RE: [RE]Initrd boot phase errors reading bogus partition on ATA raiddrive

Thank you so much!
It took me half a day to go from uber-newbie to being able to run the new
image ( I've never done the patch -> rpmbuild thing ).
Strangely enough though, this patch only dramatically REDUCED the number of 
"Buffer I/O error on device hde1, logical block BLA-BLA-BLA" errors, it did
not eliminate them.
So,now I'm more than tempted to go into nash and turn off all attempts to
access hde, sil60, or whatever I have to in order to get it to boot quickly
and quietly. After all, on this machine,  I know exactly where the root
filesystem is.
 
Any advice?

Charlweed
________________________________

From: James Olson [mailto:[email protected]] 
Sent: Thursday, October 26, 2006 6:39 PM
To: ATARAID (eg, Promise Fasttrak, Highpoint 370) related discussions
Subject: [RE]Initrd boot phase errors reading bogus partition on ATA
raiddrive


I had a similar problem some months ago. It was caused by the redhat nash
program's mount command in the initrd probing drives when it shouldn't (like
when you mount the /proc filesystem). I wrote a patch to the nash source
code to fix it on my system. 

# diff -Naur block.c.orig block.c 
--- block.c.orig 2006-03-08 11:46:59.000000000 -0800 
+++ block.c 2006-03-30 02:49:19.000000000 -0800 
@@ -337,6 +337,23 @@ 
return NULL; 
} 

+static char * 
+block_populate_cache() 
+{ 
+ bdev_iter biter; 
+ bdev dev = NULL; 
+ blkid_dev bdev = NULL; 
+ 
+ biter = block_sysfs_iterate_begin("/sys/block"); 
+ while(block_sysfs_next(biter, &dev) >= 0) { 
+ 
+ bdev = blkid_get_dev(cache, dev->dev_path, BLKID_DEV_FIND); 
+ } 
+ block_sysfs_iterate_end(&biter); 
+ 
+ return NULL; 
+} 
+ 
char * 
block_find_fs_by_label(const char *label) 
{ 
@@ -356,7 +373,7 @@ 

if (!access("/sys/block", F_OK)) { 
/* populate the whole cache */ 
- block_find_fs_by_keyvalue("unlikely","unlikely"); 
+ block_populate_cache(); 

/* now look our device up */ 
bdev = blkid_get_dev(cache, name, BLKID_DEV_NORMAL); 


---------[ Received Mail Content ]---------- 
>Subject : Initrd boot phase errors reading bogus partition on ATA raid
drive 
>Date : Thu, 26 Oct 2006 15:19:01 -0700 
>From : "charlweed" 
>To : 
> 
>Hi gurus! 
> 
>Booting my Fedora 5 (2.6.18-1.2200.fc5) system takes an extra couple of 
>minutes because of disk errors. The system is trying to "do something" with

>a partition on a drive that is part of a raid set, and failing. These
errors 
>occur during the initrd boot phase. After the system boot, the system is 
>apparently ok. 
> 
>When I boot Linux, I get several screens worth of the following error 
> 
> Buffer I/O error on device hde1, logical block 625153152 
> 
> 
> 
>hde is part of a 2 disk striped raid set. I can stop the errors if I add 
>"hde=noprobe hdf=noprobe" as a kernel boot parameter, but then I cannot use

>dmraid to access my raid partition, because /dev/hde & /dev/hde don't
exist. 
>I tried adding boot parameter "hde=19457,255,63" but the device remains 
>invisible. 
> 
> 
> 
>My naive guess is that I can solve my problem by either 
> 
>1) Getting dmraid to see my drives after a boot that uses "noprobe" 
> 
>2) Stopping whatever program is trying to access hde1 during the initrd 
>boot. 
> 
>All my attempts at 1) have failed, and I have no idea how to do 2). 
> 
> 
> 
>My System has 
> 
> An Abit AN7 nforce chipset motherboard with 
> 
> 2 ATA onboard channels 
> 
> onboard Si3112 SATA Raid ( I use SATA, but not the SATA raid) 
> 
> A SiI0680 (CMD 680) pci ATA card 
> 
> 
> 
>The drive layout is 
> 
> hda onboard ata [hda1=/boot ,ext3 : hda2=unmounted, ntfs] 
> 
> hdc onboard ata [ hdc1= /, ext3 : hdc2=swap : hdc3=/var, ext3] 
> 
>** hde SiI0680 striped raid set 1 
> 
> hdf SiI0680 [ hdf1=unmounted, ntfs ] 
> 
>** hdg SiI0680 striped raid set 1 
> 
> sda onboard SATA [sda1=unmounted, ntfs] 
> 
>** dm-0 raid 1[dm-0p1=unmounted, ntfs] 
> 
>Thanks! 
> 
>Charlweed 
> 
> 
> 
>