Random Kernel Panic on Dreamplug (FS related)

[Mattia Rossi]

Am 30.09.2014 16:05, schrieb Ian Lepore:
> On Tue, 2014-09-30 at 04:34 -0700, John-Mark Gurney wrote:
>> Ian Lepore wrote this message on Mon, Sep 29, 2014 at 07:49 -0600:
>>> On Sun, 2014-09-28 at 21:01 -0700, John-Mark Gurney wrote:
>>>> Mattia Rossi wrote this message on Fri, Sep 26, 2014 at 14:19 +0200:
>>>>> This might be part of the weird FFS issues the Dreamplug has and no-one
>>>>> knows why they're happening.
>>>> Are you running w/ FFS journaling?  If so, try turning it off, but
>>>> keeping softupdates on..
>>>>
>>> It's not an SU+J problem, or even an SU problem.  fsck finds
>>> non-existant errors on filesystems known to be clean, and if
>>> write-enabled it will corrupt the good filesystem when attempting to
>>> correct those "errors".  This is on armv4 only, not v6.  I tested with
>>> and without softupdates on.  I tested with UFS1 and UFS2 filesystems.
>>> You can even do a newfs followed immediately by an fsck on it and it
>>> will corrupt the fs.
>>>
>>> The one thing I haven't done is opened a PR for this.
>> Hmm... I just tested this on my AVILA board, and I don't see this on
>> either UFS1 or UFS2...  Are you doing this via HD or md?  My testing was
>> via a 64MB md as I don't have a good way to attach external storage to
>> my board...
>>
>> If you really are seeing immediate corruption to an SD card, then I'd
>> make sure that the card is getting the correct data written to it...
>>
> There isn't actually any corruption on the filesystem until fsck starts
> trying to fix what it thinks is wrong.  That is, fsck -n will report
> problems, you then move that drive to a non-armv4 system and fsck there
> reports no problems.  If you let armv4 fsck "fix" problems then move the
> drive to another system and re-check, the filesystem IS corrupted at
> that point.
>
>> I'd suggest trying to run ZFS since it checksums everything it writes,
>> but not sure if it'd run, and if so, how well...
>>
> afaik, nobody has ever tried zfs on any arm platform.  Maybe it just
> works.  I'd love to hear from someone about it.  I don't have time
> myself to learn to configure and administer it.

I think there are at least a few people that tried it 
(http://lists.freebsd.org/pipermail/freebsd-arm/2014-February/007455.html). 
I've tried it as well (on an external ESATA disk), but had some 
performance problems, which I think could be resolved if I would run a 
bootloader (which I don't). The other solution would be to apply the 
following patch, and tweak performance via sysctls on a running system: 
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=187594

Anyhow, first things first..