[arch-general] Segmentation fault and (possibly) file system bug while removing gnome-extra

Discussion:

Robert W0ng via arch-general

2017-05-13 23:19:05 UTC

Hi, All.

For several reasons I decided to remove the group 'gnome-extra'
from my system. During the uninstallation, something went wrong.

For a 'funny' reason, which I may explain later, I wasn't able
to post the error log in the form of text here, so I will upload
a photo as an attachment[1] with this mail...

So, I switched to a getty and started the uninstallation. No
errors occourred until it reached the 15th package. The package
name was 'telepathy-idle'. All of a sudden the system printed
a large section of dmesg, saying something like it had run into
a kernel bug(?). The pacman was then killed, leaving the zsh
reporting a segmentation fault...

I felt the possibility that I wasn't able to handle this problem
on my own, so I took a photograph using my mobile without even
thinking, which proved to be useful later. Then I ran

# dmesg >> /root/dmesg.output

hoping to save the error log. I tried to restart the
uninstallation, but the pacman freezed. Even pressing CTRL-C
didn't work. I switched to another getty attemptting to terminal
in, but after I entered 'root' and pressed ENTER, no any output.
The course was just left there, blinking. Having no choice, I
reboot the system using the magic sysrq keys...

Rebooted, I checked the package changes. Confusingly, no packages
was removed, even the previous should-be-removed 15 ones! Bad
feeling about this, I checked the /root directory, and here came
the most confusing part: The dmesg.output file was GONE! It
seemed that all the data that should have been written to the
disk were discarded...

(And that's the reason why I have to post the error log in the
form of shooted image... Somehow, I found myself a potential
prophet... :P JUST KID...)

Judged by this, together with the previous error log, I would
say it's a file system bug... But, confused, how can an
uninstallation of packages end up like this way? :(

I would appreciate any help from you guys. Truly would. Thank
you in advance! :)

PS: I'm not going to post my pacman.conf here because all
modification I had ever done to this file was to enable the
color output. I suppose it can be excluded from the possible
causes... And if necessary, I would recur the
uninstallation with '--debug' flag enabled and report
it to the upstream developer...

[1] Upadte: Unable to upload the image file because it's too
big to be attached in this mailing list... So I
uploaded it to a local website and I will post the
image address here.

Image Address: http://imgsrc.baidu.com/forum/pic/item/7bf9f2c4b74543a9b
9da16cf14178a82bb0114d4.jpg

RW

Robert W0ng via arch-general

2017-05-14 02:29:07 UTC

Permalink

Hello again, All.

Update:

I ran fsck in archiso this morning, and found sth interesting.

To sum up, fsck.f2fs exit with signal 11, arguing segmentation fault.
Just the same as the error I faced last night. (I wrote the previous
mail last night and sent it this morning.)

Now that I knew it's DEFINITELY a file system error. And I wonder if
there's any solution except formatting the whole partition.

Again, I shooted the error log and I'll post it here:
http://imgsrc.baidu.com/forum/pic/item/2318034c510fd9f90c88e96d2f2dd42a
2934a47c.jpg

Lots of thx!!!

RW

Post by Robert W0ng via arch-general
Hi, All.
For several reasons I decided to remove the group 'gnome-extra'
from my system. During the uninstallation, something went wrong.
For a 'funny' reason, which I may explain later, I wasn't able
to post the error log in the form of text here, so I will upload
a photo as an attachment[1] with this mail...
So, I switched to a getty and started the uninstallation. No
errors occourred until it reached the 15th package. The package
name was 'telepathy-idle'. All of a sudden the system printed
a large section of dmesg, saying something like it had run into
a kernel bug(?). The pacman was then killed, leaving the zsh
reporting a segmentation fault...
I felt the possibility that I wasn't able to handle this problem
on my own, so I took a photograph using my mobile without even
thinking, which proved to be useful later. Then I ran
# dmesg >> /root/dmesg.output
hoping to save the error log. I tried to restart the
uninstallation, but the pacman freezed. Even pressing CTRL-C
didn't work. I switched to another getty attemptting to terminal
in, but after I entered 'root' and pressed ENTER, no any output.
The course was just left there, blinking. Having no choice, I
reboot the system using the magic sysrq keys...
Rebooted, I checked the package changes. Confusingly, no packages
was removed, even the previous should-be-removed 15 ones! Bad
feeling about this, I checked the /root directory, and here came
the most confusing part: The dmesg.output file was GONE! It
seemed that all the data that should have been written to the
disk were discarded...
(And that's the reason why I have to post the error log in the
form of shooted image... Somehow, I found myself a potential
prophet... :P JUST KID...)
Judged by this, together with the previous error log, I would
say it's a file system bug... But, confused, how can an
uninstallation of packages end up like this way? :(
I would appreciate any help from you guys. Truly would. Thank
you in advance! :)
PS: I'm not going to post my pacman.conf here because all
modification I had ever done to this file was to enable the
color output. I suppose it can be excluded from the possible
causes... And if necessary, I would recur the
uninstallation with '--debug' flag enabled and report
it to the upstream developer...
[1] Upadte: Unable to upload the image file because it's too
big to be attached in this mailing list... So I
uploaded it to a local website and I will post the
image address here.
Image Address: http://imgsrc.baidu.com/forum/pic/item/7bf9f2c4b74543a
9b
9da16cf14178a82bb0114d4.jpg
RW

Dragon ryu via arch-general

2017-05-14 03:23:27 UTC

Permalink

2017/05/14 午前11:29 "Robert W0ng via arch-general" <
arch-***@archlinux.org>:

Hello again, All.

