From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.6) Gecko/20050324 Firefox/1.0.2 Red Hat/1.0.2-1.4.1.centos4 Description of problem: There is a bug in /etc/rc.sysinit script. The regular expression in awk statement will catch only first SCSI controler in the system (as returned by modprobe -c). The fix is trivial, and I'll include the patch as attachment. Usually this is not a big issue since device drivers for all SCSI controlers (regardless if they are needed to mount root file system or not) will be loaded from initrd image anyhow earlier in the boot process. Anyhow, might be worth fixing regardless of that. BTW, I've also made a bug report for mkinitrd (bug 154857), suggesting it should load sg driver too if there are SCSI devices in the system. Alternative would be if sg is loaded from rc.sysinit from "initialize hardware / storage" section (would be trivial to implement). Version-Release number of selected component (if applicable): initscripts-7.93.11.EL-1.centos4 How reproducible: Always Steps to Reproduce: 1. On system with > 1 SCSI contoler execute: /sbin/modprobe -c | \ awk '/^alias[[:space:]]+scsi_hostadapter[[:space:]]/ { print $3 }' It'll only return first controler. Additional info:
Created attachment 113176 [details] check for all scsi_hostadapter lines
This is fixed in 7.93.12.EL-1. *** This bug has been marked as a duplicate of 145432 ***
Ah, I searched only under RHEL tree... Anyhow, what is your opinion about loading sg device driver from rc.sysinit script if SCSI host adapters are detected (the additional question from the end of bug report)?
sg should be loaded by the hotplug system (see /etc/hotplug/scsi.agent) when necessary.
Hm, well, it isn't... And looking at scsi.agent it seems as it should. I'll look around a bit. Should I report a bug against hotplug system?
What device type? Also, make sure you don't have the version of scsi.agent with the typo (MODULE vs. MOUDLE, or something like that...)
Yup, it is the one with a typo in the line for SCSI type 8 (changer). Which is my device (changer in tape library). I guess when I fix the typo, module will be automatically loaded on boot. And looking at it more closely, I see possible typo in line that initializes TYPE_ATTR (unless DEVPATH is guaranteed to start with "/").
The hotplug typo is fixed in 2004_04_01-7.5. That and the initscripts fix should both be in U1. I believe DEVPATH is always propagated by the kernel starting with '/'/