> > /usr/local/bin/ogd_pci_analyzer_capture | gzip --best > > > > The display program can generate this file, and then execute it: > > > > /data/ogd_pci_analyzer/test_run_67/upload_script > > > /data/ogd_pci_analyzer/test_run_67/part5000.gz > > > > If the support machine isn't the same as the display machine, then run > > > > rsh support_machine sh < > > /data/ogd_pci_analyzer/test_run_67/upload_script > > > /data/ogd_pci_analyzer/test_run_67/p >art5000.gz > > So, say OGD1 royally screws the PCI bus beyond recognition... at > sample 5499. The 5499 most useful samples for debugging and fixing > the problem are in the card's RAM, a dead bus away from a display, > nic, or hard drive controller. > > Or am I missing something?
I think you are missing something. You have a test machine. The test machine has a PCI device being tested. The test machine also has a OGD card with the PCI analyzer package. You have a support machine. The support machine has some sort of connection to the OGD card PCI analyzer (RS-232, USB, Ethernet, ... not decided yet). This requires adding a RS-232 (or whatever) port to the OGD card. You have a machine running the analyzer_data_display program. In most cases this will be the same machine as the support machine, but it could be a third machine located in another country. The PCI device being tested "royally screws the PCI bus beyond recognition". The OGD1 can still transfer data to the support machine via RS-232/USB/Ethernet/... > One last thought... we're spying on the PCI bus... so sending data > over the PCI bus is not only counter-productive, but we'd have to > 'capture' it and send it along with the regular data... again over > the PCI bus... even with compression, that sets up some sort of > nearly-infinite loop. You are not sending the capture data over the PCI bus while the analyser is capturing data. During capture, the OGD just captures data and stores it in RAM. If the RS-232 (or whatever) link is added you don't have to transfer the capture data over PCI at all. One option is to work out a way so that the OGD's RAM survives a reboot. Then if the PCI bus hangs, you reboot the test machine and then transfer the capture data over the PCI bus. Depending of how bad the PCI device under test is behaving, it might be necessary to disable/disconnect the device under test (e.g. pull the board out). The advantage to this method is that transferring the data across PCI would be a LOT faster. Disadvantage is you'd have to work out a way so that the OGD's RAM survives a reboot, perhaps a PCI extender card with the power traces cut, and then power the OGD card externally. And if the card under test was misbehaving badly enough, pulling/replacing the card for every test run would be a hassle. Perhaps a modified extender card could eliminate this problem? _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
