On Wednesday 31 December 2008, Bruno Prémont wrote:
> On Wed, 31 December 2008 Bruno Prémont wrote:
> > Unfortunately the re-discovery of the drive causes at least XFS to
> > error and shutdown its mounts :(
> >
> > Is it possible to block any access to the devices on the scanned port
> > until the scan has completed? Otherwise this renders rescanning
> > on port with mounted (e.g. /) partition to suicide...
> >
> > I also wonder why it took so long and there is that complaint about
> > lost interrupt + failure. Was there some operation in progress that
> > got "killed" by the scan?
>
> In addition after a second/third attempt with reboot in between as /
> is no more accessible I ended up with garbage in first sectors of the
> disc!
>
> That time I did:
> cd /sys/class/ide_port/ide0/
> sync; sleep 1; echo -n 1 > delete_devices; sleep 1; echo -n 1 > scan
> caused:
> I/O error for sleep (unknown which one but my guess is it would be
> the second).
> Previously I just did which only made XFS panic:
> echo -n 1 > scan
>
> Matching kernel logs:
> Dec 31 20:21:59 venus [ 246.114799] Probing IDE interface ide0...
> Dec 31 20:21:59 venus [ 246.420137] hda: FUJITSU MHY2250BH, ATA DISK drive
> Dec 31 20:22:00 venus [ 246.780071] hda: host max PIO5 wanted PIO255(auto-tune) selected PIO4
> Dec 31 20:22:00 venus [ 246.780621] hda: UDMA/100 mode selected
> Dec 31 20:22:19 venus [ 266.100064] hda: dma_timer_expiry: DMA status (0x20)
> Dec 31 20:22:19 venus [ 266.100388] hda: lost interrupt
> Dec 31 20:22:19 venus [ 266.100588] hda: ide_dma_intr: bad DMA status (0x30)
> Dec 31 20:22:19 venus [ 266.100886] hda: dma_intr: status=0x50 { DriveReady SeekComplete }
> Dec 31 20:22:19 venus [ 266.101298] ide: failed opcode was: unknown
> Dec 31 20:22:19 venus [ 266.454984] hda: max request size: 512KiB
> Dec 31 20:22:19 venus [ 266.455254] hda: 488397168 sectors (250059 MB) w/8192KiB Cache, CHS=30401/255/63
> Dec 31 20:22:19 venus [ 266.455972] hda: cache flushes supported
> Dec 31 20:22:19 venus [ 266.493825] request_module: runaway loop modprobe binfmt-0000
> Dec 31 20:22:19 venus [ 266.494218] request_module: runaway loop modprobe binfmt-0000
> Dec 31 20:22:19 venus [ 266.494933] request_module: runaway loop modprobe binfmt-0000
> Dec 31 20:22:19 venus [ 266.495294] request_module: runaway loop modprobe binfmt-0000
> Dec 31 20:22:19 venus [ 266.496303] request_module: runaway loop modprobe binfmt-0000
> Dec 31 20:25:07 venus [ 432.720055] INFO: task udevd:630 blocked for more than 120 seconds.
> Dec 31 20:25:07 venus [ 432.720308] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> Dec 31 20:25:07 venus [ 432.720637] udevd D c0249ea0 2572 630 1
> Dec 31 20:25:07 venus [ 432.720867] f6903e18 00000086 c024b74d c0249ea0 f6903e04 f70c2d80 f6903e0c c0249e9c
> Dec 31 20:25:07 venus [ 432.721233] f6903e50 f6d79818 f6903e58 f6903e20 c03e675e f6903e28 c01487d5 f6903e44
> Dec 31 20:25:07 venus [ 432.721580] c03e6a7f c01487a0 c17821b0 f6903e50 f6d79818 00000000 f6903e70 c01477b8
> Dec 31 20:25:07 venus [ 432.721928] Call Trace:
> Dec 31 20:25:07 venus [ 432.722036] [<c024b74d>] ? __generic_unplug_device+0x1d/0x30
> Dec 31 20:25:07 venus [ 432.722245] [<c0249ea0>] ? blk_backing_dev_unplug+0x0/0x10
> Dec 31 20:25:07 venus [ 432.722446] [<c0249e9c>] ? blk_unplug+0xc/0x10
> Dec 31 20:25:07 venus [ 432.722619] [<c03e675e>] io_schedule+0xe/0x20
> Dec 31 20:25:07 venus [ 432.722788] [<c01487d5>] sync_page+0x35/0x40
> Dec 31 20:25:07 venus [ 432.722949] [<c03e6a7f>] __wait_on_bit_lock+0x3f/0x70
> Dec 31 20:25:07 venus [ 432.723135] [<c01487a0>] ? sync_page+0x0/0x40
> Dec 31 20:25:07 venus [ 432.723299] [<c01477b8>] __lock_page+0x68/0x70
> Dec 31 20:25:07 venus [ 432.730742] [<c01315e0>] ? wake_bit_function+0x0/0x60
> Dec 31 20:25:07 venus [ 432.738085] [<c014783b>] find_lock_page+0x4b/0x60
> Dec 31 20:25:07 venus [ 432.745577] [<c0148b17>] filemap_fault+0x297/0x3a0
> Dec 31 20:25:07 venus [ 432.753061] [<c025437e>] ? prio_tree_insert+0x1e/0x240
> Dec 31 20:25:07 venus [ 432.760556] [<c015548b>] __do_fault+0x3b/0x390
> Dec 31 20:25:07 venus [ 432.767962] [<c0152502>] ? vma_prio_tree_insert+0x22/0x40
> Dec 31 20:25:07 venus [ 432.775508] [<c015907d>] ? vma_link+0x3d/0x50
> Dec 31 20:25:07 venus [ 432.782967] [<c0155eb5>] handle_mm_fault+0xe5/0x570
> Dec 31 20:25:07 venus [ 432.790448] [<c0116ac0>] ? do_page_fault+0x0/0x670
> Dec 31 20:25:07 venus [ 432.797882] [<c0116d2c>] do_page_fault+0x26c/0x670
> Dec 31 20:25:07 venus [ 432.805292] [<c0258985>] ? copy_to_user+0x35/0x50
> Dec 31 20:25:07 venus [ 432.812644] [<c0116ac0>] ? do_page_fault+0x0/0x670
> Dec 31 20:25:07 venus [ 432.819910] [<c03e7a7a>] error_code+0x6a/0x70
> Dec 31 20:25:07 venus [ 432.827132] INFO: task bash:1918 blocked for more than 120 seconds.
