Thread: Bus Reset Observations
Brought to you by:
aeb,
bencollins
From: Mark K. <mk...@co...> - 2000-05-26 18:12:58
|
Hi, I don't know whether to report this as a stack problem or a gscanbus problem. (I think stack...) I am seeing a very specific and repeatable problem. To me, at a high level, this looks like incorrect operation at the stack level. On my systems I believe that when the Linux based PC is not connected to any 1394 devices, and a cable to a 1394 device is inserted into the OHCI card, the 1394 stack does not recognize a cable insertion. Can someone verify whether they are seeing or not seeing the same results? Thanks, Mark TESTS...... I have set up 3 PCs with RH 6.2, downloaded the 2.2.15 kernel, built it first with the patch, and then with the CVS tree. I have downloaded and installed 1394libraw from the CVS tree and installed that. I have downloaded gscanbus and am running it on all three PCs. Messages in /var/log look pretty reasonable based on how many devices are on the bus. I can set up a network with the 3 PCs and a few other peripherals, and sometimes get gscanbus to correctly see the topology identically on all three machines. (And sometimes not!) For gscanbus I built a guid-resolv.conf file with the GUIDs of each PC and put the same file on all three PCs. ***** Test 1 ***** Connect 2 PCs and a Sony CD drive. Start gscanbus. Observations: I see 2 PCs and one unknown device. If I pull the cable on the drive, I see 2 PCs. (Bus Reset detected) If I insert the cable, the unknown device comes back. (Bus Reset Detected) If I power off the drive, the unknown device disappears. (Bus Reset Detected) If I power the drive back up, the unknown device reappears. (Bus Reset Detected) ***************** ***** Test 2 ***** Connect 1 PC and the Sony CD drive. Start gscanbus. Observations: I see 1 PC and one unknown device. If I pull the cable, I see 1 PC. (Bus reset detected) If I insert the drive cable, the drive does not come back. (Bus Reset is not detected) If I power off the drive, no change in the screen, because the screen is incorrect right now. (Bus Reset Detected) If I power the drive back up, then unknown device reappears. (Bus Reset Detected) |
From: Olivier D. <dai...@el...> - 2000-05-26 19:20:31
|
I noticed EXACTLY the same problem. Here is my debug info: (I added the output of the Node Identification and status register (11'h0E8) May 26 15:05:50 milk kernel: ohci1394_0: Bus reset May 26 15:05:50 milk kernel: ohci1394_0: SelfID process finished but NodeID not valid: 0800FFC0 May 26 15:05:50 milk kernel: ohci1394_0: irq_handler timeout event=0x00020000 May 26 15:05:50 milk kernel: ohci1394_0: Got phy packet ctx=0 ... discarded The problem is that the device DOES recognize the cable insertion, but the SelfID process terminates with an invalid NodeID. (it should be something like 0x8800FFC0 instead of 0x0800FFC0). The MSB (bit 31) of the quadlet says if the 1394 OHCI has a valid node number. The 27nd bit is the CPS (Cable Power Status OK). I noticed that when there is 2 OHCI cards (in Linux boxes) on the bus, there is always one card that ends with an invalid NodeID and the other ends with a valid NodeID. My knowledge of the ohci driver (and ieee1394 in general) is not high enough to track down the problem by now. Olivier Mark Knecht wrote: > Hi, > I don't know whether to report this as a stack problem or a gscanbus > problem. (I think stack...) I am seeing a very specific and repeatable > problem. To me, at a high level, this looks like incorrect operation at the > stack level. On my systems I believe that when the Linux based PC is not > connected to any 1394 devices, and a cable to a 1394 device is inserted into > the OHCI card, the 1394 stack does not recognize a cable insertion. > > Can someone verify whether they are seeing or not seeing the same > results? > > Thanks, > Mark > > > TESTS...... > > I have set up 3 PCs with RH 6.2, downloaded the 2.2.15 kernel, built it > first with the patch, and then with the CVS tree. I have downloaded and > installed 1394libraw from the CVS tree and installed that. I have downloaded > gscanbus and am running it on all three PCs. Messages in /var/log look > pretty reasonable based on how many devices are on the bus. I can set up a > network with the 3 PCs and a few other peripherals, and sometimes get > gscanbus to correctly see the topology identically on all three machines. > (And sometimes not!) > > For gscanbus I built a guid-resolv.conf file with the GUIDs of each PC > and put the same file on all three PCs. > > ***** Test 1 ***** > > Connect 2 PCs and a Sony CD drive. Start gscanbus. > > Observations: > I see 2 PCs and one unknown device. > If I pull the cable on the drive, I see 2 PCs. (Bus Reset detected) > If I insert the