On Sat, Oct 24, 2009 at 07:07:00PM +0000, b. f. wrote: > >That is: it seems to work fine for some fraction of a minute > >(doesn't seem to be longer than a minute, anyway), and then > >stops completely for several minutes (processes reading or > >writing sit in "D" state in ps) and then starts again, after > >logging "Request completed with CAM_REQ_CMP_ERR\nRetrying > >Command". > > In the past week or so, Alexander Motin ([email protected]) and Andrew > Thompson ([email protected]) have made a number of related changes > to cam and usb in the P4 repository, and in 9-CURRENT. Some of these > may address your problem. I'm not sure when they will be back-ported > to 8.X. You may wish to try out the latest version of -CURRENT, to > see if it solves your problem(s); or to contact them.
I've done this, and it seems to have worked. It seems possible that the bulk throughput (measured by systat while doing a cat /backup/bigfile >/dev/null) might even have increased a bit, but maybe not. The big improvement is that the transfer isn't pausing any more. No more CAM_REQ_CMP_ERR messages. Thanks b. f. for the suggestion, and thanks Alexander and Andrew for the fixes! I haven't run -current for, probably, ten years, and the occasional messages about lock-order-reversals worry me a bit, but don't seem to be doing any harm. Should I report them? In a PR? Please count this message as a vote for MFC'ing those cam and usb changes to 8-STABLE. Cheers, -- Andrew _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[email protected]"
