On Sat, 18 Feb 2006, Fredrik Roubert wrote: > Here is the dmesg output from that test run: > > [262653.162293] dc395x: Tekram DC395(U/UW/F), DC315(U) - ASIC TRM-S1040 v2.05, 2004/03/08 > [262653.162314] ACPI: PCI Interrupt 0000:04:05.0[A] -> GSI 20 (level, low) -> IRQ 19 > [262653.404392] dc395x: Used settings: AdapterID=07, Speed=0(20.0MHz), dev_mode=0x57 > [262653.404398] dc395x: AdaptMode=0x0f, Tags=4(16), DelayReset=1s > [262653.404403] dc395x: Connectors: Termination: Auto Low High > [262653.404466] dc395x: Performing initial SCSI bus reset > [262653.404985] scsi4 : Tekram DC395(U/UW/F), DC315(U) - ASIC TRM-S1040 v2.05, 2004/03/08 > [262655.033158] Vendor: HP Model: C2520A Rev: 3503 > [262655.033171] Type: Processor ANSI SCSI revision: 02 > [262655.034138] 4:0:2:0: Attached scsi generic sg2 type 3 > [262675.061545] drivers/scsi/dc395x.c[2572]: 2, 0, 0 > [262676.044080] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.073044] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.102258] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.131515] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.353526] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.402529] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.418709] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.434949] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.451199] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.467416] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.483808] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.500094] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.520097] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.536219] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.599101] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.615137] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.631210] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.647318] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.663580] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.683896] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.747918] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.774339] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.810508] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.813529] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.838307] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.864369] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.936742] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.952765] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.968897] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262676.985074] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262677.001171] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.504018] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.530334] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.566400] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.569421] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.594330] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262683.620404] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262687.173095] drivers/scsi/dc395x.c[2572]: 2, 0, 0 > [262687.198697] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262713.038424] drivers/scsi/dc395x.c[2572]: 2, 0, 0 > [262713.064027] drivers/scsi/dc395x.c[2572]: 4, 0, 0 > [262725.911499] dc395x: reselect: w/o disconnected cmds <02-0> > [262725.911521] dc395x: disconnect: Unexpected reselection <02-0> > [262845.687267] dc395x: eh_abort: (pid#1007) target=<02-0> cmd=ffff810075fe3dc0 > [262845.687274] dc395x: eh_abort: Command in progress<6>dc395x: eh_abort: (pid#1008) target=<02-0> cmd=ffff810075fe3880 > [262845.687280] dc395x: eh_abort: Command was waiting > [262855.668103] dc395x: eh_abort: (pid#1008) target=<02-0> cmd=ffff810075fe3880 > [262855.668111] dc395x: eh_abort: Command was waiting > [262855.668119] dc395x: eh_bus_reset: (pid#1007) target=<02-0> cmd=ffff810075fe3dc0 > [262855.668640] dc395x: doing_srb_done: pids G:1007(02-0) > [262865.663561] drivers/scsi/dc395x.c[2572]: 2, 0, 0 Ok, this tells us, that on your system PIO is triggered only on the "write" path and when scattergather are not used. This means that, 1) we won't be able to test the most interesting part of the proposed PIO fix - scattergather remapping, and 2) your problem might be in hardware rather than in software. For the latter I have to ask you if your cables are ok? Termination? Maybe replace the cable? Do you have any other SCSI card where you also could test your scanner? If you don't think it's a hardware problem, we might try to look more exactly how this error (unexpected reselect) occurs, how it is related to failing DMA (that's when PIO is invoked), and verify if the PIO code is correct / fix it. After you confirm that your cables and termination are fine, I would ask you to do a few more tests to look more exactly at what's happening, and to verify at least the non-sg path of my patch. So, would you agree to help me further with the tests or would you rather or would you prefer to stop now? Thanks Guennadi --- Guennadi Liakhovetski - : 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
