Re: Subject: zoned: transaction abort with -EAGAIN (-11) on host-managed SMR (WD HC670), fs forced read-only

[Johannes Thumshirn <johannes.thumshirn@wdc.com>]

On Fri, Nov 06, 2026 at 08:36:16AM +0100, Jochen Philipp wrote:
> Hi Dongjiang,
> 
> thanks for reaching out - yes, I'm willing to give it a try.
> 
> Johannes asked me to keep the on-disk state untouched while he
> looks at the zone report, so I'd like to coordinate with him
> first: Johannes, are you fine with me testing Dongjiang's series
> on this fs, or do you need anything else from the current state
> before I do?
> 
> For testing I can set up a separate machine (bare metal, no
> out-of-tree modules), so my main system isn't affected and
> switching kernels is easy. I'd move the drive over there.
> 
> Two practical questions, since I've never built a patched kernel
> for testing:
> - Which base should I apply the series to (e.g. 7.3, or a
>   specific btrfs branch)?
> - Is there a git branch with the series applied that I could
>   use instead?

I think I could re-create the issue and unfortunately Dongjiang's patch didn't
fix it for me (I've only applied patch 2 not the whole branch). What fixes it
for me is:


diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c
index 3192b86669b9..3724cc0120d9 100644
--- a/fs/btrfs/zoned.c
+++ b/fs/btrfs/zoned.c
@@ -3115,8 +3115,12 @@ void btrfs_check_active_zone_reservation(struct btrfs_fs_info *fs_info)
 		      (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_SYSTEM)))
 			continue;
 
-		for (int i = 0; i < map->num_stripes; i++)
-			map->stripes[i].dev->zone_info->reserved_active_zones--;
+		for (int i = 0; i < map->num_stripes; i++) {
+			struct btrfs_device *dev = map->stripes[i].dev;
+
+			if (btrfs_dev_is_sequential(dev, map->stripes[i].physical))
+				dev->zone_info->reserved_active_zones--;
+		}
 	}
 	spin_unlock(&fs_info->zone_active_bgs_lock);
 }