This could be anything from a renegade module corrupting kernel memory, to
bad memory chips.
If you want to rule out the latter, get memtest86
(http://www.teresaudio.com/memtest86/) and run it for a few hours on your
system. A few times in the past problems I attributed to faulty code were
actually due to bad memory chips.
-Calin
On Thu, 13 Sep 2001, Stephen Bridges wrote:
> I am using multiple PHANToMs on a 2.4.4-rtl kernel and rtlinux 3.1.
> By using lots of threads and slightly changing kabi's module source I
> have got them working together satisfactorally.
>
> However there seems to be some memory leak in the system which seems
> to be linked to either the module or my programs.
>
> I get errors such as:
>
> Unable to handle kernel paging request at virtual address c997f280
> printing eip:
> c997f280
> *pde = 018b9067
> *pte = 00000000
> Oops: 0000
> CPU: 0
> EIP: 0010:[scsi_mod:proc_scsi_R79a04975+85235276/72222440]
> EIP: 0010:[<c997f280>]
> EFLAGS: 00010282
> eax: c047cedc ebx: c2cbceb4 ecx: c0478000 edx: c997f280
> esi: c116bf74 edi: c116a558 ebp: c116a000 esp: c116bf6c
> ds: 0018 es: 0018 ss: 0018
> Process keventd (pid: 2, stackpage=c116b000)
> Stack: c011509a c0478000 c047cedc c2cbceb4 00000296 00000000 00000000
> c011c0a1
> c0210e20 0000000a c116a550 c116a560 c116bfd0 c02525d8 00000001
> c0114d4b
> c0114d59 00000001 00000000 00000000 00010000 00000000 00000700
> c011bf80
> Call Trace: [__run_task_queue+90/112] [context_thread+289/432]
> [do_softirq+91/128] [do_softirq+105/128] [context_thread+0/432]
> [init+0/320] [kernel_thread+38/48]
> Call Trace: [<c011509a>] [<c011c0a1>] [<c0114d4b>] [<c0114d59>]
> [<c011bf80>] [<c0105000>] [<c01054d6>]
> [context_thread+0/432]
> [<c011bf80>]
>
> Code: Bad EIP value.
>
> Now if the computer starts these types of crashes and is not rebooted
> then certain programs will stop working producing different crashes
> but still the same type at the same address, although this address
> will change after rebooting.
>
> Is this a memory leak? If it isn't, what is it? The process shown
> crashing here is keventd which is normally the most likely although
> others have crashed as well. Normally make, ls, grep, and anything
> that uses lots of memory.
>
> I am going to try to debug this by getting my programs to log every
> variable they create and search these logs on crashes, and probably do
> a similar thing with the module and use printk.
>
> Would people recommend this or a different approach?
>
> Finally, I'd like to mention the people of New York, in what must be a
> difficult time for them. My thoughts go out.
>
> Stephen
>
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
--
For more information on Real-Time Linux see:
http://www.rtlinux.org/
