On Tue, Mar 29, 2016 at 07:04:10PM -1000, Jean-Denis Girard wrote: > Hi list, > > I just started to use send / receive for backups to another drive. > That's a great feature, but unfortunately I'm getting page allocation > failure, see below. > > My backup script does something like this for 11 sub-volumes: > btrfs subvolume snapshot -r vol /snaps > btrfs fi sync /snaps > btrfs send -p /snaps/vol_old /snaps/vol | btrfs receive -v /mnt/backup > btrfs fi sync /mnt/backup > btrfs subvolume delete -c /snaps/vol_old > mv /snaps/vol /snaps/vol_old > btrfs subvolume delete -c /backup/vol_old > btrfs subvolume snapshot -r :backup/vol \ > /backup/vol_$(date +'%Y%m%d') > btrfs fi sync /backup > mv /backup/vol /backup/vol_old > > This is on a up-to-date Fedora 23 system, with kernel > 4.4.6-300.fc23.x86_64, and btrfs-progs v4.4.1 (recompiled on the > system). The system is mostly idle when the error happens. The backup > file system seems clean: btrfs check or scrub report no errors. > > [ 3734.651439] btrfs: page allocation failure: order:4, mode:0x2404040
Order 4 is 64k, and most probably it's the allocation of a nodesize, the IP offset in the function is close to beginning, there are two other allocations that are served from the slab. So do you have a filesystem with a 64k nodesize? Just checking. The memory is fragmented so a contiguous 64k cannot be found, what we can do is a fallback to vmalloc, that can assemble th 64k memory from smaller pages. I'll send a patch. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
