inconsistent behaviour with onenand with 2.6.24 kernel

inconsistent behaviour with onenand with 2.6.24 kernel

[Abhijit Singh]

Hi,

Sorry my earlier message was rejected because of suspicious header.

I am facing some problem with onenand with the kernel 2.6.24.
Everything used to work with 2.6.20 version but after upgrading the
kernel to this version there is some inconsistent behaviour. If we
keep copying small file ( ~= 20 Kb ) to onenand,it hangs after some
time. Copying some big file ( ~= 10 MB ) leads to system hang even for
the first time. I am not able to figure out whether the problem is
with onenand or jffs2.
Do i need to apply some patch to fix it ? has any one observed the behaviour ?
I have enabled the MTD_DEBUG. But all it generated was a big file. It
hangs in read, sometimes in write.

Onenand chip number that i am using is : K5W1G1GACM-DL60.
here is the boot time message from the kernel having all the details
about onenand.

Muxed OneNAND 128MB 1.8V 16-bit (0x30)
OneNAND version = 0x0221
WRITESIZE = 2048
Scanning device for bad blocks
Bad eraseblock 19 at 0x00260000
cmdlinepart partition parsing not available
mtd: Giving out device 0 to onenand.0

Our onenand is SLC.

Any pointer for this from anyone ( KP ? ).

Thanks
Abhijit

Re: inconsistent behaviour with onenand with 2.6.24 kernel

[Adrian Hunter]

Abhijit Singh wrote:
> I am facing some problem with onenand with the kernel 2.6.24.

Yes it is broken.

Here is a list on OneNAND patches in 2.6.25

69d79186dc48ca22a0ce69511bef8ef6c2465ada [MTD] [OneNAND] Use pre-alloced oob buffer instead of local buffer
978cb38a296fceac82a8a757f6387d7ef2a21ac6 [MTD] [OneNAND] Use the u_char instead of char in oobbuf
b2581be291aa8595eadf3d6933d04d0f1d01b46d [MTD] [OneNAND] Check the initial bad block using ONENAND_CTRL_ERROR
e71f04fc9234b14636887ceb5862755f1690642c [MTD] [OneNAND] Get correct density from device ID
b21b72cf33bb212414c1d967850e261b795befa4 [MTD] [OneNAND] Consolidate OneNAND operation order
9d2f0b7a3de28d06ba4011b835b9a7e772553f0d [MTD] [OneNAND] fix call to onenand_verify when writing subpages
d0b36d8cc8f345ec5faadd15daaecfb409b94523 [MTD] [OneNAND] Do not release chip twice

The last one "[MTD] [OneNAND] Do not release chip twice" is absolutely essential, so here it is:


commit d0b36d8cc8f345ec5faadd15daaecfb409b94523
Author: Adrian Hunter <ext-adrian.hunter@nokia.com>
Date:   Thu Jan 24 12:48:37 2008 +0200

    [MTD] [OneNAND] Do not release chip twice
    
    Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
    Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>

diff --git a/drivers/mtd/onenand/onenand_base.c b/drivers/mtd/onenand/onenand_base.c
index ed9f9c0..e5882b6 100644
--- a/drivers/mtd/onenand/onenand_base.c
+++ b/drivers/mtd/onenand/onenand_base.c
@@ -1447,9 +1447,6 @@ static int onenand_write_ops_nolock(struct mtd_info *mtd, loff_t to,
                buf += thislen;
        }
 
-       /* Deselect and wake up anyone waiting on the device */
-       onenand_release_device(mtd);
-
        ops->retlen = written;
 
        return ret;

Re: inconsistent behaviour with onenand with 2.6.24 kernel

[Abhijit Singh]

