Skip to contentSkip to conversation

Assorted blkcg fixes

PATCH In review Tejun Heo 3 participants 10 messages Jun 13, 2019 – Jun 19, 2019

Subsystems (4)
lore
Share & tools
lore

Tejun Heo

author maintainer
#0 lore raw

PATCH 1/1 Assorted blkcg fixes

Hello,

This patchset contains the following blkcg fixes.

0001-blk-iolatency-clear-use_delay-when-io.latency-is-set.patch 0002-blkcg-update-blkcg_print_stat-to-handle-larger-outpu.patch 0003-blkcg-perpcu_ref-init-exit-should-be-done-from-blkg_.patch 0004-blkcg-blkcg_activate_policy-should-initialize-ancest.patch 0005-blkcg-writeback-dead-memcgs-shouldn-t-contribute-to-.patch

Please refer to each patch's description for details. Patchset is also available in the following git branch.

git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup.git review-blkcg-fixes

Thanks. diffstat follows.

 block/blk-cgroup.c    |   24 ++++++++++++------------
 block/blk-iolatency.c |    4 +++-
 fs/fs-writeback.c     |    8 +++++++-
 3 files changed, 22 insertions(+), 14 deletions(-)

--
tejun

author maintainer
#1 lore raw

PATCH 1/5 blk-iolatency: clear use_delay when io.latency is set to zero

If use_delay was non-zero when the latency target of a cgroup was set to zero, it will stay stuck until io.latency is enabled on the cgroup again. This keeps readahead disabled for the cgroup impacting performance negatively.

Cc Josef Bacik <[email protected]>
Fixes d70675121546 ("block: introduce blk-iolatency io controller")
Cc [email protected] # v4.19+
---
 block/blk-iolatency.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
block/blk-iolatency.c +3 -1
index d22e61bced86..17896bb3aaf2 100644
@@ -778,8 +778,10 @@ static int iolatency_set_min_lat_nsec(struct blkcg_gq *blkg, u64 val)
 
 	if (!oldval && val)
 		return 1;
-	if (oldval && !val)
+	if (oldval && !val) {
+		blkcg_clear_delay(blkg);
 		return -1;
+	}
 	return 0;
 }
 
Signature · 1 line
2.17.1
author maintainer
#2 lore raw

PATCH 2/5 blkcg: update blkcg_print_stat() to handle larger outputs

In reply to Tejun Heo

Depending on the number of devices, blkcg stats can go over the default seqfile buf size. seqfile normally retries with a larger buffer but since the ->pd_stat() addition, blkcg_print_stat() doesn't tell seqfile that overflow has happened and the output gets printed truncated. Fix it by calling seq_commit() w/ -1 on possible overflows.

Fixes 903d23f0a354 ("blk-cgroup: allow controllers to output their own stats")
Cc [email protected] # v4.19+
Cc Josef Bacik <[email protected]>
---
 block/blk-cgroup.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
block/blk-cgroup.c +6 -2
index 1f7127b03490..e4715b35d42c 100644
@@ -1006,8 +1006,12 @@ static int blkcg_print_stat(struct seq_file *sf, void *v)
 		}
 next:
 		if (has_stats) {
-			off += scnprintf(buf+off, size-off, "\n");
-			seq_commit(sf, off);
+			if (off < size - 1) {
+				off += scnprintf(buf+off, size-off, "\n");
+				seq_commit(sf, off);
+			} else {
+				seq_commit(sf, -1);
+			}
 		}
 	}
 
Signature · 1 line
2.17.1
author maintainer
#3 lore raw

PATCH 3/5 blkcg: perpcu_ref init/exit should be done from blkg_alloc/free()

In reply to Tejun Heo

blkg alloc is performed as a separate step from the rest of blkg creation so that GFP_KERNEL allocations can be used when creating blkgs from configuration file writes because otherwise user actions may fail due to failures of opportunistic GFP_NOWAIT allocations.

While making blkgs use percpu_ref, 7fcf2b033b84 ("blkcg: change blkg reference counting to use percpu_ref") incorrectly added unconditional opportunistic percpu_ref_init() to blkg_create() breaking this guarantee.

