I am testing the attachment of larger storage systems to Linux hosts. Challenging the Linux SCSI error recovery capabilities we experience infinite reset loops which sometimes lead to system crashes. Trigger for the problem is a target reset of 20 attached volumes. We see the same problems also when we use scsiinfo to send unsupported commands to the subsystems. I am looking for someone willing to help solve this problem. It is serious and impacts the reliabilty of any Linux system with a large disk subsystem attached. Configuration: Server: IBM Netfinity 8500R (MachineType:8681 Model:5RY), 8 x Intel Xeon 550MHz, 1,5GByte RAM SCSI : Adaptec AHA-2944UW (firmware/bios: v2.20.0) Linux driver: aic7xxx We tried - Kernel 2.2.16 SMP, TCQ on, Queue depth between 8 and 32, 128 LUN support - Kernel 2.4.0-test10 - Kernel 2.2.18 - Kernel 2.2.16 SMP and Fibre Channel Attachment through QLogic 2200 (qlocicfc driver) All show the same behaviour. The problem seems to go away when we turn off TCQ. Kernel 2.4.0 with the new AIC7xxx driver version 6.08BETA is not running stable enough to test error recovery. Alexander Warmuth IBM Germany Tel. +49-6131-84-2887, Fax -6433 - To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [EMAIL PROTECTED]