Hi Adrian,

On Fri, Apr 18, 2008 at 12:49 PM, Adrian Hunter
<ext-adrian.hunter@nokia.com> wrote:
> Abhijit Singh wrote:
> > I am facing some problem with onenand with the kernel 2.6.24.
> >
>
> Yes it is broken.
>
> Here is a list on OneNAND patches in 2.6.25
>
> 69d79186dc48ca22a0ce69511bef8ef6c2465ada [MTD] [OneNAND] Use pre-alloced oob
> buffer instead of local buffer
> 978cb38a296fceac82a8a757f6387d7ef2a21ac6 [MTD] [OneNAND] Use the u_char
> instead of char in oobbuf
> b2581be291aa8595eadf3d6933d04d0f1d01b46d [MTD] [OneNAND] Check the initial
> bad block using ONENAND_CTRL_ERROR
> e71f04fc9234b14636887ceb5862755f1690642c [MTD] [OneNAND] Get correct density
> from device ID
> b21b72cf33bb212414c1d967850e261b795befa4 [MTD] [OneNAND] Consolidate OneNAND
> operation order
> 9d2f0b7a3de28d06ba4011b835b9a7e772553f0d [MTD] [OneNAND] fix call to
> onenand_verify when writing subpages
> d0b36d8cc8f345ec5faadd15daaecfb409b94523 [MTD] [OneNAND] Do not release chip
> twice
>
> The last one "[MTD] [OneNAND] Do not release chip twice" is absolutely
> essential, so here it is:

I have tried all the above mentioned patches. But the status is same.
behaviour is inconsistent.
It still leads to system hangs while writing big files. Any other pointers ?
But it is working fine on 2.6.20 kernel.


Abhijit

Re: inconsistent behaviour with onenand with 2.6.24 kernel

[Adrian Hunter]

Abhijit Singh wrote:
> I have tried all the above mentioned patches. But the status is same.
> behaviour is inconsistent.
> It still leads to system hangs while writing big files. Any other pointers ?
> But it is working fine on 2.6.20 kernel.

That was the only problem I knew of.

You will need to dump the task list (using sysrq) when it hangs and try
to work out where it is stuck.

Re: inconsistent behaviour with onenand with 2.6.24 kernel

[Kyungmin Park]

On Mon, Apr 21, 2008 at 5:33 PM, Adrian Hunter
<ext-adrian.hunter@nokia.com> wrote:
> Abhijit Singh wrote:
>  > I have tried all the above mentioned patches. But the status is same.
>  > behaviour is inconsistent.
>  > It still leads to system hangs while writing big files. Any other pointers ?
>  > But it is working fine on 2.6.20 kernel.
>
>  That was the only problem I knew of.
>
>  You will need to dump the task list (using sysrq) when it hangs and try
>  to work out where it is stuck.
>

To verify which has a bug, try to another approaches,
Could you replace 2.6.20 OneNAND driver with 2.6.24. or vice versa, If
the problem is same, it's OneNAND bug.
If not, it could be another problem. not OneNAND drivers.

Thank you,
Kyungmin Park

Re: inconsistent behaviour with onenand with 2.6.24 kernel

[David Woodhouse]

On Wed, 2008-04-23 at 09:53 +0900, Kyungmin Park wrote:
> On Mon, Apr 21, 2008 at 5:33 PM, Adrian Hunter
> <ext-adrian.hunter@nokia.com> wrote:
> > Abhijit Singh wrote:
> >  > I have tried all the above mentioned patches. But the status is same.
> >  > behaviour is inconsistent.
> >  > It still leads to system hangs while writing big files. Any other pointers ?
> >  > But it is working fine on 2.6.20 kernel.
> >
> >  That was the only problem I knew of.
> >
> >  You will need to dump the task list (using sysrq) when it hangs and try
> >  to work out where it is stuck.
> >
> 
> To verify which has a bug, try to another approaches,
> Could you replace 2.6.20 OneNAND driver with 2.6.24. or vice versa, If
> the problem is same, it's OneNAND bug.

You could also use git-bisect to find exactly which changeset introduced
the problem. But first show us the results of SysRq-P and SysRq-T which
might give some clue where it's hanging.

-- 
dwmw2