(Originally reported by Hitachi) Test machine: 8wayIA-64 Bug:Adaptec SCSI driver Reason: The new version 6.08BETA of the driver is used in Fisher,actually in this version, since distributed commands are too many,buffer(SW IO TLB) of kernel became not enough and kernel got panic. Soultions: Change the default number of distributed commands from 254 to 3. The patch is attached.
Created attachment 10540 [details] kernel panic when access to scsi
This defect is considered MUST-FIX for Florence Release-Candidate #2
Please test with wolverine kernel and both aic7xxx and aic7xxx_new drivers and report any problems with either driver.
understood.
Excuse me,which version do you mean?There seems no aicxxx_new included in Wolverine version. bash-2.04$ rpm -qlp kernel* | grep aic /lib/modules/2.4.1-0.1.9/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9BOOT/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9enterprise/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9smp/kernel/drivers/scsi/aic7xxx.o /lib/modules/2.4.1-0.1.9smp/kernel/drivers/scsi/aic7xxx.o bash-2.04$ pwd /mnt/test/wolverine/i386/RedHat/RPMS and the test platform is IA64.....
Sorry,you mean in rawhide..
The patch to change the aic7xxx queue depth is the *wrong* thing to do here. You need to fix the core ia64 kernel code, the aic7xxx driver isn't doing anything wrong. If needed, someone should take this problem to the ia64 mailing list and have the kernel people there write a proper fix. I'm closing this bug out since it isn't an aic7xxx problem, someone should open a new bug against the ia64 kernel that the DMA mapping code uses an insufficient map table size and doesn't dynamically expand it as needed.