[PATCH v5 0/2] btrfs: fix sleep from invalid context bug in update_qgroup_limit_item()

[PATCH v5 0/2] btrfs: fix sleep from invalid context bug in update_qgroup_limit_item()

[ChenXiaoSong]

At least 3 places might sleep in update_qgroup_limit_item(), as shown below:

  update_qgroup_limit_item
    btrfs_alloc_path
      /* allocate memory non-atomically, might sleep */
      kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS)
    btrfs_search_slot
      setup_nodes_for_search
        reada_for_balance
          btrfs_readahead_node_child
            btrfs_readahead_tree_block
              btrfs_find_create_tree_block
                alloc_extent_buffer
                  kmem_cache_zalloc
                    /* allocate memory non-atomically, might sleep */
                    kmem_cache_alloc(GFP_NOFS|__GFP_NOFAIL|__GFP_ZERO)
              read_extent_buffer_pages
                submit_extent_page
                  /* disk IO, might sleep */
                  submit_one_bio

Fix this by calling qgroup_dirty() on @dstqgroup, and update limit item in
btrfs_run_qgroups() later.

By the way, add might_sleep() to btrfs_search_slot().

ChenXiaoSong (2):
  btrfs: add might_sleep() to btrfs_search_slot()
  btrfs: qgroup: fix sleep from invalid context bug in
    update_qgroup_limit_item()

 fs/btrfs/ctree.c  |  2 ++
 fs/btrfs/qgroup.c | 29 ++++++++++++-----------------
 2 files changed, 14 insertions(+), 17 deletions(-)

-- 
2.31.1

[PATCH v5 1/2] btrfs: add might_sleep() to btrfs_search_slot()

[ChenXiaoSong]

At least 2 places might sleep in btrfs_search_slot(), as shown below:

  btrfs_search_slot
    setup_nodes_for_search
      reada_for_balance
        btrfs_readahead_node_child
          btrfs_readahead_tree_block
            btrfs_find_create_tree_block
              alloc_extent_buffer
                kmem_cache_zalloc
                  /* allocate memory non-atomically, might sleep */
                  kmem_cache_alloc(GFP_NOFS|__GFP_NOFAIL|__GFP_ZERO)
            read_extent_buffer_pages
              submit_extent_page
                /* disk IO, might sleep */
                submit_one_bio

As the potential sleeping under spin lock is hard to spot, we should add
might_sleep() to btrfs_search_slot().

Signed-off-by: ChenXiaoSong <chenxiaosong2@huawei.com>
---
 fs/btrfs/ctree.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index a9543f01184c..809053e9cfde 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -1934,6 +1934,8 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	int min_write_lock_level;
 	int prev_cmp;
 
+	might_sleep();
+
 	lowest_level = p->lowest_level;
 	WARN_ON(lowest_level && ins_len > 0);
 	WARN_ON(p->nodes[0] != NULL);
-- 
2.31.1

[PATCH v5 2/2] btrfs: qgroup: fix sleep from invalid context bug in update_qgroup_limit_item()

[ChenXiaoSong]

Syzkaller reported BUG as follows:

  BUG: sleeping function called from invalid context at
       include/linux/sched/mm.h:274
  Call Trace:
   <TASK>
   dump_stack_lvl+0xcd/0x134
   __might_resched.cold+0x222/0x26b
   kmem_cache_alloc+0x2e7/0x3c0
   update_qgroup_limit_item+0xe1/0x390
   btrfs_qgroup_inherit+0x147b/0x1ee0
   create_subvol+0x4eb/0x1710
   btrfs_mksubvol+0xfe5/0x13f0
   __btrfs_ioctl_snap_create+0x2b0/0x430
   btrfs_ioctl_snap_create_v2+0x25a/0x520
   btrfs_ioctl+0x2a1c/0x5ce0
   __x64_sys_ioctl+0x193/0x200
   do_syscall_64+0x35/0x80

Fix this by calling qgroup_dirty() on @dstqgroup, and update limit item in
btrfs_run_qgroups() later.

