.c @@ -1840,7 +1840,6 @@ static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh, BUFFER_TRACE(bh, "entry"); -retry: /* * It is safe to proceed here without the j_list_lock because the * buffers cannot be stolen by try_to_free_buffers as long as we are @@ -1935,14 +1934,11 @@ retry: * for commit and try again. */ if (partial_page) { - tid_t tid = journal->j_committing_transaction->t_tid; - jbd2_journal_put_journal_head(jh); spin_unlock(&journal->j_list_lock); jbd_unlock_bh_state(bh); write_unlock(&journal->j_state_lock); - jbd2_log_wait_commit(journal, tid); - goto retry; + return -EBUSY; } /* * OK, buffer won't be reachable after truncate. We just set @@ -2003,21 +1999,23 @@ zap_buffer_unlocked: * @page: page to flush * @offset: length of page to invalidate. * - * Reap page buffers containing data after offset in page. - * + * Reap page buffers containing data after offset in page. Can return -EBUSY + * if buffers are part of the committing transaction and the page is straddling + * i_size. Caller then has to wait for current commit and try again. */ -void jbd2_journal_invalidatepage(journal_t *journal, - struct page *page, - unsigned long offset) +int jbd2_journal_invalidatepage(journal_t *journal, + struct page *page, + unsigned long offset) { struct buffer_head *head, *bh, *next; unsigned int curr_off = 0; int may_free = 1; + int ret = 0; if (!PageLocked(page)) BUG(); if (!page_has_buffers(page)) - return; + return 0; /* We will potentially be playing with lists other than just the * data lists (especially for journaled data mode), so be @@ -2031,9 +2029,11 @@ void jbd2_journal_invalidatepage(journal_t *journal, if (offset <= curr_off) { /* This block is wholly outside the truncation point */ lock_buffer(bh); - may_free &= journal_unmap_buffer(journal, bh, - offset > 0); + ret = journal_unmap_buffer(journal, bh, offset > 0); unlock_buffer(bh); + if (ret < 0) + return ret; + may_free &= ret; } curr_off = next_off; bh = next; @@ -2044,6 +2044,7 @@ void jbd2_journal_invalidatepage(journal_t *journal, if (may_free && try_to_free_buffers(page)) J_ASSERT(!page_has_buffers(page)); } + return 0; } /* -- cgit v1.1