This patch moves percpu_ref_init() to blkg_alloc() so makes it use @gfp_mask that blkg_alloc() is called with. Also, percpu_ref_exit() is moved to blkg_free() for consistency.

Fixes 7fcf2b033b84 ("blkcg: change blkg reference counting to use percpu_ref")
Cc Dennis Zhou <[email protected]>
---
 block/blk-cgroup.c | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)
block/blk-cgroup.c +4 -9
index e4715b35d42c..04d286934c5e 100644
@@ -79,6 +79,7 @@ static void blkg_free(struct blkcg_gq *blkg)
 
 	blkg_rwstat_exit(&blkg->stat_ios);
 	blkg_rwstat_exit(&blkg->stat_bytes);
+	percpu_ref_exit(&blkg->refcnt);
 	kfree(blkg);
 }
 
@@ -86,8 +87,6 @@ static void __blkg_release(struct rcu_head *rcu)
 {
 	struct blkcg_gq *blkg = container_of(rcu, struct blkcg_gq, rcu_head);
 
-	percpu_ref_exit(&blkg->refcnt);
-
 	/* release the blkcg and parent blkg refs this blkg has been holding */
 	css_put(&blkg->blkcg->css);
 	if (blkg->parent)
@@ -132,6 +131,9 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
 	if (!blkg)
 		return NULL;
 
+	if (percpu_ref_init(&blkg->refcnt, blkg_release, 0, gfp_mask))
+		goto err_free;
+
 	if (blkg_rwstat_init(&blkg->stat_bytes, gfp_mask) ||
 	    blkg_rwstat_init(&blkg->stat_ios, gfp_mask))
 		goto err_free;
@@ -244,11 +246,6 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
 		blkg_get(blkg->parent);
 	}
 
-	ret = percpu_ref_init(&blkg->refcnt, blkg_release, 0,
-			      GFP_NOWAIT | __GFP_NOWARN);
-	if (ret)
-		goto err_cancel_ref;
-
 	/* invoke per-policy init */
 	for (i = 0; i < BLKCG_MAX_POLS; i++) {
 		struct blkcg_policy *pol = blkcg_policy[i];
@@ -281,8 +278,6 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg,
 	blkg_put(blkg);
 	return ERR_PTR(ret);
 
-err_cancel_ref:
-	percpu_ref_exit(&blkg->refcnt);
 err_put_congested:
 	wb_congested_put(wb_congested);
 err_put_css:
Signature · 1 line
2.17.1
author maintainer
#4 lore raw

PATCH 4/5 blkcg: blkcg_activate_policy() should initialize ancestors first

In reply to Tejun Heo

When blkcg_activate_policy() is creating blkg_policy_data for existing blkgs, it did in the wrong order - descendants first. Fix it. None of the existing controllers seem affected by this.

---
 block/blk-cgroup.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
block/blk-cgroup.c +2 -1
index 04d286934c5e..440797293235 100644
@@ -1390,7 +1390,8 @@ int blkcg_activate_policy(struct request_queue *q,
 
 	spin_lock_irq(&q->queue_lock);
 
-	list_for_each_entry(blkg, &q->blkg_list, q_node) {
+	/* blkg_list is pushed at the head, reverse walk to init parents first */
+	list_for_each_entry_reverse(blkg, &q->blkg_list, q_node) {
 		struct blkg_policy_data *pd;
 
 		if (blkg->pd[pol->plid])
Signature · 1 line
2.17.1
author maintainer
#5 lore raw

PATCH 5/5 blkcg, writeback: dead memcgs shouldn't contribute to writeback ownership arbitration

In reply to Tejun Heo

wbc_account_io() collects information on cgroup ownership of writeback pages to determine which cgroup should own the inode. Pages can stay associated with dead memcgs but we want to avoid attributing IOs to dead blkcgs as much as possible as the association is likely to be stale. However, currently, pages associated with dead memcgs contribute to the accounting delaying and/or confusing the arbitration.

Fix it by ignoring pages associated with dead memcgs.

