[Devel] [PATCH RHEL8 COMMIT] ploop: Care about timeout on grow and snapshot

[Konstantin Khorenko]

The commit is pushed to "branch-rh8-4.18.0-80.1.2.vz8.3.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh8-4.18.0-80.1.2.vz8.3.2
------>
commit 30689b512ecc243e78a351d4e167a6920d48adb0
Author: Kirill Tkhai <ktkhai at virtuozzo.com>
Date:   Tue Mar 10 18:44:57 2020 +0300

    ploop: Care about timeout on grow and snapshot
    
    Signed-off-by: Kirill Tkhai <ktkhai at virtuozzo.com>
    
    =====================
    Patchset description:
    
    ploop: Make service operations care about timeout
    
    Recently in vz7 we met a situation, when fuse daemon lost a request.
    This patch set makes ploop service operations use timeout to prevent
    a hang in such the situations.
    
    Kirill Tkhai (7):
          ploop: Add timeout to ploop_inflight_bios_ref_switch()
          ploop: Care about timeout on grow and snapshot
          ploop: Care about timeout on tracking start
          ploop: Care about timeout on discard
          ploop: Care about timeout on process_notify_delta_merged()
          ploop: Care about timeout on process_merge_latest_snapshot_cmd()
          ploop: Care about timeout on process_update_delta_index()
---
 drivers/md/dm-ploop-cmd.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/drivers/md/dm-ploop-cmd.c b/drivers/md/dm-ploop-cmd.c
index 0d245ef505aa..75b4ade68097 100644
--- a/drivers/md/dm-ploop-cmd.c
+++ b/drivers/md/dm-ploop-cmd.c
@@ -227,7 +227,7 @@ static int ploop_grow_relocate_cluster(struct ploop *ploop,
 {
 	struct bio *bio = cmd->resize.bio;
 	unsigned int new_dst, cluster, dst_cluster;
-	bool is_locked;
+	bool is_locked, defer_bio_count = false;
 	int ret = 0;
 
 	dst_cluster = cmd->resize.dst_cluster;
@@ -252,7 +252,10 @@ static int ploop_grow_relocate_cluster(struct ploop *ploop,
 
 	/* Redirect bios to kwork and wait inflights, which may use @cluster */
 	force_defer_bio_count_inc(ploop);
-	ploop_inflight_bios_ref_switch(ploop, false);
+	defer_bio_count = true;
+	ret = ploop_inflight_bios_ref_switch(ploop, true);
+	if (ret < 0)
+		goto out;
 
 	/* Read full cluster sync */
 	ret = ploop_read_cluster_sync(ploop, bio, dst_cluster);
@@ -294,7 +297,7 @@ static int ploop_grow_relocate_cluster(struct ploop *ploop,
 	/* Zero new BAT entries on disk. */
 	ret = ploop_write_zero_cluster_sync(ploop, bio, dst_cluster);
 out:
-	if (cluster != UINT_MAX)
+	if (defer_bio_count)
 		force_defer_bio_count_dec(ploop);
 
 	return ret;
@@ -1009,9 +1012,12 @@ static int ploop_update_delta_index(struct ploop *ploop, unsigned int level,
 static void process_switch_top_delta(struct ploop *ploop, struct ploop_cmd *cmd)
 {
 	unsigned int i, size, bat_clusters, level = ploop->nr_deltas;
+	int ret;
 
 	force_defer_bio_count_inc(ploop);
-	ploop_inflight_bios_ref_switch(ploop, false);
+	ret = ploop_inflight_bios_ref_switch(ploop, true);
+	if (ret)
+		goto out;
 
 	/* If you add more two-stages-actions, you must cancel them here too */
 	cancel_discard_bios(ploop);
@@ -1036,9 +1042,10 @@ static void process_switch_top_delta(struct ploop *ploop, struct ploop_cmd *cmd)
 
 	ploop->nr_deltas++;
 	write_unlock_irq(&ploop->bat_rwlock);
+out:
 	force_defer_bio_count_dec(ploop);
 
-	cmd->retval = 0;
+	cmd->retval = ret;
 	complete(&cmd->comp); /* Last touch of cmd memory */
 }