Well what a Christmas day! If you're in a Christmas culture, then
Happy Christmas.
It looks like I might finally have the FSX monkey off my back.
I found and fixed a problem in the way trees with wide tnodes were being pruned.
To manifest, needs quite a complicated sequence of truncated down,
then up, then down again... the sort of stuff triggered by fsx but
very seldom in real-wold applications which probably explains why this
bug has been around for 4 or so years.
This did not fail my earlier fsx tests because those were on smaller
partitions that did not trigger the problem.
The fixed code seems to be surviving fsx pretty well.
Note that this is a fix and does not need the wideTnodeDisable hack
mentioned below.That was just a debugging hack.
Changes are checked in as yaffs_guts V1.100. I like the neatness of that.
Thanks particularly to JiSheng Zang for helping fix this problem.
I'll go be a Family Man now... great weather here in New Zealand, 32 deg C....
-- CHarles
On Thu, Dec 24, 2009 at 6:54 PM, JiSheng Zhang <
jszhang3@gmail.com> wrote:
> Hi Charles,
>
> 2009/12/24 Charles Manning <manningc2@actrix.gen.nz>:
>> Hello
>>
>> I wonder if you can try something for me?
>>
>> In yaffs_fs.c there is a set of lines
>> #if CONFIG_YAFFS_DISABLE_WIDE_TNODE
>> dev->wideTnodeDisable=1;
>> #endif
>>
>> By default this is off.
>>
>> Can you try setting the compile flag to have wideTnodeDisable = 1 and see if
>> that changes things.
>>
>> I found that this changes things in my test harness and I shall be
>> investigating why.
>>
>> -- CHarles
>
> wow!!
> I have tested for about 1 hour even without yesterday yaffs_fs.c fix,
> no failure so far!!! It seems that there's bug in logical
> address=>Tnode translation.
>
>
> Cheers,
> Jisheng
>
> _______________________________________________
> yaffs mailing list
> yaffs@lists.aleph1.co.uk
> http://lists.aleph1.co.uk/cgi-bin/mailman/listinfo/yaffs
>
(text/plain)