Release Engineering

Tasklist

FS#13381 - [archiso] Kernel null pointer reference error when passing malformed arguments

Attached to Project: Release Engineering
Opened by Ole Rasmussen (elumineX) - Thursday, 19 February 2009, 18:48 GMT
Last edited by Aaron Griffin (phrakture) - Tuesday, 01 December 2009, 22:29 GMT
Task Type Bug Report
Category ArchISO
Status Closed
Assigned To Tobias Powalowski (tpowa)
Aaron Griffin (phrakture)
Thomas Bächler (brain0)
Gerhard Brauer (GerBra)
Dieter Plaetinck (Dieter_be)
Architecture All
Severity Low
Priority Normal
Reported Version 2009.02
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
Kernel null pointer reference error at Arch Linux 2009.02 x86_64 installation (with vga="773") boot parameter in VMWare Workstation v6.5.1 (b. 126130).
This error does not happen without the vga parameter.

Additional info:
* Screenshot of the error attatched

Steps to reproduce:
1) Boot the FTP cd
2) Press [e] to edit boot parameters
3) Press [e] again to edit kernel line and append 'vga=773' as proposed in the Beginners guide.
4) Press [enter] then [b]
6) When ready to login, type 'root' [enter]
7) Kernel error....
This task depends upon

Closed by  Aaron Griffin (phrakture)
Tuesday, 01 December 2009, 22:29 GMT
Reason for closing:  Fixed
Additional comments about closing:  Fixed in git. Shows an error message. git commit 2b999307
Comment by Ole Rasmussen (elumineX) - Thursday, 19 February 2009, 18:50 GMT
Screenshot of error attached (retrying)
   arch_kernel_error.jpg (280.8 KiB)
Comment by Ole Rasmussen (elumineX) - Friday, 20 February 2009, 20:41 GMT
The error happens on my laptop as well so it doesn't seem to be an issue related to VMWare but generally the vga=773 parameter.
Comment by Ole Rasmussen (elumineX) - Monday, 23 February 2009, 22:18 GMT
False alarm.
It seems I actually misunderstood the last setting int the boot params (with the %). I accidently added vga=773 before the percentage meaning it was vga=773%.

Adding the setting properly works like a charm!
Comment by Aaron Griffin (phrakture) - Monday, 23 February 2009, 22:34 GMT
  • Field changed: Attached to Project (Arch Linux → Release Engineering)
Not sure if we want to close this...
I imagine it actually ran "ramdisk_size=75 vga=773%" - we should check to see if there's any way to fix this
Comment by Johannes Jordan (FoPref) - Friday, 27 February 2009, 16:02 GMT
This should also be reported to upstream, kernel should be able to parse it or give error message instead of crash like this.
Comment by Aaron Griffin (phrakture) - Friday, 27 February 2009, 20:50 GMT
It might not be kernel related - it might by our script parsing the ramdisk_size param incorrectly
Comment by Dieter Plaetinck (Dieter_be) - Friday, 13 March 2009, 18:03 GMT
Okay, it may be a kernel bug, but it also could be a bug in our scripts like Aaron said. I assume this is about the early userspace tools (initcpio hooks and things), so I'll move this ticket to the Arch Linux project. (and lower the severity)
Comment by Dieter Plaetinck (Dieter_be) - Friday, 15 May 2009, 07:22 GMT
  • Field changed: Percent Complete (100% → 0%)
could you please elaborate why you think this is not a bug?

it may be an upstream bug, but that doesn't mean it's not a bug we can track.
Comment by Ole Rasmussen (elumineX) - Sunday, 24 May 2009, 08:57 GMT
You're probably right. I was too fast on the close button. It would of course be much better to report an error instead of a kernel crash.
Comment by Gerardo Exequiel Pozzi (djgera) - Sunday, 26 July 2009, 15:43 GMT
The problem here is the space reserved for ramdisk_size that is _too small_ (75Kb), and not the "%" in the vga parameter. And when tmpfs is mounted with the size=75 create a tmpfs of 4 Kbytes.
You can see that when booting there are at least one message saying out of disk space. The crash was with unionfs, now with aufs2 the crash is a kernel panic instead of null pointer dereference.
I think that we can do nothing here other than message saying: "Hey guy the ramdisk is too small!" ;)

@Aaron: One sidenote here, as you can see, the "ramdisk_size" when the size is without percent is in Kbytes, but the "size" parameter for tmpfs is in bytes.
   ramdisk_size.too-small.txt (1 KiB)
Comment by Gerardo Exequiel Pozzi (djgera) - Monday, 30 November 2009, 22:33 GMT
any decision on this?

I don't know if can be really useful testing if tmpfs size is too small to avoid a kernel panic -> show a message, then stop. This is like booting on system with a small memory.

The check should be against parameter passed to tmpfs_size (ex ramdisk_size) it can be in bytes, [k]bytes, [m]bytes, [g]bytes or % of ram, and comparing with some minimal permited. Or check if mounting the aufs root fails because can't write to tmpfs. I think that the first option will make much more complicated the archiso hook, second option maybe aceptable...

opinions?
Comment by Gerardo Exequiel Pozzi (djgera) - Monday, 30 November 2009, 23:15 GMT
attached a patch that check exit status of mount
http://github.com/djgera/archiso/commit/90e0a728fc95441eebf291ed7c84e184f548224a
   0003-Add-an-error-message-whe... (1.3 KiB)

Loading...