From: mroos@xxxxxxxx Date: Thu, 16 Oct 2014 09:43:34 +0300 (EEST) > [ 158.730919] ESP: tgt[6] lun[0] scsi_cmd [ 12 00 00 00 24 00 ] > [ 158.799397] ESP: intr sreg[87] seqreg[00] sreg2[00] ireg[18] Target 6 responds, Bus Service + Function Done interrupt. Status register 1 indicates interrupt pending, phase is MESSAGE IN. > [ 158.866964] ESP: intr sreg[87] seqreg[02] sreg2[00] ireg[08] Another interrupt, just plain Function Done, we're getting the message-in byte. Phase is unchanged. > [ 158.934651] ESP: Got msgin byte 7 Receive MESSAGE_REJECT (0x7). > [ 158.974329] ESP: intr sreg[86] seqreg[02] sreg2[00] ireg[10] Bus Service interrupt, we've moved to Message Out phase. > [ 159.041950] ESP: Sending message [ 06 ] Sending ABORT_TASK_SET message. > [ 159.087894] ESP: intr sreg[80] seqreg[02] sreg2[00] ireg[20] Disconnect interrupt. > [ 159.155522] ESP: start data addr[c01f1100] len[36] write(0) > [ 159.222248] ESP: intr sreg[80] seqreg[02] sreg2[00] ireg[40] > [ 159.289891] ESP: unexpected IREG 40 We attempt some kind of data transfer (maybe for the status block) and the device gives us an illegal-command interrupt. More details in the event log that gets dumped: [ 161.208873] esp: esp0: ent[31] EVENT val[0d] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[0c] Check Phase [ 161.318255] esp: esp0: ent[0] EVENT val[06] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[0d] Message In [ 161.426599] esp: esp0: ent[1] CMD val[01] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[06] Give FIFO flush ESP command [ 161.532860] esp: esp0: ent[2] CMD val[10] sreg[87] seqreg[02] sreg2[00] ireg[18] ss[00] event[06] Give "transfer information" ESP command. [ 161.639125] esp: esp0: ent[3] CMD val[1a] sreg[87] seqreg[02] sreg2[00] ireg[08] ss[00] event[06] Give "Set ATN" command to move to message-out to give the chip the ABORT message. [ 161.745386] esp: esp0: ent[4] CMD val[12] sreg[87] seqreg[02] sreg2[00] ireg[08] ss[00] event[06] Give "Message OK" command to ACK the message-in from the device. [ 161.851643] esp: esp0: ent[5] EVENT val[0d] sreg[87] seqreg[02] sreg2[00] ireg[08] ss[00] event[06] Check Phase. [ 161.959990] esp: esp0: ent[6] EVENT val[09] sreg[86] seqreg[02] sreg2[00] ireg[10] ss[00] event[0d] Message Out. [ 162.068332] esp: esp0: ent[7] CMD val[01] sreg[86] seqreg[02] sreg2[00] ireg[10] ss[00] event[09] Flush FIFO [ 162.174593] esp: esp0: ent[8] CMD val[10] sreg[86] seqreg[02] sreg2[00] ireg[10] ss[00] event[09] Transfer Information [ 162.280854] esp: esp0: ent[9] EVENT val[0a] sreg[86] seqreg[02] sreg2[00] ireg[10] ss[00] event[09] Message Out done. [ 162.389204] esp: esp0: ent[10] EVENT val[0d] sreg[80] seqreg[02] sreg2[00] ireg[20] ss[00] event[0a] Check Phase. [ 162.498588] esp: esp0: ent[11] EVENT val[04] sreg[80] seqreg[02] sreg2[00] ireg[20] ss[00] event[0d] Data Out. [ 162.607977] esp: esp0: ent[12] CMD val[90] sreg[80] seqreg[02] sreg2[00] ireg[20] ss[00] event[04] Perform DMA based Transfer Information [ 162.715275] esp: esp0: ent[13] EVENT val[05] sreg[80] seqreg[02] sreg2[00] ireg[20] ss[00] event[04] Data done. What's really weird is that we disconnected as per the next-to-last interrupt in the top-level trace, but then we try to send the device data. I wonder why we don't pick up on this :-/ It seems the interrupt processing only picks up cleanly if we have not selected successfully yet, and this code path things we have (via esp->select_state). I'll try to see if there is anything about the blk-mq code paths that might shed some more light on this. -- 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
