> Hi,
>
> On 13 Sep 2007, at 21:35, mar...@es... wrote:
>> When copying large amounts of data (> 25GB) from an NTFS mount, my
>> kernel
>> panics.
>
> Ouch. Would it be possible for you to write down and email the
> actual panic text? (Or capture it via serial console if you don't
> want to type it by hand...)
I guess I may have stated that wrong, I don't think the ntfs driver is
making the kernel panic directly, I think it is using up all of the memory
and then when another process tries to use some memory, that invokes
oom-killer. It doesn't seem like it always makes the kernel panic, but I
think it is init that invokes oom-killer because there is no more memory:
init invoked oom-killer
Is that correct?
Here is the message I get when the kernel panics:
Kernel panic - not syncing: Out of memory and no killable processes.
..
Kernel panic - not syncing: Out of memory and no killable processes.
>
>> escabot log # uname -a
>> Linux escabot 2.6.21-gentoo-r4 #3 SMP PREEMPT Thu Sep 13 14:44:12
>> EDT 2007
>> i686 AMD Opteron(tm) Processor 280 AuthenticAMD GNU/Linux
>
> Out of curiosity. Why are you running a 32-bit kernel on such a
> large 64-bit machine?!? Your machine would run a lot quicker with a
> 64-bit kernel (and 64-bit user space)...
>
I really don't have a good reason, I probably should be. To be honest, I
kind of forgot this was an Opteron until you mentioned that. =)
>> I watch top while copying the files and it just sucks up all of the
>> memory. It gets to a point for a while where it seems to copy fine
>> leaving
>> just a few megabytes of memory free. If I copy smaller sets of
>> data, it
>> seems to work fine (but still sucks all the memory it can). This
>> machine
>> has 12GB of memory.
>>
>> I have tried several times and I can pretty much get it to always
>> dump if
>> I copy enough stuff over. The last time I recompiled the kernel
>> with NTFS
>> debug support and enabled the debugging messages. I am attaching a
>> gzip'd
>> file of the kernel messages.
>
> Great, thanks.
>
>> If attaching a file and sending it to the mailing list is improper
>> etiquette, let me know and I can post it on a web page or do something
>> else.
>
> No that is fine!
>
>> I'm not sure if this is a bug or something I'm doing wrong.
>>
>> Thank you for your time.
>
> The kernel panic()ing is always a kernel bug!
>
> The question is what are you copying to?
>
I was copying to an ext3 file system.
> You said you are copying those large files from NTFS but never said
> what file system is the destination...
>
> To see whether it is NTFS at fault (I don't think it is, I think it
> is either your destination file system or a bug in the VFS/VM not a
> in a file system at all) try to copy to /dev/null, i.e. if you are
> copying using "cp" command then do:
>
> cp /path/to/ntfs/volume/huge26gbfile /dev/null
>
> If you use "dd" then do:
>
> dd if=/path/to/ntfs/volume/huge26gbfile of=/dev/null [insert your
> usual dd options here]
>
> Also out of interest what command are you exactly running to do the
> copying?
>
Before trying /dev/null just a: cp -R IMAGES /root
> That will effectively read in the whole file but throw what it reads
> away instead of writing it to another file system.
>
> Does this work or also panic?
>
escabot CAD # ls -ltrh
total 49G
-r-------- 1 root root 2.0G Oct 7 2004 RTC26001.GHS
-r-------- 1 root root 2.0G Oct 7 2004 RTC26002.GHS
-r-------- 1 root root 2.0G Oct 7 2004 RTC2604.GHO
-r-------- 1 root root 620M Oct 7 2004 RTC26003.GHS
-r-------- 1 root root 4 Sep 14 09:32 asdf.txt
-r-------- 1 root root 8 Sep 14 09:33 rty.txt
-r-------- 1 root root 42G Sep 14 10:36 bigfile.gho
escabot CAD # cp bigfile.gho /dev/null
> My expectation is that it will work fine but there is only one way to
> find out and that is for you to try it! (-:
>
> Best regards,
>
> Anton
> --
> Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
> Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
> Linux NTFS maintainer, http://www.linux-ntfs.org/
>
Thank you for your time and help. I look forward to hearing from you soon.
--Marc
>
|