Bug 111783 - cannot boot vmlinuz-2.4.9-e.31summit on x440
cannot boot vmlinuz-2.4.9-e.31summit on x440
Status: CLOSED CURRENTRELEASE
Product: Red Hat Enterprise Linux 2.1
Classification: Red Hat
Component: kernel (Show other bugs)
2.1
i686 Linux
medium Severity high
: ---
: ---
Assigned To: Jim Paradis
Brian Brock
:
Depends On:
Blocks: 107566
  Show dependency treegraph
 
Reported: 2003-12-09 21:12 EST by keith mannth
Modified: 2013-08-05 21:03 EDT (History)
3 users (show)

See Also:
Fixed In Version: RHEL 4 U4
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Environment:
Last Closed: 2006-08-21 10:32:17 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---


Attachments (Terms of Use)
var/log/messages (109.81 KB, text/plain)
2003-12-09 21:19 EST, keith mannth
no flags Details
a patch to fix irq affinity for summint and the e.35 summit kernel (1.22 KB, patch)
2004-02-09 18:10 EST, keith mannth
no flags Details | Diff

  None (edit)
Description keith mannth 2003-12-09 21:12:45 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.2)
Gecko/20030716

Description of problem:
This is bug is against AS2.1 Update 3 beta1 (As far as I now beta I)
The summit kernel will not boot on my system. It is an 4-way x440 with
under 4 gig on memory.

This disto installed ok but the summit kennel will not boot.  It gets
as far as bring up sshd and xinitd but it will not pass this.  I am
still able to ping to box but I can't ssh into it.  The key board is
non responsive.  The last kernel I had access to e.29summit booted ok
but was not stable on the system (it fell down after about an hour of
stress test).  

I booted the up kernel but was unable to bulid a debug kernel (thanks
alot for including kdb in you kenel tree).  When I did a make
modules_install i get this mess. (after bzImage and modules)

[root@elm3a2 linux-2.4]# make modules_install
make -C  kernel modules_install
make[1]: Entering directory `/usr/src/linux-2.4.9-e.31/kernel'
make[1]: Nothing to be done for `modules_install'.
make[1]: Leaving directory `/usr/src/linux-2.4.9-e.31/kernel'
mkdir -p  /lib/kernel/2.4.9-e.31custom/
install -m 755 ulib/libredhat-kernel.so.1.0.1
/lib/kernel/2.4.9-e.31custom/
install: cannot stat `ulib/libredhat-kernel.so.1.0.1': No such file or
directorymake: *** [_modinst_kernel] Error 1
[root@elm3a2 linux-2.4]#

so I am a little stuck for more info at this time.  I need to build
kernels.  

That is wher I am at with the AS2.1 beta 1 release. 

Version-Release number of selected component (if applicable):
kernel-2.4.9-e.31

How reproducible:
Always

Steps to Reproduce:
1.Load AS2.1 Update 3 beta 1 
2.Boot summit kernel
3.Observe the box not boot. 
    

Actual Results:  Box hangs while bringing up xinitd. 

Expected Results:  box should boot. 

Additional info:

I will attach output files to this bug
Comment 1 keith mannth 2003-12-09 21:19:40 EST
Created attachment 96438 [details]
var/log/messages

This is the /var/log/messages from the system.	The last kenel to boot is the
up kernel.  There are 3-4 bad boots in this log.
Comment 2 Arjan van de Ven 2003-12-10 13:18:12 EST
does this happen if you chkconfig --level 345 irqbalance off
?

if so, then the kernel is bust in the irq affinity settings on summit;
something that we almost never used before U3
Comment 3 keith mannth 2003-12-10 14:58:31 EST
If I run without the irqbalance off I am able to boot just file.  
Does the irqbalance demon use the /proc irqaffinity?   If so there may
be issues for clusterd apic boxes.  I fixed this in 2.5 a while ago
but I don't have a patch for this 2.4.9 kernel tree.  
Comment 4 keith mannth 2003-12-10 16:31:16 EST
Sorry just read my comment.  It should read.

If I run with irqbalance off I am able to boot just fine. 
....
Comment 5 Mike McLean 2003-12-10 18:20:10 EST
the e.34 kernel works on the x440 in the test lab
Comment 6 keith mannth 2003-12-10 18:39:47 EST
  Where can I get the e.34 kernel?
Comment 7 Jason Baron 2003-12-11 10:36:36 EST
http://people.redhat.com/~jbaron/.private/testing/2.4.9-e.34/

we ended up making writes to smp_affinity be noops. any testing is
appreciated. We also haven't seen any stability issues as hinted at
above. 

Comment 8 keith mannth 2003-12-11 13:17:21 EST
  I'll give this kenel some testing and let you know if I have
problems.  Would you take a patch to fix clusered apic boxes and irq
affinity?  I did this work for v2.6 I should be able to fix it here as
well. 
Comment 9 Jason Baron 2003-12-11 14:03:10 EST
definitely, we can queue those for QU4.
Comment 10 Jason Baron 2003-12-11 14:06:09 EST
or should say Update 4.
Comment 13 Jason Baron 2004-01-29 10:58:30 EST
Keith, do you have a patch for this for U4?
Comment 14 keith mannth 2004-02-09 18:10:31 EST
Created attachment 97544 [details]
a patch to fix irq affinity for summint and the e.35 summit kernel
Comment 15 keith mannth 2004-02-09 18:14:51 EST
Sorry for the delay.  This fixes the clusterd apic for the summit
kernel.  It is a simple fix.  Apics in v2.4 and summit boxes can only
only address one cpu at a time. You can only set affinity to one cpu
at a time.  
  The current irq deamon tries to mask to multiple cpus at a time. 
With summit this maps to getting masked to the first cpu in the mask.
 It looks like cpus (with you current irqdeamon) are only getting
mapped to 1/2 the cpus.
  Let me know what you think.

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