On Jul 03 Stefan Richter wrote: > On Jul 02 Alan Stern wrote: > > Also, I have no idea why this shows up with USB drives but not other > > SCSI transports. A fluke of timing? > > A while ago I frequently observed oopses at removal of a FireWire > CompactFlash card reader (an sd device which exposes itself as device with > removable medium). At that time I wasn't motivated to track down whether > the bug resided in the block or SCSI or firewire subsystem. The bug was > apparently triggered because hald was polling the device like crazy for > media changes, and that polling coincided with device hot unplug. > > Furthermore, FireWire CD-ROM removal never has been an exactly glitch-free > experience because the SCSI stack occasionally went on to issue command > retries for many minutes after device removal a.k.a. DID_NO_CONNECT. I > don't remember crashes during FireWire CD-ROM removal, but it has been a > while that I last used CD-ROM drives. > > I might test the card reader and a CD-ROM later next week on 3.0-rc6 if I > find the time and the issue hadn't been resolved by then. I don't have > hald anymore, but there are certainly other ways to force accesses during > device shutdown. I have not tested FireWire CD-ROMs or card reader yet. But today I did get the following crash on 3.0-rc6-71-g4dd1b49 x86-64 plus Alan's "USB: additional regression fix for device removal" when I powered down an USB hub with an empty USB card reader attached (or when I powered it up, impossible for me to say): general protection fault: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC PID: 8299, comm: kworker/0:0 Not tainted RIP: [...] elv_put_request+0xb/0x1b [...] Call Trace: __blk_put_request ? blk_put_request blk_put_request scsi_execute scsi_execute_req ? sd_check_events scsi_test_unit_ready ? kmem_cache_alloc ? sd_check_events sd_check_events disk_events_workfn [...] After this trace, the following messages were logged: generic-usb 0003:046D:C51.0059: can't reset device, 0000:00:12.2-3.1.3/input/0, status -71 usb 1-3.1: clear tt 1 (0530) error -19 BUG: unable to handle kernel paging request at fffffffffffffff8 IP: [...] kthread_data+0xb/0x11 Pid: 8299, comm: kworker/0:0 Tainted: G D RIP: 0010:[<ffffffff8104aa8f>] I can upload a screenshot of this crash if desired. The 2.6.39 and 3.0 block and/or SCSI layer are in a really sad state. -- Stefan Richter -=====-==-== -=== -=--- http://arcgraph.de/sr/ -- 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
