Re: read time tree block corruption detected

[Nikolay Borisov <nborisov@suse.com>]

On 16.01.20 г. 15:40 ч., Peter Luladjiev wrote:
> Hello,
> 
> I'm having a "read time tree block corruption detected" error after
> upgrading to kernel 5.4, reverting to kernel 5.3 fixes the problem.
> I'm using openSuse Tumbleweed, btrfs-progs v5.4

The reason is that the read-time validator got added in v5.4. 
Looking at the provided dump: 


item 70 key (1497006080 EXTENT_ITEM 4096) itemoff 11757 itemsize 37
		refs 1 gen 17889 flags DATA
		shared data backref parent 51611369472 count 1
	item 71 key (1497010176 EXTENT_ITEM 4096) itemoff 11720 itemsize 37
		refs 1 gen 17890 flags DATA
		shared data backref parent 51611369472 count 1
	item 72 key (1497014272 EXTENT_ITEM 4096) itemoff 11683 itemsize 37
		refs 1 gen 17892 flags DATA
		shared data backref parent 51611369472 count 1
	item 73 key (1497018368 EXTENT_ITEM 4096) itemoff 11646 itemsize 37
		refs 72057183177116417 gen 17894 flags DATA
		shared data backref parent 51611369472 count 1
	item 74 key (1497022464 EXTENT_ITEM 4096) itemoff 11609 itemsize 37
		refs 1 gen 17896 flags DATA
		shared data backref parent 51611369472 count 3087007745


Clearly the problematic extent has an abnormal count: 

3087007745, in binary this makes: 
10111000000000000000000000000001

so the topmost 5 bits (apart from 1) has been set to 1 whereas 
I'd assume  only the LSB (1) should be set, judging based on the the rest
of the extents that have 51611369472 as their parent. 

Can you run btrfs check on /dev/mapper/system-root and post the log here. 
Note this will be a read-only operation to you don't risk breaking
 anything. You might also have to set the --force option since this looks 
like to be your root fs which is always mounted (barring working from a rescue
disk).