Randy.Dunlap wrote:
trivial to reproduce:
modprobe aic7xxx
mount -t ext3 -o data=writeback /dev/sda4 /media/zip
ls -l /media/zip
umount /media/zip
rmmod aic7xxx ---> oops
[...]
Oct 1 21:51:27 midway kernel: [42958200.670000] Synchronizing SCSI cache for disk sda:
Oct 1 21:51:27 midway kernel: [42958200.670000] Unable to handle kernel paging request at virtual address e4bff163
[...]
Oct 1 21:51:27 midway kernel: [42958200.670000] EIP is at ahc_linux_run_command+0x58f/0x6ce [aic7xxx]
[...]
Oct 1 21:51:27 midway kernel: [42958200.670000] [<f9f7193c>] ahc_linux_queue+0x51/0x5c [aic7xxx]
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02c22cb>] scsi_dispatch_cmd+0x209/0x293
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02c7c0c>] scsi_request_fn+0x287/0x31b
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02865ba>] __generic_unplug_device+0x20/0x25
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02865d7>] generic_unplug_device+0x18/0x28
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02871a9>] blk_execute_rq_nowait+0x38/0x3d
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c0287254>] blk_execute_rq+0xa6/0xcd
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02c6ae4>] scsi_execute+0xaf/0xc9
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02c6b4f>] scsi_execute_req+0x51/0x76
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02cfbd9>] sd_sync_cache+0x4f/0xab
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02d114e>] sd_shutdown+0x34/0x39
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02d108e>] sd_remove+0x1c/0x5f
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c02803be>] __device_release_driver+0x56/0x70
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c0280400>] device_release_driver+0x28/0x3b
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c027fcd9>] bus_remove_device+0x57/0x69
Oct 1 21:51:27 midway kernel: [42958200.670000] [<c027eeef>] device_del+0x2f/0x5e
Oct 1 21:51:28 midway kernel: [42958200.670000] [<c02ca764>] __scsi_remove_device+0x2e/0x69
Oct 1 21:51:28 midway kernel: [42958200.670000] [<c02c9a44>] scsi_forget_host+0x3b/0x62
Oct 1 21:51:28 midway kernel: [42958200.670000] [<c02c2dd1>] scsi_remove_host+0x76/0xde
Oct 1 21:51:28 midway kernel: [42958200.670000] [<f9f7289e>] ahc_platform_free+0x29/0x148 [aic7xxx]
Oct 1 21:51:28 midway kernel: [42958200.670000] [<f9f679ca>] ahc_free+0x79/0xc8 [aic7xxx]
Oct 1 21:51:28 midway kernel: [42958200.670000] [<f9f75345>] ahc_linux_pci_dev_remove+0x41/0x48 [aic7xxx]
[...]
The solution is probably to add calls to scsi_remove_device() for each
device on the host as first action in ahc_linux_pci_dev_remove(). That
way, scsi high-level drivers are able to perform their shutdown
functions before the host adapter ceases to operate.
More care needs to be taken for hotswappable hardware. I don't know if
aic7xxx falls into this category. If yes, ahc_linux_queue() must be made
aware of that the adapter is gone and has to immediately complete any
commands with DID_NO_CONNECT. Similarly, the eh handlers should take
shortcuts when the adpater is gone.
AFAIK.
--
Stefan Richter
-=====-=-=-= =-=- ---=-
http://arcgraph.de/sr/
-
: 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
