Bug 6695

Summary: Dell CERC ATA100/4ch not supported by any driver in 2.6.16
Product: IO/Storage Reporter: Torsten Krah (krah.tm)
Component: SCSIAssignee: io_scsi
Status: CLOSED CODE_FIX    
Severity: low CC: bunk, hare, jgarzik, protasnb, sandro.wefel, sumant.patro
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.16 Subsystem:
Regression: --- Bisected commit-id:
Attachments: patch file
megaraid_mbox-dell-cerc-support

Description Torsten Krah 2006-06-15 07:20:20 UTC
Most recent kernel where this bug did not occur: 2.6.15.6
Distribution: Debian
Hardware Environment: Dell Power Edge 1600SC
Software Environment:
Problem Description: The CERC Controller is not supported anymore by 2.6.16.

Steps to reproduce:compile 2.6.16.x and boot.

After searching for years i found a debian bug report for 2.6.16, its not really
old but is addressing some issue with megaraid controller, i got since 2.6.16 is
out.

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=369378

I did not found the patch on latest patchset for 2.6.17 ( maybe i've searched
not as good as i thought so excuse me if its already in there ) , so i would
love to see it in 2.6.17.

kind regards

Torsten
Comment 1 Torsten Krah 2006-12-06 07:34:04 UTC
Created attachment 9745 [details]
patch file

Still not done - i'll add the patch from the post, maybe someone is going to
integrate this.
Comment 2 Natalie Protasevich 2007-07-23 00:37:05 UTC
I don't see this patch integrated. I suppose you still have this problem with current kernel?
Thanks.
Comment 3 Torsten Krah 2007-07-23 00:47:19 UTC
Yes you guess right.
I still patch the kernel everytime with this patch, until someone may integrate this "small" change ;-)
Comment 4 Andrew Morton 2007-07-24 12:53:48 UTC
Thanks, I merged this into -mm.

I never knew this patch existed.  Please, submit patches via email in future?

It'll probably take some time to get this patch into kernel.org:
scsi patches are like that, I'm afraid.
Comment 5 Torsten Krah 2007-07-24 13:10:19 UTC
Oh ok, i will submit them via email in the future.
Thx merging it, i'll can wait that time :-)
Comment 6 Andrew Morton 2007-08-10 17:19:20 UTC
reopened and rejected.  Sumant saye to use the new driver.

If we add this ID to the old driver, the two could conflict.
Comment 7 Torsten Krah 2007-08-13 01:24:55 UTC
Umhhh, i'll take a look at the new driver - but i've tested the new driver once (since the old one did not work before patching it) and the new one didn't work either (did not found the raid array, although this was on 2.6.16 times, maybe something has changed till now).
I'll take a look for the new driver this week and will report if it does work or not.
Comment 8 Torsten Krah 2007-08-13 03:21:49 UTC
Hm - i did not tested it yet, but looking at the new mbox driver code (2.6.22.2) - there is no subsys id 0x0511.
I wonder how the megaraid_mbox driver should know, that it should be responsible for this device.
Sumant, can you please tell me where this id is already defined? 
thx.
Comment 9 Sumant Patro 2007-08-13 12:30:40 UTC
It is there in PCI table definition (megaraid_mbox):

         {
                 PCI_VENDOR_ID_AMI,
                 PCI_DEVICE_ID_AMI_MEGARAID3,
                 PCI_ANY_ID,
                 PCI_ANY_ID,
         }
Comment 10 Torsten Krah 2007-08-14 11:22:41 UTC
Hi. Thx.
I tried it a few seconds ago and its still the same like 2.6.16 - the new driver does not find my raid drive.
What informations do you need to get involved, as you told this should work with the new one? Its a productive machine and i can only do such things in the evening.
I'll reopen the bug until this is fixed, if its okay.
Comment 11 Torsten Krah 2007-08-22 01:04:11 UTC
Another alternative would be, removing the ID from the new driver and add it to the old one, because the old one works - thoughts?
Comment 12 Natalie Protasevich 2007-09-22 20:45:48 UTC
Sumant, seems like the logic is broken somewhere; is it just a matter of adding PCI_SUBSYS_ID_CERC_ATA100 as Torsten mentioned or is it more involved then that?
Thanks.
Comment 13 Hannes Reinecke 2007-10-31 07:40:51 UTC
No, it's actually handling of the 'random deletion' feature which is different from megaraid and megaraid_mbox.
megaraid shifts the target ID for logical values by 0x80 for I/O commands, and megaraid_mbox shifts the target ID for _all_ commands.
So the INQUIRY command is being sent to the wrong channel and no devices are detected.
Comment 14 Hannes Reinecke 2007-10-31 07:42:03 UTC
Created attachment 13360 [details]
megaraid_mbox-dell-cerc-support

Patch to disable 'random deletion' for Dell CERC with F/W >= 6.62.
Comment 15 Torsten Krah 2007-11-16 12:11:48 UTC
Patch works for me, thx.
Comment 16 Natalie Protasevich 2008-03-24 17:07:49 UTC
Was this patch submitted? Or was this problem was handled in megaraid some other way?
Comment 17 Torsten Krah 2008-03-25 02:32:47 UTC
Don't know.
I am still using this patch every time i update my kernel.
Don't know who should and where to submit this patch. Thought its enough to track it here.
Comment 18 Adrian Bunk 2008-05-03 10:35:26 UTC
fixed by commit 69cd39e94669e2994277a29249b6ef93b088ddbb (will be in 2.6.26)