JFFS2 wrong data CRC message

JFFS2 wrong data CRC message

[Adrian Hunter]

Hi

Can someone explain why this message is not an error?

JFFS2 notice: (17573) read_dnode: wrong data CRC in data node at 0x07e85800: read 0x86c6656c, calculated 0xaf4f2bc5.

Re: JFFS2 wrong data CRC message

[David Woodhouse]

On Wed, 2007-05-02 at 12:19 +0300, Adrian Hunter wrote:
> Can someone explain why this message is not an error?
> 
> JFFS2 notice: (17573) read_dnode: wrong data CRC in data node at
> 0x07e85800: read 0x86c6656c, calculated 0xaf4f2bc5. 

Because it's almost certainly harmless. It usually just means that you
rebooted or reset without cleanly unmounting -- and thus without the
writebuffer getting flushed to finish the node it was writing at the
time. So the data CRC fails, and the node is ignored.

-- 
dwmw2

Re: JFFS2 wrong data CRC message

[Adrian Hunter]

ext David Woodhouse wrote:
> On Wed, 2007-05-02 at 12:19 +0300, Adrian Hunter wrote:
>> Can someone explain why this message is not an error?
>>
>> JFFS2 notice: (17573) read_dnode: wrong data CRC in data node at
>> 0x07e85800: read 0x86c6656c, calculated 0xaf4f2bc5. 
> 
> Because it's almost certainly harmless. It usually just means that you
> rebooted or reset without cleanly unmounting -- and thus without the
> writebuffer getting flushed to finish the node it was writing at the
> time. So the data CRC fails, and the node is ignored.
> 
Unfortunately I started with an empty flash and have not rebooted at all.

I've been running stress tests constantly for about a week, and this is
the only JFFS2 message.  The tests do unmount and remount, but there are
no error messages.

So unless you can offer an alternative explanation I am going to have to
assume it is an error.

Re: JFFS2 wrong data CRC message

[David Woodhouse]

On Wed, 2007-05-02 at 16:02 +0300, Adrian Hunter wrote:
> Unfortunately I started with an empty flash and have not rebooted at all.
> 
> I've been running stress tests constantly for about a week, and this is
> the only JFFS2 message.  The tests do unmount and remount, but there are
> no error messages.
> 
> So unless you can offer an alternative explanation I am going to have to
> assume it is an error.

That's certainly possible -- and in fact is why I've resisted the
persistent requests _not_ to print that message.

Can I see a hexdump of the offending node?

-- 
dwmw2

Re: JFFS2 wrong data CRC message

[Vitaly Wool]

On 5/2/07, Adrian Hunter <ext-adrian.hunter@nokia.com> wrote:
> >
> > Because it's almost certainly harmless. It usually just means that you
> > rebooted or reset without cleanly unmounting -- and thus without the
> > writebuffer getting flushed to finish the node it was writing at the
> > time. So the data CRC fails, and the node is ignored.
> >
> Unfortunately I started with an empty flash and have not rebooted at all.

I'd suspect some caching issues.

Vitaly

Re: JFFS2 wrong data CRC message

[Jörn Engel]

On Wed, 2 May 2007 14:06:24 +0100, David Woodhouse wrote:
> On Wed, 2007-05-02 at 16:02 +0300, Adrian Hunter wrote:
> > Unfortunately I started with an empty flash and have not rebooted at all.
> > 
> > I've been running stress tests constantly for about a week, and this is
> > the only JFFS2 message.  The tests do unmount and remount, but there are
> > no error messages.
> > 
> > So unless you can offer an alternative explanation I am going to have to
> > assume it is an error.
> 
> That's certainly possible -- and in fact is why I've resisted the
> persistent requests _not_ to print that message.

You could add a new node type to mark "safe" crc errors.  Something
roughly like this:
- Print error for crc errors, unless the next node in eraseblock is a
  "known crc error" node.
- Write "known crc error" node on mount for any node that is the last in
  an eraseblock if there still is enough room to write further data.

That should be able to eliminate most false positive.  If you care.

Jörn

-- 
The wise man seeks everything in himself; the ignorant man tries to get
everything from somebody else.
-- unknown

Re: JFFS2 wrong data CRC message

[David Woodhouse]

On Wed, 2007-05-02 at 15:51 +0200, Jörn Engel wrote:
> You could add a new node type to mark "safe" crc errors.  Something
> roughly like this:
> - Print error for crc errors, unless the next node in eraseblock is a
>   "known crc error" node.
> - Write "known crc error" node on mount for any node that is the last in
>   an eraseblock if there still is enough room to write further data.
> 
> That should be able to eliminate most false positive.  If you care.

I'm not sure we even need to go that far. If the data CRC fails and the
last data bytes are 0xFF rather than anything else, then it's almost
certain it's an unfinished write. I just don't care -- I'm happy enough
seeing the warning. 

-- 
dwmw2                       09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0

Re: JFFS2 wrong data CRC message

[Adrian Hunter]

ext David Woodhouse wrote:
> Can I see a hexdump of the offending node?

Sorry it is long gone.  The problem is most likely the OneNAND driver anyway.