On 04/10/2012 11:37 AM, Mike Christie wrote: > On 04/10/2012 03:16 AM, Bart Van Assche wrote: >> On 04/10/12 01:22, Elric Fu wrote: >> >>> After debugging the code, I found the issue happened while the driver ran to >>> line 782 in scsi_send_eh_cmnd(). >>> >>> 778 static int scsi_send_eh_cmnd(struct scsi_cmnd *scmd, unsigned char *cmnd, >>> 779 int cmnd_size, int timeout, unsigned >>> sense_bytes) >>> 780 { >>> 781 struct scsi_device *sdev = scmd->device; >>> 782 struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd); >>> 783 struct Scsi_Host *shost = sdev->host; >>> 784 DECLARE_COMPLETION_ONSTACK(done); >>> 785 unsigned long timeleft; >>> 786 struct scsi_eh_save ses; >>> 787 int rtn; >>> >>> I know the code is submitted by you. I don't familiar with the scsi core. >>> It seems like the conversion process from scsi command to scsi driver >>> encounter a NULL pointer. Any idea? >> >> >> I have observed crashes at the same point while testing device removal >> with the ib_srp driver. As far as I can see that code was added through >> commit 18a4d0a22ed6c54b67af7718c305cd010f09ddf8 (February 9, 2012). The >> approach of that patch looks questionable to me: what guarantees that >> the struct scsi_driver will be available at the time the SCSI error >> handler needs it ? > > If a scsi scan IO timesout then the driver will not be set yet. It is ok > to use scsi_cmd_to_driver in scsi_finish_cmd because of the req block pc > check (scsi scan IO set that flag and so do not hit that path). I meant to say that all REQ_TYPE_BLOCK_PC will have a NULL driver, so at the very least there should be a check in scsi_send_eh_cmnd for NULL sdrv or for the IO being a REQ_TYPE_BLOCK_PC like is done in scsi_finish_command. -- 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