Update:

I ran fsck in archiso this morning, and found sth interesting.

To sum up, fsck.f2fs exit with signal 11, arguing segmentation fault.
Just the same as the error I faced last night. (I wrote the previous
mail last night and sent it this morning.)

Now that I knew it's DEFINITELY a file system error. And I wonder if
there's any solution except formatting the whole partition.

Again, I shooted the error log and I'll post it here:
http://imgsrc.baidu.com/forum/pic/item/2318034c510fd9f90c88e96d2f2dd42a
2934a47c.jpg

Lots of thx!!!

RW

uhm, I'd suggest you to write proper link

Robert Wong via arch-general

2017-05-14 05:59:55 UTC

Permalink

Well... It seems that all the links were divided into two lines... But after
putting them in one line accordingly, it should be proper links... Sorry. :P

------------------
Regards,
Robert S. Wong

------------------ Original message ------------------
From: "Dragon ryu via arch-general";
Sendtime: Sunday, May 14, 2017 11:23 AM
To: "General Discussion about Arch Linux";
Cc: "Dragon ryu";
Subject: Re: [arch-general] Segmentation fault and (possibly) file systembug while removing gnome-extra

2017/05/14 午前11:29 "Robert W0ng via arch-general" <
arch-***@archlinux.org>:

Hello again, All.

Update:

I ran fsck in archiso this morning, and found sth interesting.

To sum up, fsck.f2fs exit with signal 11, arguing segmentation fault.
Just the same as the error I faced last night. (I wrote the previous
mail last night and sent it this morning.)

Now that I knew it's DEFINITELY a file system error. And I wonder if
there's any solution except formatting the whole partition.

Again, I shooted the error log and I'll post it here:
http://imgsrc.baidu.com/forum/pic/item/2318034c510fd9f90c88e96d2f2dd42a
2934a47c.jpg

Lots of thx!!!

RW

uhm, I'd

Guus Snijders via arch-general

2017-05-14 13:18:50 UTC

Permalink

Op 14 mei 2017 04:29 schreef "Robert W0ng via arch-general" <
arch-***@archlinux.org>:

Hello again, All.

Update:

I ran fsck in archiso this morning, and found sth interesting.

To sum up, fsck.f2fs exit with signal 11, arguing segmentation fault.

AFAIK segfaults are memory errors (where $program tries to read or write
memory that isn't his).

If there is a problem with the fs, fsck should complain and if possible fix
it, not just crash.

Try memtest86 (give it a few hours) and see if something comes up.

Mvg, Guus Snijders

Robert W0ng via arch-general

2017-05-14 14:56:32 UTC

Permalink

Hi, Guus.

I've just run the memtest from archiso. Just passed, no error reported. So I'm afraid it has to be an fs bug or other hardware error... I'm looking forward to running a SMART test later. But the current Arch installation is on a Kingstom USB drive, is that okay with the SMART test?

In fact, what confused me most now is why both pacman and fsck reported segmentation fault... It can't just be a coincidence, right?

Oh, remembered one more thing. I tried to remove the gnome-extra group again this afternoon. The same error, the same output. More specifically, even exactly the same time the error happened (when attemptting to remove package telepathy-idle). Funny enough, Uhm? ;)

Thx a lot!

RW

Post by Guus Snijders via arch-general
Op 14 mei 2017 04:29 schreef "Robert W0ng via arch-general" <
Hello again, All.
I ran fsck in archiso this morning, and found sth interesting.
To sum up, fsck.f2fs exit with signal 11, arguing segmentation fault.
AFAIK segfaults are memory errors (where $program tries to read or write
memory that isn't his).
If there is a problem with the fs, fsck should complain and if possible fix
it, not just crash.
Try memtest86 (give it a few hours) and see if something comes up.
Mvg, Guus Sn

Neven Sajko via arch-general

2017-05-16 17:32:38 UTC

Permalink

Did you notice the "kernel BUG at fs/f2fs/segment.c:992" line in the
kernel log? You should probably report this to the linux people. (But
I'm not sure how useful those traces are to them without the rest of
the line numbers...)

On another note, how long did you run memtest, I think under five or
six passes usually is not considered comprehensive enough.

Robert W0ng via arch-general

2017-05-17 13:35:57 UTC

Permalink

It would be my honor to do it if they accept image file............ :(

OMG!? One pass is not enough!? Oh, look at the SILLY thing I've done!!

BTW, switched to BTRFS last night. It seemed that the new file system agrees more with my Kingston drive than F2FS. :P

RW

Post by Neven Sajko via arch-general
Did you notice the "kernel BUG at fs/f2fs/segment.c:992" line in the
kernel log? You should probably report this to the linux people. (But
I'm not sure how useful those traces are to them without the rest of
the line numbers...)
On another note, how long did you run memtest, I think under five or
six passes usually is not considered compre

Bartłomiej Piotrowski

2017-05-17 13:37:32 UTC

Permalink

Post by Robert W0ng via arch-general
It would be my honor to do it if they accept image file............ :(
OMG!? One pass is not enough!? Oh, look at the SILLY thing I've done!!
BTW, switched to BTRFS last night. It seemed that the new file system agrees more with my Kingston drive than F2FS. :P
RW

Can you do something with your e-mail client so it stops breaking
threading every time you send a message?

Thanks,
Bartłomiej

Robert W0ng via arch-general

2017-05-17 13:56:36 UTC

Permalink

Terribly sorry... I'll pay attention to it next time.

;)