KDL "ExtentStream::FindBlock() invalid header" when mounting

[humdinger]

This is hrev53099, 32bit.

I recently updated from a fairly recent hrev (can't be sure, not older than hrev53060 I think) to hrev53099 (32bit).

I now get a KDL "ExtentStream::FindBlock() invalid header" (see attachment) when trying to mount a specific partition writable and a "Bad data" alert when mounting read-only. For the latter, the syslog says:

KERN: ahci: sg_memcpy phyAddr 0x191f8890, size 8
KERN: ext2: get_vnode: InitCheck() failed. Error: Bad data
KERN: ext2: could not create root node: get_vnode() failed!
KERN: ext2: Failed mounting the volume. Error: Bad data

Which is strange, because it's not ext2, but a BFS partition. Booting a 64bit Haiku Beta1 image I have laying around, I can mount the partition, no problem. A "checkfs" when booted in the Beta1 doesn't find anything wrong.

[humdinger]

KDL when mounting writable

[diver]

Interestingly, some other user on Telegram reported that reiserfs caused KDL for him. However, he didn't have any reiser partitions at all.

[waddlesplash]

Diver, that's an unrelated KDL, and I believe it's a duplicate of a different ticket on here; it appears to be caused by CD media changes conflicting with the IO scheduler and partition rescans.

[waddlesplash]

humdinger, is this reproducible across reboots?

[diver]

I don't recall any ticket like this. Do you have a link?

[humdinger]

Replying to waddlesplash:

humdinger, is this reproducible across reboots?

Yes, 100%.
Any hint where to start my binary hunt for the breaking hrev?

[humdinger]

I've pin-pointed the offending commit: it's hrev53082

[waddlesplash]

Probably that just exposed an old bug in the ext2 filesystem, but cc jessicah anyway.

[jessicah]

Ugh, alright, I'll look into reverting and finding a better way to fix this.

[waddlesplash]

What? No, don't revert it. This just exposed some bug in ext2. If the disk was misidentified by ext2 and it was a non BFS disk we'd have this same problem. We need to just fix ext2.

[jessicah]

Axel wanted it fixed in a better/more intelligent way anyway, which I can understand.

[humdinger]

@jessicah, any news? I wouldn't hate accessing my two mis-indentified partitions again... :)

[humdinger]

FYI, as I needed the disk space, I had to delete/reformat the affected partitions and won't be able to test possible future attempts to fix the issue.

[un_spacyar]

I get the same issue after upgrading from Beta to Nightly (x86_gcc2). When booting from a USB stick, my Haiku partition get recognized as EXT2 by DriveSetup, after the update.

Thanks to @waddlesplash help, I managed to boot it again by blacklisting the EXT2 driver.

Just ask me if you need to test some fix.

[waddlesplash]

hrev53401 may fix this.

[waddlesplash]

Actually fixed in hrev53429.

[nielx]

Assign tickets with status=closed and resolution=fixed within the R1/beta2 development window to the R1/beta2 Milestone