---
 fs/fs-writeback.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
fs/fs-writeback.c +7 -1
index e41cbe8e81b9..9ebfb1b28430 100644
@@ -715,6 +715,7 @@ void wbc_detach_inode(struct writeback_control *wbc)
 void wbc_account_io(struct writeback_control *wbc, struct page *page,
 		    size_t bytes)
 {
+	struct cgroup_subsys_state *css;
 	int id;
 
 	/*
@@ -726,7 +727,12 @@ void wbc_account_io(struct writeback_control *wbc, struct page *page,
 	if (!wbc->wb)
 		return;
 
-	id = mem_cgroup_css_from_page(page)->id;
+	css = mem_cgroup_css_from_page(page);
+	/* dead cgroups shouldn't contribute to inode ownership arbitration */
+	if (!(css->flags & CSS_ONLINE))
+		return;
+
+	id = css->id;
 
 	if (id == wbc->wb_id) {
 		wbc->wb_bytes += bytes;
Signature · 1 line
2.17.1
reviewer
#6 lore raw

Re: [PATCH 5/5] blkcg, writeback: dead memcgs shouldn't contribute to writeback ownership arbitration

Quoted 8 lines · Tejun Heo

wbc_account_io() collects information on cgroup ownership of writeback pages to determine which cgroup should own the inode. Pages can stay associated with dead memcgs but we want to avoid attributing IOs to dead blkcgs as much as possible as the association is likely to be stale. However, currently, pages associated with dead memcgs contribute to the accounting delaying and/or confusing the arbitration.

Quoted 4 lines

Fix it by ignoring pages associated with dead memcgs.

I see Jens has already pulled the changes so this is mostly informative but the patch looks good to me.

Honza

Quoted 33 lines
---
 fs/fs-writeback.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
fs/fs-writeback.c +7 -1
index e41cbe8e81b9..9ebfb1b28430 100644
@@ -715,6 +715,7 @@ void wbc_detach_inode(struct writeback_control *wbc)
 void wbc_account_io(struct writeback_control *wbc, struct page *page,
 		    size_t bytes)
 {
+	struct cgroup_subsys_state *css;
 	int id;
 
 	/*
@@ -726,7 +727,12 @@ void wbc_account_io(struct writeback_control *wbc, struct page *page,
 	if (!wbc->wb)
 		return;
 
-	id = mem_cgroup_css_from_page(page)->id;
+	css = mem_cgroup_css_from_page(page);
+	/* dead cgroups shouldn't contribute to inode ownership arbitration */
+	if (!(css->flags & CSS_ONLINE))
+		return;
+
+	id = css->id;
 
 	if (id == wbc->wb_id) {
 		wbc->wb_bytes += bytes;
Signature · 1 line
2.17.1
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
maintainer
#7 lore raw
In reply to Tejun Heo
Quoted 10 lines · Tejun Heo

Hello,

This patchset contains the following blkcg fixes.

0001-blk-iolatency-clear-use_delay-when-io.latency-is-set.patch 0002-blkcg-update-blkcg_print_stat-to-handle-larger-outpu.patch 0003-blkcg-perpcu_ref-init-exit-should-be-done-from-blkg_.patch 0004-blkcg-blkcg_activate_policy-should-initialize-ancest.patch 0005-blkcg-writeback-dead-memcgs-shouldn-t-contribute-to-.patch

Quoted 6 lines

Please refer to each patch's description for details. Patchset is also available in the following git branch.

git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup.git review-blkcg-fixes

Thanks. diffstat follows.

Are you fine with these hitting 5.3?

Signature · 1 line
Jens Axboe
author maintainer
#8 lore raw

Hello,

Quoted 8 lines · Jens Axboe
Quoted 6 lines

Please refer to each patch's description for details. Patchset is also available in the following git branch.

git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup.git review-blkcg-fixes

Thanks. diffstat follows.

Are you fine with these hitting 5.3?

Yeah, none of them are very urgent. 5.3 should be fine.

Thanks.

Signature · 1 line
tejun
maintainer
#9 lore raw
Quoted 6 lines · Tejun Heo

Hello,

Quoted 3 lines · Jens Axboe
Quoted 3 lines

Please refer to each patch's description for details. Patchset is also available in the following git branch.

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup.git review-blkcg-fixes

Thanks. diffstat follows.

Are you fine with these hitting 5.3?

Yeah, none of them are very urgent. 5.3 should be fine.

OK great, added for 5.3, thanks.

Signature · 1 line
Jens Axboe