Bug 682079 - Request for add "--omit-cciss-modules" to mkdumprd
Summary: Request for add "--omit-cciss-modules" to mkdumprd
Keywords:
Status: CLOSED WORKSFORME
Alias: None
Product: Red Hat Enterprise Linux 6
Classification: Red Hat
Component: kexec-tools
Version: 6.0
Hardware: All
OS: Linux
unspecified
medium
Target Milestone: rc
: ---
Assignee: Cong Wang
QA Contact: Kernel Dump QE
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-03-04 02:58 UTC by masanari iida
Modified: 2013-09-30 02:22 UTC (History)
2 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2011-03-14 04:36:33 UTC
Target Upstream Version:


Attachments (Terms of Use)
kdump.conf with blacklist option test result. Console output. (5.86 KB, text/plain)
2011-03-07 10:36 UTC, masanari iida
no flags Details

Description masanari iida 2011-03-04 02:58:04 UTC
Description of problem:

(1) When we configure kdump for network (SSH or NFS) on HP Proliant, 
we don't need to load cciss or hpsa driver. 
Currently some type of SmartArray failed to reset device and it hung 
for ever. 
So I don't want to load unnecessary driver into the 2nd kernel.
I have tried --omit-scsi-modules, but cciss.ko is still in the img.

In order not to load cciss.ko and hpsa.ko in initrd-kdump.img, I proposed 
to add "--omit-cciss-modules" option to mkdumprd.


(2) Add mkdumprd_options into kdump startup script
In /etc/init.d/kdump  script, currently we can not specify additional 
options for mkdumprd.

For example, (from RHEL5)

                /sbin/mkdumprd -d -f $kdump_initrd $kdump_kver

I prefer to have $mkdumprd_options 

                $MKDUMPRD $mkdumprd_options -d -f $kdump_initrd $kdump_kver

Then I can specify kdump_options="--omit-cciss-modules".
Otherwise, I need to modify the script by myself.


Version-Release number of selected component (if applicable):
kexec-tools-2.0.0-145.el6.i686.rpm

Additional Information:
These 2 modifications are valid for RHEL5 as well.

Comment 2 Cong Wang 2011-03-04 03:12:09 UTC
What's the problem of using "blacklist" in kdump.conf? Can't you blacklist cciss modules in kdump.conf?

Comment 3 masanari iida 2011-03-04 04:02:20 UTC
Hi,
I never known kexec-tool have such an option.
In man page, kexec-kdump-howto.txt page and kdump.conf,
it is not documented.  
Could you add the info?

One more question.
Is this "blacklist" option available for kexec-tools on RHEL5?

Comment 4 Cong Wang 2011-03-04 05:35:57 UTC
(In reply to comment #3)
> Hi,
> I never known kexec-tool have such an option.
> In man page, kexec-kdump-howto.txt page and kdump.conf,
> it is not documented.  
> Could you add the info?

It is fixed in -159.

> 
> One more question.
> Is this "blacklist" option available for kexec-tools on RHEL5?

Hmm, probably not, but we can backport it to RHEL5.

Comment 5 masanari iida 2011-03-04 09:46:09 UTC
(In reply to comment #2)
> What's the problem of using "blacklist" in kdump.conf? Can't you blacklist
> cciss modules in kdump.conf?

Confirmed, the blacklist does what I want to do.
Thanks for your advice.

>> Could you add the info?
>
>It is fixed in -159.

OK.

>> One more question.
>> Is this "blacklist" option available for kexec-tools on RHEL5?
>
>Hmm, probably not, but we can backport it to RHEL5.

Please backport the function to RHEL5 before the Production 2 
support phase end.
We sometimes suffering the SmartArray hung durint the kdump symptom.
Use blacklist on RHEL5 may become our good workaround for this symptom.

Comment 6 masanari iida 2011-03-07 10:33:58 UTC
My colleague tested kdump over the network with "blacklist hpsa" 
on real hard ware. 
Then he found an issue with the option.
In short, the init script within initrd-kdump.img  still waiting
for _some_ block device to come up, so it doesn't start copying 
the memory image over the network.

Environment
OS: Kernel 2.6.32-71.el6
Driver:  hpsa   (SmartArray P410i is used)

== kdump.conf ==

net 192.168.1.50:/var/crash
path .
core_collector makedumpfile -c --message-level 1 -d 31
blacklist hpsa cciss

========

Detail: 
My colleage has tested the blacklist option in kdump.conf on RHEL6 (hpsa).
Then he found that even with the blacklist option, the  init script 
in initrd-kdump.img still waiting for a block device to come up.  
(See attached console output during the kdump) 
It seemed following part of the init script cause this symptom.

for i in `cat /etc/critical_disks | awk '{print $1}'`
do
    IDSTRING=`grep $i /etc/critical_disks | awk '{print $2}'`
    COUNT=`grep $i /etc/critical_disks | awk '{print $3}'`
    found=0

    echo -n "Waiting for $COUNT $i-like device(s)..."
    while true
    do

.....snip

    done
    echo Found
    found=0
done  

For your information,  contents of /etc/critical_disks (in initrd) is

# cat etc/critical_disks
sda HPLOGICALVOLUME3.660 1
#

Workaround:
In this case, my colleague had to specify blacklist and disk_timeout 
option in kdump.conf and these option made the script expired.  

Question:
So I come back to the starting point of this case.
As you know, my goal is, if I want to do kdump over the network, 
I don't want the 2nd kernel to recognize _any_ block devices. 
What is the rocommended setting?

Comment 7 masanari iida 2011-03-07 10:36:47 UTC
Created attachment 482642 [details]
kdump.conf with blacklist option test result. Console output.

kdump.conf with blacklist option test result.
Console output.

Comment 8 Cong Wang 2011-03-07 10:51:24 UTC
It is the fix for this problem, otherwise it will always wait for a disk which never comes up.

Comment 9 Cong Wang 2011-03-14 04:36:33 UTC
I am closing this bug as WORKSFORME. If you request to backport "blacklist" to RHEL5, please report a new bug against RHEL5.

Thanks.


Note You need to log in before you can comment on or make changes to this bug.