cable, the unknown device comes back. (Bus Reset Detected) > If I power off the drive, the unknown device disappears. (Bus Reset > Detected) > If I power the drive back up, the unknown device reappears. (Bus Reset > Detected) > > ***************** > > ***** Test 2 ***** > > Connect 1 PC and the Sony CD drive. Start gscanbus. > > Observations: > > I see 1 PC and one unknown device. > If I pull the cable, I see 1 PC. (Bus reset detected) > > If I insert the drive cable, the drive does not come back. (Bus Reset is not > detected) > > If I power off the drive, no change in the screen, because the screen is > incorrect right now. (Bus Reset Detected) > If I power the drive back up, then unknown device reappears. (Bus Reset > Detected) > > > _______________________________________________ > mailing list lin...@li... > http://lists.sourceforge.net/mailman/listinfo/linux1394-devel -- Olivier Daigle Projet Harfang (514) 396-8800 ext.7699 |
From: Olivier D. <dai...@el...> - 2000-05-26 19:23:34
|
I forgot to say... If the cable is connected to both computers and I load ieee1394 modules on the machines, the load of modules generates a "soft" reset on the bus and SelfID always complete successfully. Olivier Mark Knecht wrote: > Hi, > I don't know whether to report this as a stack problem or a gscanbus > problem. (I think stack...) I am seeing a very specific and repeatable > problem. To me, at a high level, this looks like incorrect operation at the > stack level. On my systems I believe that when the Linux based PC is not > connected to any 1394 devices, and a cable to a 1394 device is inserted into > the OHCI card, the 1394 stack does not recognize a cable insertion. > > Can someone verify whether they are seeing or not seeing the same > results? > > Thanks, > Mark > > > TESTS...... > > I have set up 3 PCs with RH 6.2, downloaded the 2.2.15 kernel, built it > first with the patch, and then with the CVS tree. I have downloaded and > installed 1394libraw from the CVS tree and installed that. I have downloaded > gscanbus and am running it on all three PCs. Messages in /var/log look > pretty reasonable based on how many devices are on the bus. I can set up a > network with the 3 PCs and a few other peripherals, and sometimes get > gscanbus to correctly see the topology identically on all three machines. > (And sometimes not!) > > For gscanbus I built a guid-resolv.conf file with the GUIDs of each PC > and put the same file on all three PCs. > > ***** Test 1 ***** > > Connect 2 PCs and a Sony CD drive. Start gscanbus. > > Observations: > I see 2 PCs and one unknown device. > If I pull the cable on the drive, I see 2 PCs. (Bus Reset detected) > If I insert the cable, the unknown device comes back. (Bus Reset Detected) > If I power off the drive, the unknown device disappears. (Bus Reset > Detected) > If I power the drive back up, the unknown device reappears. (Bus Reset > Detected) > > ***************** > > ***** Test 2 ***** > > Connect 1 PC and the Sony CD drive. Start gscanbus. > > Observations: > > I see 1 PC and one unknown device. > If I pull the cable, I see 1 PC. (Bus reset detected) > > If I insert the drive cable, the drive does not come back. (Bus Reset is not > detected) > > If I power off the drive, no change in the screen, because the screen is > incorrect right now. (Bus Reset Detected) > If I power the drive back up, then unknown device reappears. (Bus Reset > Detected) > > > _______________________________________________ > mailing list lin...@li... > http://lists.sourceforge.net/mailman/listinfo/linux1394-devel -- Olivier Daigle Projet Harfang (514) 396-8800 ext.7699 |
From: Ichiro S. <ic...@fn...> - 2000-06-01 21:06:17
|
Hi all, I also had the same (or similar) problem. Connecting two OHCI cards directly, Node-ID 0 is assigned to one card and 2 (instead of 1) is assigned to the other. Then, the Self-ID process fails. Sometime, I observed wrong Seld-ID, 0800FFC0 on one card. I didn't make tests in systematic way and I'm not sure if this behavior is alway same, sorry. And, I found a workaround, use non-OHCI card as a root node. When I connect these OHCI cards to a PCILynx card or an unknown chip-set card, these cards successfully finished a Self-ID process and valid Node-IDs are assigned. My configurations are; o OHCI card: D-Link DFW-500 (TI's TSB12LV23 and TSB41LV03) o PCILynx: TI TSBKPCI403 (TSB12LV21B and TSB21LV03 ?) o Unknown chip-set: On-board I/F of Compaq SP700 o Linux 2.2.14, ieee1394-2.2.14-20000318, libraw1394-0.6 Thanks, ----------------------------------------------------------------- Ichiro Suzuki ic...@fn..., http://home.fnal.gov/~ichiro/ PPD/EPP, Fermi National Accelerator Laboratory, Batavia, Illinois ----------------------------------------------------------------- |