From f255c19b3ab46d3cad3b1b2e1036f4c926cb1d0c Mon Sep 17 00:00:00 2001
From: Gabriel Krisman Bertazi <krisman@collabora.com>
Date: Thu, 22 Oct 2020 16:58:42 -0400
Subject: [PATCH] blk-cgroup: Pre-allocate tree node on blkg_conf_prep
Git-commit: f255c19b3ab46d3cad3b1b2e1036f4c926cb1d0c
Patch-mainline: v5.10-rc2
References: git-fixes
Similarly to commit 457e490f2b741 ("blkcg: allocate struct blkcg_gq
outside request queue spinlock"), blkg_create can also trigger
occasional -ENOMEM failures at the radix insertion because any
allocation inside blkg_create has to be non-blocking, making it more
likely to fail. This causes trouble for userspace tools trying to
configure io weights who need to deal with this condition.
This patch reduces the occurrence of -ENOMEMs on this path by preloading
the radix tree element on a GFP_KERNEL context, such that we guarantee
the later non-blocking insertion won't fail.
A similar solution exists in blkcg_init_queue for the same situation.
(Coly Li: refreshed for Linux v4.12 based SUSE kernel)
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Coly Li <colyli@suse.de>
---
block/blk-cgroup.c | 14 ++++++++++++--
1 file changed, 12 insertions(+), 2 deletions(-)
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -872,6 +872,12 @@ int blkg_conf_prep(struct blkcg *blkcg,
goto fail;
}
+ if (radix_tree_preload(GFP_KERNEL)) {
+ blkg_free(new_blkg);
+ ret = -ENOMEM;
+ goto fail;
+ }
+
rcu_read_lock();
spin_lock_irq(q->queue_lock);
@@ -879,7 +885,7 @@ int blkg_conf_prep(struct blkcg *blkcg,
if (IS_ERR(blkg)) {
ret = PTR_ERR(blkg);
blkg_free(new_blkg);
- goto fail_unlock;
+ goto fail_preloaded;
}
if (blkg) {
@@ -888,10 +894,12 @@ int blkg_conf_prep(struct blkcg *blkcg,
blkg = blkg_create(pos, q, new_blkg);
if (unlikely(IS_ERR(blkg))) {
ret = PTR_ERR(blkg);
- goto fail_unlock;
+ goto fail_preloaded;
}
}
+ radix_tree_preload_end();
+
if (pos == blkcg)
goto success;
}
@@ -901,6 +909,8 @@ success:
ctx->body = body;
return 0;
+fail_preloaded:
+ radix_tree_preload_end();
fail_unlock:
spin_unlock_irq(q->queue_lock);
rcu_read_unlock();