On Thu, 21 Jan 1999, you wrote:
>btw, what PAGE_OFFSET setting do you use? (the default?)
I decided to upgrade my kernel again after receiving replies from you and
Dave Miller to 2.2.0-pre8. I made one change to this kernel and that was
to patch the de4x5 driver to not do any EISA probes (since they kill the
system badly and yes, I have notified the author about this problem).
I then built 4 different kernels with different values for PAGE_OFFSET.
Default (0xC0000000) results in no problems.
Changed to the number (0x70000000) as directed in the comments in page.h
resulted in a kernel panic. Here is an exerpt of the error message:
Unable to handle kernel paging request at virtual address fffc39e2.
Current->tss.cr3 = 00101000, %cr3 = 00101000
*pde = 00257067
*pte = 00000000
Oops: 0000
CPU: 1
EIP: 0010:[<701cd309>]
>From the System.map, this cooresponds to:
701cd2ec t BusLogic_ScanIncomingMailboxes
701cd364 t BusLogic_ProcessCompletedCCBs
and then a final message
Aiee, killing interrupt handler
Kernel panic: Attempted to kill the idle task!
In interrupt handler - not syncing
I then change the PAGE_OFFSET to 0x7C000000 like in the 2.0.36 big
RAM patch from Leonard Zubkoff with the same result except the EIP
was
EIP: 0010:[<7c1cd309>]
Again, from the System.map:
7c1cd2ec t BusLogic_ScanIncomingMailboxes
7c1cd364 t BusLogic_ProcessCompletedCCBs
Finally, I changed the PAGE_OFFSET to 0x90000000 and was able to
boot, but the test I described earlier resulted in the same file
system corruption.
I will try the same tests with 2.2.0-pre9 later today.
Tom
>
>On Wed, 20 Jan 1999, Thomas Schenk wrote:
>
>> We have been testing a 2.2.0 kernel on a system with 2 Gigs of RAM and are
>> seeing some serious problems. The system in question is a Dell PowerEdge
>> 6300/400. This is a quad Xeon 400Mhz system with 2Gigs of RAM, two
>> AIC 7890 SCSI controllers, one AIC 7860 SCSI controller and 4 9 Gig disks.
>> We have eliminated the SCSI controllers as the source of our problem by
>> deactivating them and using a BusLogic BT958 controller instead. Here is
>> the problem. When we do something that causes huge amounts of RAM to be
>> used as buffers (750 Meg or more), the filesystems get corrupted badly and
>> network cards stop working (the NIC is a DEC 500-BA). We have also ruled
>> out the network driver as the source of the trouble since the problem
>> occurs using either the de4x5 driver or the tulip driver.
--
+-----------------------+-------------------------+------------------------+
| Tom Schenk | Use Linux! | All opinions expressed |
| [EMAIL PROTECTED] | Friends don't let | are mine and not those |
| [EMAIL PROTECTED] | friends do Windows! | of my employer. |
+-----------------------+-------------------------+------------------------+
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
