Hello, I was trying to test some failover and reliability features, and had instrumented the iscsi low level driver to simulate failures when some one tries to read the disk. I observed that* iscsi disk is discovered and sd_probe_async() is called that ultimately tries to read the partition table using READ10 command. Now due to simulation of failures, the iscsi LLD never completes the command which results in a timeout. The iscsi driver provides an eh_timed_out() function, which is called and decides what can be done about it. Now, it seems to me that an LLD (that has provided eh_timed_out() func ptr, such as iscsi) , can impact all the other SCSI disks if it keeps on returning BLK_EH_RESET_TIMER from its timeout handler. Specifically, no SCSI disk (even handled by other LLDs) in the system can be removed. I tried removing multiple disks and pretty soon ended up with multiple threads deadlocked, because sd_remove() for the "good disks" is waiting on async_synchronize_full_domain(&scsi_sd_probe_domain). This never completes because the sd_probe_async() for the "bad disk" is trying to read the parition table, and the "bad" LLD keeps on timing out, always asking for more time (BLK_EH_RESET_TIMER). My question is have I identified this problem correctly, or if I am missing something here? Thanks, Rajat -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