Signed-off-by: ChenXiaoSong <chenxiaosong2@huawei.com>
---
 fs/btrfs/qgroup.c | 29 ++++++++++++-----------------
 1 file changed, 12 insertions(+), 17 deletions(-)

diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index 9334c3157c22..8f5c52e24430 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -2860,6 +2860,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 	bool need_rescan = false;
 	u32 level_size = 0;
 	u64 nums;
+	bool dirty_dstgrp = false;
 
 	/*
 	 * There are only two callers of this function.
@@ -2941,7 +2942,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 	dstgroup = add_qgroup_rb(fs_info, objectid);
 	if (IS_ERR(dstgroup)) {
 		ret = PTR_ERR(dstgroup);
-		goto unlock;
+		goto dirty;
 	}
 
 	if (inherit && inherit->flags & BTRFS_QGROUP_INHERIT_SET_LIMITS) {
@@ -2950,21 +2951,13 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 		dstgroup->max_excl = inherit->lim.max_excl;
 		dstgroup->rsv_rfer = inherit->lim.rsv_rfer;
 		dstgroup->rsv_excl = inherit->lim.rsv_excl;
-
-		ret = update_qgroup_limit_item(trans, dstgroup);
-		if (ret) {
-			qgroup_mark_inconsistent(fs_info);
-			btrfs_info(fs_info,
-				   "unable to update quota limit for %llu",
-				   dstgroup->qgroupid);
-			goto unlock;
-		}
+		dirty_dstgrp = true;
 	}
 
 	if (srcid) {
 		srcgroup = find_qgroup_rb(fs_info, srcid);
 		if (!srcgroup)
-			goto unlock;
+			goto dirty;
 
 		/*
 		 * We call inherit after we clone the root in order to make sure
@@ -2985,20 +2978,20 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 		dstgroup->max_excl = srcgroup->max_excl;
 		dstgroup->rsv_rfer = srcgroup->rsv_rfer;
 		dstgroup->rsv_excl = srcgroup->rsv_excl;
+		dirty_dstgrp = true;
 
-		qgroup_dirty(fs_info, dstgroup);
 		qgroup_dirty(fs_info, srcgroup);
 	}
 
 	if (!inherit)
-		goto unlock;
+		goto dirty;
 
 	i_qgroups = (u64 *)(inherit + 1);
 	for (i = 0; i < inherit->num_qgroups; ++i) {
 		if (*i_qgroups) {
 			ret = add_relation_rb(fs_info, objectid, *i_qgroups);
 			if (ret)
-				goto unlock;
+				goto dirty;
 		}
 		++i_qgroups;
 
@@ -3022,7 +3015,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 
 		if (!src || !dst) {
 			ret = -EINVAL;
-			goto unlock;
+			goto dirty;
 		}
 
 		dst->rfer = src->rfer - level_size;
@@ -3043,15 +3036,17 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
 
 		if (!src || !dst) {
 			ret = -EINVAL;
-			goto unlock;
+			goto dirty;
 		}
 
 		dst->excl = src->excl + level_size;
 		dst->excl_cmpr = src->excl_cmpr + level_size;
 		need_rescan = true;
 	}
+dirty:
+	if (dirty_dstgrp)
+		qgroup_dirty(fs_info, dstgroup);
 
-unlock:
 	spin_unlock(&fs_info->qgroup_lock);
 	if (!ret)
 		ret = btrfs_sysfs_add_one_qgroup(fs_info, dstgroup);
-- 
2.31.1

Re: [PATCH v5 2/2] btrfs: qgroup: fix sleep from invalid context bug in update_qgroup_limit_item()

[Qu Wenruo]

On 2022/11/16 21:07, ChenXiaoSong wrote:
> Syzkaller reported BUG as follows:
> 
>    BUG: sleeping function called from invalid context at
>         include/linux/sched/mm.h:274
>    Call Trace:
>     <TASK>
>     dump_stack_lvl+0xcd/0x134
>     __might_resched.cold+0x222/0x26b
>     kmem_cache_alloc+0x2e7/0x3c0
>     update_qgroup_limit_item+0xe1/0x390
>     btrfs_qgroup_inherit+0x147b/0x1ee0
>     create_subvol+0x4eb/0x1710
>     btrfs_mksubvol+0xfe5/0x13f0
>     __btrfs_ioctl_snap_create+0x2b0/0x430
>     btrfs_ioctl_snap_create_v2+0x25a/0x520
>     btrfs_ioctl+0x2a1c/0x5ce0
>     __x64_sys_ioctl+0x193/0x200
>     do_syscall_64+0x35/0x80
> 
> Fix this by calling qgroup_dirty() on @dstqgroup, and update limit item in
> btrfs_run_qgroups() later.
> 
> Signed-off-by: ChenXiaoSong <chenxiaosong2@huawei.com>
> ---
>   fs/btrfs/qgroup.c | 29 ++++++++++++-----------------
>   1 file changed, 12 insertions(+), 17 deletions(-)
> 
> diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
> index 9334c3157c22..8f5c52e24430 100644
> --- a/fs/btrfs/qgroup.c
> +++ b/fs/btrfs/qgroup.c
> @@ -2860,6 +2860,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   	bool need_rescan = false;
>   	u32 level_size = 0;
>   	u64 nums;
> +	bool dirty_dstgrp = false;

I don't know why you insist on such bool (and the extra lable).