> Dec 31 20:25:07 venus [ 432.834416] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> Dec 31 20:25:07 venus [ 432.841880] bash D f70081fc 2504 1918 1863
> Dec 31 20:25:07 venus [ 432.849416] f730ae4c 00000086 f7033fac f70081fc 00000000 f711db00 c011c3ab f730ae50
> Dec 31 20:25:07 venus [ 432.857217] 7fffffff f730aed4 00000002 f730ae94 c03e68a5 c011bede 00000000 00000001
> Dec 31 20:25:07 venus [ 432.865082] 00000003 f7008208 00000000 00000082 f700a0a0 f730ae8c 00000082 00000000
> Dec 31 20:25:07 venus [ 432.873032] Call Trace:
> Dec 31 20:25:07 venus [ 432.880908] [<c011c3ab>] ? default_wake_function+0xb/0x10
> Dec 31 20:25:07 venus [ 432.889004] [<c03e68a5>] schedule_timeout+0x75/0xc0
> Dec 31 20:25:07 venus [ 432.897171] [<c011bede>] ? __wake_up_common+0x3e/0x70
> Dec 31 20:25:07 venus [ 432.905391] [<c03e63a3>] wait_for_common+0x93/0x110
> Dec 31 20:25:07 venus [ 432.913587] [<c011c3a0>] ? default_wake_function+0x0/0x10
> Dec 31 20:25:07 venus [ 432.921818] [<c03e64b2>] wait_for_completion+0x12/0x20
> Dec 31 20:25:07 venus [ 432.930167] [<c012de8b>] call_usermodehelper_exec+0xab/0xc0
> Dec 31 20:25:07 venus [ 432.938491] [<c012e0ce>] request_module+0x9e/0xe0
> Dec 31 20:25:07 venus [ 432.946793] [<c016e462>] search_binary_handler+0x192/0x1f0
> Dec 31 20:25:07 venus [ 432.955084] [<c016f613>] do_execve+0x163/0x1b0
> Dec 31 20:25:07 venus [ 432.963311] [<c010261e>] sys_execve+0x2e/0x60
> Dec 31 20:25:07 venus [ 432.971452] [<c0103b01>] sysenter_do_call+0x12/0x25
>
> I guess the binfmt-0000 is cause by XFS reading at wrong location on disc
> and kernel seeing random data as bin format signature.
>
>
> On reboot no more bootloader.
> hexdump with rescue system showed XFS magic in very first sector of the
> disc, (rescuing the GPT worked - I assume parted used copy at end
> of disc) though the first partition also got corrupted.
>
> This looks like the scan is pretty dangerous in case anything has a
> reference to a disc/partition on the scanned channel :(
I'm really sorry to hear it but "delete_devices" / "scan" interfaces are
currently not ready for use on actively used devices (there are still some
locking issues with the core code that need to be addressed before this
can be allowed).
Even though HOWTO only mentions use for replacing devices (not rescaning)
I admit that it should be more clear about the above:
From: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
Subject: [PATCH] ide: update warm-plug HOWTO
Reported-by: Bruno Prémont <bonbons@xxxxxxxxxxxxxxxxx>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
---
Documentation/ide/warm-plug-howto.txt | 5 +++++
1 file changed, 5 insertions(+)
Index: b/Documentation/ide/warm-plug-howto.txt
===================================================================
--- a/Documentation/ide/warm-plug-howto.txt
+++ b/Documentation/ide/warm-plug-howto.txt
@@ -11,3 +11,8 @@ unplug old device(s) and plug new device
# echo -n "1" > /sys/class/ide_port/idex/scan
done
+
+NOTE: please make sure that partitions are unmounted and that there are
+no other active references to devices before doing "delete_devices" step,
+also do not attempt "scan" step on devices currently in use -- otherwise
+results may be unpredictable and lead to data loss if you're unlucky
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
