| Message ID | 1609080931-4048864-1-git-send-email-bodefang@126.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | None | expand |
On Sun, Dec 27, 2020 at 10:55:31PM +0800, Defang Bo wrote: > Similar to commmit<389305b2>, Please use full commit reference like 389305b2aa68 ("btrfs: relocation: Only remove reloc rb_trees if reloc control has been initialized") > it turns out that fs_info::reloc_ctl can be NULL , Please describe how the NULL can get there.
I'm keeping the rest of the mail as I received it, it's completely garbled and would need manual reformatting. The changelog and even the diff are completely on one line(!). On Tue, Jan 05, 2021 at 11:08:42AM +0800, bodefang wrote: > Similar to commmit<389305b2aa68>（“btrfs: relocation: Only remove reloc rb_trees if reloc control has been initialized”).it turns out that fs_info::reloc_ctl can be NULL in btrfs_recover_relocation() as we allocate relocation control after allreloc roots have been verified.,so there should be a check for rc to prevent null pointer dereference. > > Signed-off-by: Defang Bo <bodefang@126.com>Reviewed-by: David Sterba <dsterba@suse.cz>--- Oh and don't add reviewed-by unless it's explicitly stated by the person. > Changes since v1: > - More accurate description for this patch to describe how the NULL can get there. > --- > --- fs/btrfs/relocation.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 3602806..ca03571 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -626,6 +626,9 @@ static int __must_check __add_reloc_root(struct btrfs_root *root) struct mapping_node *node; struct reloc_control *rc = fs_info->reloc_ctl; + if (!rc) + return 0; + node = kmalloc(sizeof(*node), GFP_NOFS); if (!node) return -ENOMEM; @@ -703,6 +706,9 @@ static int __update_reloc_root(struct btrfs_root *root) struct rb_node *rb_node; struct mapping_node *node = NULL; struct reloc_control *rc = fs_info->reloc_ctl; + + if (!rc) + return 0; spin_lock(&rc->reloc_root_tree.lock); rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root, -- 2.7.4 > > > > > > > > > > > > > > > > > > At 2021-01-04 23:11:13, "David Sterba" <dsterba@suse.cz> wrote: > >On Sun, Dec 27, 2020 at 10:55:31PM +0800, Defang Bo wrote: > >> Similar to commmit<389305b2>, > > > >Please use full commit reference like 389305b2aa68 ("btrfs: relocation: > >Only remove reloc rb_trees if reloc control has been initialized") > > > >> it turns out that fs_info::reloc_ctl can be NULL , > > > >Please describe how the NULL can get there.
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 3602806..ca03571 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -626,6 +626,9 @@ static int __must_check __add_reloc_root(struct btrfs_root *root) struct mapping_node *node; struct reloc_control *rc = fs_info->reloc_ctl; + if (!rc) + return 0; + node = kmalloc(sizeof(*node), GFP_NOFS); if (!node) return -ENOMEM; @@ -703,6 +706,9 @@ static int __update_reloc_root(struct btrfs_root *root) struct rb_node *rb_node; struct mapping_node *node = NULL; struct reloc_control *rc = fs_info->reloc_ctl; + + if (!rc) + return 0; spin_lock(&rc->reloc_root_tree.lock); rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root,
Similar to commmit<389305b2>, it turns out that fs_info::reloc_ctl can be NULL , so there should be a check for rc to prevent null pointer dereference. Signed-off-by: Defang Bo <bodefang@126.com> --- fs/btrfs/relocation.c | 6 ++++++ 1 file changed, 6 insertions(+)