On 2018/11/17 上午12:17, Stephan Olbrich wrote:
> Hi,
> 
> a few days ago my root file system (simple btrfs on a SSD, no RAID or anything) suddenly became read only. 
> Looking at dmsg, I found this:
> 
> [   19.285020] BTRFS error (device sda2): bad tree block start, want 705757184 have 82362368

Does this only occurs once?

If only once, are you using SINGLE metadata profile (default for SSD)?
If using DUP/RAID1 it may have a chance to recover.

> [   19.285042] BTRFS: error (device sda2) in __btrfs_free_extent:6804: errno=-5 IO failure

The problem is in extent tree.

If there is no other problem, your data should be OK.

You could still try to mount the fs RO to salvage data.

> [   19.285048] BTRFS info (device sda2): forced readonly
> [   19.285051] BTRFS: error (device sda2) in btrfs_run_delayed_refs:2934: errno=-5 IO failure
> [   19.287213] BTRFS error (device sda2): pending csums is 41889792
> 
> Late on I got the same errors for my /home partition (on the same drive) as well.
> I have snapshots of all partitions on another drive made by btrbk. To get a working system, I made new (rw) snapshots
> of the most recent backup and setup grub and fstab, so my system would boot from the other drive.
> Unfortunately now I got the "bad tree block start" error again at least once in dmesg but I didn't save it and it's not in syslog :-(
> What I remember is, that it was followed by other btrfs error messages saying something about correcting something.
> And the filesystem was still read/write this time.
> At the moment I can't reproduce it.
> 
> Is there any way to find out, which files are affected by the errors above?

No files are affected, but an essential tree, extent tree, is corrupted.

Normally this may prevent RW mount, and even it mounts it can still
cause problem when doing any write.
It could even prevent RO mount if the corrupted leaf contains block
group item.

But your data should be OK if there is no other corruption, and in that
case btrfs-restore should work well.

> I don't really trust the data on the drive I'm using at the
> moment, as it has shown errors as well, but I have a less current backup on yet another drive but at it is a few weeks old, I don't
> want to use it to setup the system on the SSD again, but just copy the relevant files if possible.
> Or is it possible to repair the original file system?

At least we need "btrfs check" output.

> 
> Some information about my system:
> Kubuntu 18.04
> Kernel 4.19.1 when the problem occured, now 4.19.2
> btrfs-tools 4.15.1

And "btrfs check" should be executed using latest version.

Thanks,
Qu

> 
> Regards,
> Stephan
> 
> 
> 
> 
>