Re: [PATCH V2 2/2] mtd: msm_nand: Add initial msm nand driver support.

[Murali Nalajala <mnalajal@codeaurora.org>]

On 4/27/2011 6:08 PM, Matthieu CASTET wrote:

Hi,

> Hi,
>
> Murali Nalajala a écrit :
>> On 4/18/2011 1:02 PM, Matthieu CASTET wrote:
>>> Hi,
>>>
>>> Murali Nalajala a écrit :
>>>> Add initial msm nand driver support for Qualcomm MSM platforms.
>>>> This driver is capable of handling both 2k and 4k page support
>>>> nand devices.
>>>>
>>>> This driver was originally developed by Arve Hjønnevåg at google.
>>>> Its source is available at
>>>> http://android.git.kernel.org/?p=kernel/msm.git under
>>>> android-msm-2.6.35 branch.
>>>>
>>>> CC: Dima Zavin<dima@android.com>
>>>> CC: Brian Swetland<swetland@google.com>
>>>> CC: Arve Hjønnevåg<arve@android.com>
>>>> Signed-off-by: Murali Nalajala<mnalajal@codeaurora.org>
>>>> ---
>>>> Changes in V2
>>>> * Turn most of the pr_info() calls into pr_debug().
>>>>
>>>>    drivers/mtd/devices/Kconfig    |   11 +
>>>>    drivers/mtd/devices/Makefile   |    1 +
>>>>    drivers/mtd/devices/msm_nand.c | 1597 ++++++++++++++++++++++++++++++++++++++++
>>>>    drivers/mtd/devices/msm_nand.h |   77 ++
>>>>    4 files changed, 1686 insertions(+), 0 deletions(-)
>>>>    create mode 100644 drivers/mtd/devices/msm_nand.c
>>>>    create mode 100644 drivers/mtd/devices/msm_nand.h
>>>>
>>> For the record, I am not sure it is a great idea to make nand drivers
>>> independent of the nand layer.
>>> By doing that you could change the expected behavior, and introduce bug
>>> difficult to debug.
>>>
>>> Why do you need that ? Nand drivers can override callbacks of nand framework.
>>>
>>> Maybe because the nand layer call the driver page per page, and  you want to
>>> work with more than one page. But in this case you can extend nand layer.
>>>
>>> Or does your controller don't allow sending raw command ?
>>> And in that case all the nand logic is in the nand controller ?
>>
>> Our controller doesn't allow sending RAW commands to the flash. All
>> these commands are abstracted by the controller logic. Please refer the
>> msm_nand.h file where you will find the commands which are different
>> from the actual NAND device commands.
> Ok, but does the controller does all other check : status check, ...

Yes, it does the automatic update of the status registers for every 
read/writes do report any ECC errors.

>
> In the nand framework there is already controller like yours : see the denali.c
> as example.

I had gone through it, here in driver i saw something like 
"nand_scan_ident" calls which still use the RAW commands.

So i need to do most of the custom nand scan implementation which i am 
doing it right now in my code!!!

>>
>> eg: To send read/write data command to the device we configure the
>> controller with page address&  controller specific command. Then the
>> controller itself takes care of generating the internal h/w signals
>> required by the flash and it polls for the R/B signal&  wait states are
>> internally controlled by the h/w.
>>
> So you provide your own
>      chip->ecc.read_page
>      chip->ecc.read_page_raw
>      chip->ecc.read_oob
>      chip->write_page
>      chip->ecc.write_oob

I don't see great advantage of re-writing these call backs in my driver 
( Which are very well written and tested ). Also the DMA handling is bit 
different than what i had seen in the denali.c code.
In my case most of the limitation comes from the controller.

We have used nand base kind of implementation wherever is possible.
eg: We have leveraged the code from nand base to detect the nand device 
properties.

>
> and your done.
>
>
> Matthieu


Thanks,
Murali N

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.