From 58e08e8d83ab03a1ca25d53420bd0b87f2dfe458 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Tue, 9 Jun 2020 09:55:11 -0400 Subject: gfs2: fix trans slab error when withdraw occurs inside log_flush Log flush operations (gfs2_log_flush()) can target a specific transaction. But if the function encounters errors (e.g. io errors) and withdraws, the transaction was only freed it if was queued to one of the ail lists. If the withdraw occurred before the transaction was queued to the ail1 list, function ail_drain never freed it. The result was: BUG gfs2_trans: Objects remaining in gfs2_trans on __kmem_cache_shutdown() This patch makes log_flush() add the targeted transaction to the ail1 list so that function ail_drain() will find and free it properly. Cc: stable@vger.kernel.org # v5.7+ Signed-off-by: Bob Peterson Signed-off-by: Andreas Gruenbacher --- fs/gfs2/log.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'fs/gfs2/log.c') diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index 3e4734431783..2b05415bbc13 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -1002,6 +1002,16 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl, u32 flags) out: if (gfs2_withdrawn(sdp)) { + /** + * If the tr_list is empty, we're withdrawing during a log + * flush that targets a transaction, but the transaction was + * never queued onto any of the ail lists. Here we add it to + * ail1 just so that ail_drain() will find and free it. + */ + spin_lock(&sdp->sd_ail_lock); + if (tr && list_empty(&tr->tr_list)) + list_add(&tr->tr_list, &sdp->sd_ail1_list); + spin_unlock(&sdp->sd_ail_lock); ail_drain(sdp); /* frees all transactions */ tr = NULL; } -- cgit From 7542486b89b2e321ffe0de82163b425d6a38bc72 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 17 Jun 2020 07:47:34 -0500 Subject: gfs2: eliminate GIF_ORDERED in favor of list_empty In several places, we used the GIF_ORDERED inode flag to determine if an inode was on the ordered writes list. However, since we always held the sd_ordered_lock spin_lock during the manipulation, we can just as easily check list_empty(&ip->i_ordered) instead. This allows us to keep more than one ordered writes list to make journal writing improvements. This patch eliminates GIF_ORDERED in favor of checking list_empty. Signed-off-by: Bob Peterson --- fs/gfs2/log.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'fs/gfs2/log.c') diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index 2b05415bbc13..a76e55bc28eb 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -613,6 +613,12 @@ static int ip_cmp(void *priv, struct list_head *a, struct list_head *b) return 0; } +static void __ordered_del_inode(struct gfs2_inode *ip) +{ + if (!list_empty(&ip->i_ordered)) + list_del_init(&ip->i_ordered); +} + static void gfs2_ordered_write(struct gfs2_sbd *sdp) { struct gfs2_inode *ip; @@ -623,8 +629,7 @@ static void gfs2_ordered_write(struct gfs2_sbd *sdp) while (!list_empty(&sdp->sd_log_ordered)) { ip = list_first_entry(&sdp->sd_log_ordered, struct gfs2_inode, i_ordered); if (ip->i_inode.i_mapping->nrpages == 0) { - test_and_clear_bit(GIF_ORDERED, &ip->i_flags); - list_del(&ip->i_ordered); + __ordered_del_inode(ip); continue; } list_move(&ip->i_ordered, &written); @@ -643,8 +648,7 @@ static void gfs2_ordered_wait(struct gfs2_sbd *sdp) spin_lock(&sdp->sd_ordered_lock); while (!list_empty(&sdp->sd_log_ordered)) { ip = list_first_entry(&sdp->sd_log_ordered, struct gfs2_inode, i_ordered); - list_del(&ip->i_ordered); - WARN_ON(!test_and_clear_bit(GIF_ORDERED, &ip->i_flags)); + __ordered_del_inode(ip); if (ip->i_inode.i_mapping->nrpages == 0) continue; spin_unlock(&sdp->sd_ordered_lock); @@ -659,8 +663,7 @@ void gfs2_ordered_del_inode(struct gfs2_inode *ip) struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); spin_lock(&sdp->sd_ordered_lock); - if (test_and_clear_bit(GIF_ORDERED, &ip->i_flags)) - list_del(&ip->i_ordered); + __ordered_del_inode(ip); spin_unlock(&sdp->sd_ordered_lock); } -- cgit From b57bc0fb2fe403b687142d12fba4ad085f8be6ee Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 22 Jul 2020 10:19:52 -0500 Subject: gfs2: Fix inaccurate comment The comment regarding journal flush thresholds is wrong. This patch fixes it. Signed-off-by: Bob Peterson Signed-off-by: Andreas Gruenbacher --- fs/gfs2/log.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/gfs2/log.c') diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index a76e55bc28eb..a58333e3980d 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -1092,7 +1092,7 @@ static void log_refund(struct gfs2_sbd *sdp, struct gfs2_trans *tr) * or the total number of used blocks (pinned blocks plus AIL blocks) * is greater than thresh2. * - * At mount time thresh1 is 1/3rd of journal size, thresh2 is 2/3rd of + * At mount time thresh1 is 2/5ths of journal size, thresh2 is 4/5ths of * journal size. * * Returns: errno -- cgit