3.8.13 bone66 mmcqd related kernel panic under disk load

[rainforest1155]

With netconsole enabled I was finally able to capture some details on the random freezes I'm getting on my BBB during it's overnight backup. At times it crashed the same night I last rebooted it while at other times it worked up to 3 weeks before crashing. I've had the issue with previous kernel versions as well, e.g. I was on bone47 and bone62. Pretty much since I had fully set up my Beagle at the beginning of the year.
This is what I had in the netconsole log today:

[224948.385572] INFO: task mmcqd/0:71 blocked for more than 60 seconds. 
[224948.392495] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[224948.400964] mmcqd/0         D c04a09ff     0    71      2 0x00000000
[224948.408086] Kernel panic - not syncing: hung_task: blocked tasks
[224948.414593] [<c001125d>] (unwind_backtrace+0x1/0x98) from [<c049bb93>] (panic+0x5b/0x160)
[224948.423351] [<c049bb93>] (panic+0x5b/0x160) from [<c00724ed>] (watchdog+0x1a1/0x1c8)
[224948.431661] [<c00724ed>] (watchdog+0x1a1/0x1c8) from [<c004485f>] (kthread+0x6b/0x74)
[224948.440058] [<c004485f>] (kthread+0x6b/0x74) from [<c000c7fd>] (ret_from_fork+0x11/0x34)
[224948.448708] drm_kms_helper: panic occurred, switching back to text console

I'm running Debian 7 with the latest updates where the system runs from the eMMC, but there's also a SD card for storage purposes and an USB stick for swap, temporary files etc.

Doing a search, I did come across another report of the issue:
https://github.com/archlinuxarm/PKGBUILDs/issues/749
and that also links back to a thread on the beagleboard group:
http://comments.gmane.org/gmane.comp.hardware.beagleboard.user/43479

Apart from this issue, I'm pretty happy with the 3.8.13 kernel. Is there a way that a fix for it could be backported from one of the newer kernel versions?

Or should I just update to one of the newer versions? If so, what specific kernel version would you recommend? It should be pretty stable as the Beagle is supposed to run as a headless server on its own. I don't need any cape support, but would like to have basic ACPI support so I can initiate a Debian shutdown with the power button on the BBB. And on finished shutdown, the Beagle should turn itself off.

Thanks,
Sebastian

EDIT: Just fixed the log entry by marking it as code so it's not partially filtered. Sorry for not noticing that before.

[rainforest1155]

Actually, it wasn't entirely correct that I had all the updates for Debian installed. On Aug 30 I updated these packages:
libc-bin libc-dev-bin locales multiarch-support libc6-dev libc6
to 2.13-38+deb7u4
I restarted the system and overnight, the same freeze occurred. I thought of it as coincidence, but then the next night it happened again.
So I rolled the packages back to 2.13-38+deb7u3 which helped that the system wasn't crashing each night, but less frequently again.
Last night I decided to give the 2.13-38+deb7u4 version of the packages another try and the crash happened again over night. Here's the net console log:

[56582.698248] INFO: task mmcqd/0:71 blocked for more than 60 seconds.
[56582.705082] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[56582.713434] mmcqd/0         D c04a09ff     0    71      2 0x00000000
[56582.720398] Kernel panic - not syncing: hung_task: blocked tasks
[56582.726811] [<c001125d>] (unwind_backtrace+0x1/0x98) from [<c049bb93>] (panic+0x5b/0x160)
[56582.735476] [<c049bb93>] (panic+0x5b/0x160) from [<c00724ed>] (watchdog+0x1a1/0x1c8)
[56582.743696] [<c00724ed>] (watchdog+0x1a1/0x1c8) from [<c004485f>] (kthread+0x6b/0x74)
[56582.752021] [<c004485f>] (kthread+0x6b/0x74) from [<c000c7fd>] (ret_from_fork+0x11/0x34)
[56582.760578] drm_kms_helper: panic occurred, switching back to text console 

So for now, I rolled the packages back to deb7u3.

I'd appreciate any suggestions that you can offer.

Sebastian