Kernel 5.0 disk-errors on boot, can't boot up: is it just me?

RandomTroll · 03-04-2019, 04:54 PM

I just built kernel 5.0. I had been running 4.20.13. I unpatched with the 4.20.13 patch then applied the 5.0 patch. It ran without complaint. When I try to boot into it, I get hard disk errors. I can still boot into 4.20.13. e2fsck run from a boot flash drive finds no errors. I had disk errors before; this may be a problem only I have. Has it worked for everyone else?

Later: I built the kernel from the 5.0 tarball, get the same result.

Quote:

EXT4-fs error in ext4_reserve_inode_write: Journal has aborted
EXT4-fs ext4_journal_check_start [whoops! forgot to write this down, but it fails]

then it mounts read-only.

bathory · 03-05-2019, 05:07 AM

Quote:

Originally Posted by RandomTroll

I just built kernel 5.0. I had been running 4.20.13. I unpatched with the 4.20.13 patch then applied the 5.0 patch. It ran without complaint. When I try to boot into it, I get hard disk errors. I can still boot into 4.20.13. e2fsck run from a boot flash drive finds no errors. I had disk errors before; this may be a problem only I have. Has it worked for everyone else?

Later: I built the kernel from the 5.0 tarball, get the same result.

then it mounts read-only.

I'm running 5.0 on 2 boxes running Slackware64-current without any hitch.
What patch did you apply?

RandomTroll · 03-05-2019, 09:38 AM

Quote:

Originally Posted by bathory

What patch did you apply?

patch-5.0.xz I later downloaded the whole source then built from that, had the same result.

bathory · 03-05-2019, 03:32 PM

Quote:

Originally Posted by RandomTroll

patch-5.0.xz I later downloaded the whole source then built from that, had the same result.

Doh, I too downloaded the kernel source from here, and built it using the .config of the previous running kernel.
FYI, even the nvidia latest drivers (version 418.43) worked!

RandomTroll · 03-05-2019, 05:45 PM

I always get my kernel from kernel.org. I booted from a flash drive, turned off the journal, can boot, get

Quote:

EXT4-fs error (device sda1): ext4_find_entry:1439: inode #661755: comm udevd: reading directory lblock 0

but it seems to work otherwise. Do you use journals? I don't know how to interpret this error, why udevd is involved. How much do I miss by not-journaling? Why do I get this error in kernel 5.0 but not 4.20.13?

bathory · 03-06-2019, 02:01 AM

Quote:

Originally Posted by RandomTroll

I always get my kernel from kernel.org. I booted from a flash drive, turned off the journal, can boot, get but it seems to work otherwise. Do you use journals? I don't know how to interpret this error, why udevd is involved. How much do I miss by not-journaling? Why do I get this error in kernel 5.0 but not 4.20.13?

From what I've read this error affects SSD drives, which I don't have.
Other than that, I use the default journaling capabilities of EXT4 and the kernel is compiled with:

Code:

CONFIG_EXT4_FS=y
CONFIG_EXT4_FS_POSIX_ACL=y
CONFIG_EXT4_FS_SECURITY=y
# CONFIG_EXT4_ENCRYPTION is not set
# CONFIG_EXT4_DEBUG is not set

RandomTroll · 03-06-2019, 09:14 AM

Quote:

Originally Posted by bathory

From what I've read this error affects SSD drives, which I don't have.

Where have you read this? I don't have SSD drives.

Quote:

Originally Posted by bathory

Other than that, I use the default journaling capabilities of EXT4 and the kernel is compiled with:

What does

Code:

 dumpe2fs /dev/sda1 | grep journal

return?

bathory · 03-06-2019, 12:05 PM

Quote:

Originally Posted by RandomTroll

Where have you read this? I don't have SSD drives.

Put the error you get in the search engine and the first hits were about SSD drives giving the same error.
Also what is strange is that these first hits were about ubuntu or debian distros

Quote:

What does

Code:

 dumpe2fs /dev/sda1 | grep journal

return?

Code:

dumpe2fs /dev/sda1 | grep -i journal
dumpe2fs 1.44.5 (15-Dec-2018)
Filesystem features:      has_journal ext_attr resize_inode dir_index filetype needs_recovery extent flex_bg sparse_super large_file huge_file dir_nlink extra_isize
Journal inode:            8
Journal backup:           inode blocks
Journal features:         journal_incompat_revoke
Journal size:             128M
Journal length:           32768
Journal sequence:         0x00064a64
Journal start:            1

RandomTroll · 03-06-2019, 03:42 PM

Quote:

Originally Posted by bathory

Put the error you get in the search engine and the first hits were about SSD drives giving the same error.
Also what is strange is that these first hits were about ubuntu or debian distros

These are old messages about old kernels. Your disk has a journal, so it's not a general problem. The journal on my other volume has no problem. I seem to have some subtle defects with my hardware. I don't need a journal on the boot volume, so I'll let it ride. Thanks for your help.

RandomTroll · 03-14-2019, 11:03 AM

I try again every kernel update: no luck. I notice this error:

Quote:

ext4_dirty_inode:6087: IO Failure

I find this error from kernels in 2.6, relating to a bug when JBD_DEBUG isn't being used: the kernel didn't mark the inode as dirty. Could a similar error have recrudesced? I use JBD2_DEBUG. One would think this inode would get marked dirty, which would solve the problem. e2fsck finds no errors, so the inode can't have failed because the sector in which it resides is bad.

bathory · 03-14-2019, 01:38 PM

Quote:

Originally Posted by RandomTroll

I try again every kernel update: no luck. I notice this error: I find this error from kernels in 2.6, relating to a bug when JBD_DEBUG isn't being used: the kernel didn't mark the inode as dirty. Could a similar error have recrudesced? I use JBD2_DEBUG. One would think this inode would get marked dirty, which would solve the problem. e2fsck finds no errors, so the inode can't have failed because the sector in which it resides is bad.

Perhaps you should run a smartctl check on the disk, because getting different errors any time could mean a hardware problem

RandomTroll · 03-14-2019, 01:59 PM

Quote:

Originally Posted by bathory

Perhaps you should run a smartctl check on the disk, because getting different errors any time could mean a hardware problem

I always get the same error. I finally installed 5.0.2 on my old disk and booted from it: it works, which points to my new disk having the problem. sigh.

Later: I bought a new disk because smartctl reported errors in the old one; it doesn't in the new one.

bathory · 03-15-2019, 01:59 AM

Quote:

Originally Posted by RandomTroll

I always get the same error. I finally installed 5.0.2 on my old disk and booted from it: it works, which points to my new disk having the problem. sigh.

Later: I bought a new disk because smartctl reported errors in the old one; it doesn't in the new one.

Could be a faulty cable, or (SATA) connector on the motherboard.

RandomTroll · 03-15-2019, 11:55 AM

Quote:

Originally Posted by bathory

Could be a faulty cable, or (SATA) connector on the motherboard.

It's a laptop.

RandomTroll · 07-14-2019, 12:57 PM

This error has expanded in an odd way: when I build a new kernel, the journal on /dev/sda2 (which mounts on /home) errors, the drive becomes unwritable - before I boot into the new kernel - complaining of a journal error. I re-boot and it works as it did before. /dev/sda1, which mounts on / has no problem; it has no journal.