The patch below does not apply to the 4.14-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to stable@vger.kernel.org.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
From 998ac6d21cfd6efd58f5edf420bae8839dda9f2a Mon Sep 17 00:00:00 2001
From: ethanwu ethanwu@synology.com Date: Sun, 29 Apr 2018 15:59:42 +0800 Subject: [PATCH] btrfs: Take trans lock before access running trans in check_delayed_ref
In preivous patch: Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist We avoid starting btrfs transaction and get this information from fs_info->running_transaction directly.
When accessing running_transaction in check_delayed_ref, there's a chance that current transaction will be freed by commit transaction after the NULL pointer check of running_transaction is passed.
After looking all the other places using fs_info->running_transaction, they are either protected by trans_lock or holding the transactions.
Fix this by using trans_lock and increasing the use_count.
Fixes: e4c3b2dcd144 ("Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist") CC: stable@vger.kernel.org # 4.14+ Signed-off-by: ethanwu ethanwu@synology.com Signed-off-by: David Sterba dsterba@suse.com
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index f99102063366..3871658b6ab1 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3142,7 +3142,11 @@ static noinline int check_delayed_ref(struct btrfs_root *root, struct rb_node *node; int ret = 0;
+ spin_lock(&root->fs_info->trans_lock); cur_trans = root->fs_info->running_transaction; + if (cur_trans) + refcount_inc(&cur_trans->use_count); + spin_unlock(&root->fs_info->trans_lock); if (!cur_trans) return 0;
@@ -3151,6 +3155,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, head = btrfs_find_delayed_ref_head(delayed_refs, bytenr); if (!head) { spin_unlock(&delayed_refs->lock); + btrfs_put_transaction(cur_trans); return 0; }
@@ -3167,6 +3172,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, mutex_lock(&head->mutex); mutex_unlock(&head->mutex); btrfs_put_delayed_ref_head(head); + btrfs_put_transaction(cur_trans); return -EAGAIN; } spin_unlock(&delayed_refs->lock); @@ -3199,6 +3205,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, } spin_unlock(&head->lock); mutex_unlock(&head->mutex); + btrfs_put_transaction(cur_trans); return ret; }
commit 998ac6d21cfd6efd58f5edf420bae8839dda9f2a upstream
In preivous patch: Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist We avoid starting btrfs transaction and get this information from fs_info->running_transaction directly.
When accessing running_transaction in check_delayed_ref, there's a chance that current transaction will be freed by commit transaction after the NULL pointer check of running_transaction is passed.
After looking all the other places using fs_info->running_transaction, they are either protected by trans_lock or holding the transactions.
Fix this by using trans_lock and increasing the use_count.
Fixes: e4c3b2dcd144 ("Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist") CC: stable@vger.kernel.org # 4.14+ Signed-off-by: ethanwu ethanwu@synology.com Signed-off-by: David Sterba dsterba@suse.com --- fs/btrfs/extent-tree.c | 7 +++++++ 1 file changed, 7 insertions(+)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index d227d85..1bc6229 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3171,7 +3171,11 @@ static noinline int check_delayed_ref(struct btrfs_root *root, struct btrfs_transaction *cur_trans; int ret = 0;
+ spin_lock(&root->fs_info->trans_lock); cur_trans = root->fs_info->running_transaction; + if (cur_trans) + refcount_inc(&cur_trans->use_count); + spin_unlock(&root->fs_info->trans_lock); if (!cur_trans) return 0;
@@ -3180,6 +3184,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, head = btrfs_find_delayed_ref_head(delayed_refs, bytenr); if (!head) { spin_unlock(&delayed_refs->lock); + btrfs_put_transaction(cur_trans); return 0; }
@@ -3196,6 +3201,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, mutex_lock(&head->mutex); mutex_unlock(&head->mutex); btrfs_put_delayed_ref(&head->node); + btrfs_put_transaction(cur_trans); return -EAGAIN; } spin_unlock(&delayed_refs->lock); @@ -3223,6 +3229,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root, } spin_unlock(&head->lock); mutex_unlock(&head->mutex); + btrfs_put_transaction(cur_trans); return ret; }
On Thu, May 17, 2018 at 05:01:45PM +0800, ethanwu wrote:
commit 998ac6d21cfd6efd58f5edf420bae8839dda9f2a upstream
In preivous patch: Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist We avoid starting btrfs transaction and get this information from fs_info->running_transaction directly.
When accessing running_transaction in check_delayed_ref, there's a chance that current transaction will be freed by commit transaction after the NULL pointer check of running_transaction is passed.
After looking all the other places using fs_info->running_transaction, they are either protected by trans_lock or holding the transactions.
Fix this by using trans_lock and increasing the use_count.
Fixes: e4c3b2dcd144 ("Btrfs: kill trans in run_delalloc_nocow and btrfs_cross_ref_exist") CC: stable@vger.kernel.org # 4.14+ Signed-off-by: ethanwu ethanwu@synology.com Signed-off-by: David Sterba dsterba@suse.com
fs/btrfs/extent-tree.c | 7 +++++++ 1 file changed, 7 insertions(+)
Thanks for the backport, now queued up.
greg k-h
linux-stable-mirror@lists.linaro.org