I have mentioned how qgroup_dirty() works, it can be called how ever 
many times, and the qgroup code will handle it without problem.

So the whole patch can be just as simple as:

diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index 05e79f7b4433..e0522c6c0d67 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -2965,14 +2965,7 @@ int btrfs_qgroup_inherit(struct 
btrfs_trans_handle *trans, u64 srcid,
                 dstgroup->rsv_rfer = inherit->lim.rsv_rfer;
                 dstgroup->rsv_excl = inherit->lim.rsv_excl;

-               ret = update_qgroup_limit_item(trans, dstgroup);
-               if (ret) {
-                       qgroup_mark_inconsistent(fs_info);
-                       btrfs_info(fs_info,
-                                  "unable to update quota limit for %llu",
-                                  dstgroup->qgroupid);
-                       goto unlock;
-               }
+               qgroup_dirty(fs_info, dstgroup);
         }

         if (srcid) {


>   
>   	/*
>   	 * There are only two callers of this function.
> @@ -2941,7 +2942,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   	dstgroup = add_qgroup_rb(fs_info, objectid);
>   	if (IS_ERR(dstgroup)) {
>   		ret = PTR_ERR(dstgroup);
> -		goto unlock;
> +		goto dirty;
>   	}
>   
>   	if (inherit && inherit->flags & BTRFS_QGROUP_INHERIT_SET_LIMITS) {
> @@ -2950,21 +2951,13 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   		dstgroup->max_excl = inherit->lim.max_excl;
>   		dstgroup->rsv_rfer = inherit->lim.rsv_rfer;
>   		dstgroup->rsv_excl = inherit->lim.rsv_excl;
> -
> -		ret = update_qgroup_limit_item(trans, dstgroup);
> -		if (ret) {
> -			qgroup_mark_inconsistent(fs_info);
> -			btrfs_info(fs_info,
> -				   "unable to update quota limit for %llu",
> -				   dstgroup->qgroupid);
> -			goto unlock;
> -		}
> +		dirty_dstgrp = true;
>   	}
>   
>   	if (srcid) {
>   		srcgroup = find_qgroup_rb(fs_info, srcid);
>   		if (!srcgroup)
> -			goto unlock;
> +			goto dirty;
>   
>   		/*
>   		 * We call inherit after we clone the root in order to make sure
> @@ -2985,20 +2978,20 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   		dstgroup->max_excl = srcgroup->max_excl;
>   		dstgroup->rsv_rfer = srcgroup->rsv_rfer;
>   		dstgroup->rsv_excl = srcgroup->rsv_excl;
> +		dirty_dstgrp = true;
>   
> -		qgroup_dirty(fs_info, dstgroup);
>   		qgroup_dirty(fs_info, srcgroup);
>   	}
>   
>   	if (!inherit)
> -		goto unlock;
> +		goto dirty;
>   
>   	i_qgroups = (u64 *)(inherit + 1);
>   	for (i = 0; i < inherit->num_qgroups; ++i) {
>   		if (*i_qgroups) {
>   			ret = add_relation_rb(fs_info, objectid, *i_qgroups);
>   			if (ret)
> -				goto unlock;
> +				goto dirty;
>   		}
>   		++i_qgroups;
>   
> @@ -3022,7 +3015,7 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   
>   		if (!src || !dst) {
>   			ret = -EINVAL;
> -			goto unlock;
> +			goto dirty;
>   		}
>   
>   		dst->rfer = src->rfer - level_size;
> @@ -3043,15 +3036,17 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans, u64 srcid,
>   
>   		if (!src || !dst) {
>   			ret = -EINVAL;
> -			goto unlock;
> +			goto dirty;
>   		}
>   
>   		dst->excl = src->excl + level_size;
>   		dst->excl_cmpr = src->excl_cmpr + level_size;
>   		need_rescan = true;
>   	}
> +dirty:
> +	if (dirty_dstgrp)
> +		qgroup_dirty(fs_info, dstgroup);
>   
> -unlock:
>   	spin_unlock(&fs_info->qgroup_lock);
>   	if (!ret)
>   		ret = btrfs_sysfs_add_one_qgroup(fs_info, dstgroup);