Hello,

I guess a bogus patch went in the standard kernel somewhere between 
2.6.3-mm1 and 2.6.3-mm3.

The device mapper tree (udm?) use to package the mm patchset. Starting 
with 2.6.2, I've had to reverse the mm? patch from udm? not to hit the 
following bug. 2.6.3-mm3 has the culprit in linus.patch.

The attached capture is all I can gather.

It happens sometimes during "modprobe sd_mod" after a successful 
"modprobe qla2200". Anyway, a SG inquiry triggers it  reliably. Of 
notice, there are failed device (StorageWorks' ghosts devs) detected by 
scsi_scan in REPORT_LUN mode.

Hope it helps,
cvaroqui