Re: [PATCH v8 0/3] add the GPMI controller driver for IMX23/IMX28

[Huang Shijie <b32955@freescale.com>]

Hi Koen:

I tested the driver in mx23evk board, and I do not find the DMA-timeout 
error.

The following is my test environment:
[0] Hardware:
        MX23EVK PCB REV C
[1] SW:
     kernel:git://git.linaro.org/people/shawnguo/linux-2.6.git mxs-gpmi
     .config: my_config , you can find it in my previous email.
     cmdline: console=ttyAMA0,115200 root=/dev/mmcblk0p1 rw rootwait 
gpmi_debug_init mtdparts=gpmi-nand:20m(boot),-(user)

[2] test shell script:
================================

echo 0x20 > /sys/module/gpmi_nand/parameters/gpmi_debug
flash_eraseall /dev/mtd1
ubiformat /dev/mtd1
flash_eraseall /dev/mtd1
ubiattach /dev/ubi_ctrl -m 1
ubimkvol /dev/ubi0 -N test -m
mount -t ubifs ubi0:test tmp
bonnie++ -d tmp -u 0 -s 2 -r 1
bonnie++ -d tmp -u 0 -s 2 -r 1
bonnie++ -d tmp -u 0 -s 2 -r 1
umount tmp
ubidetach /dev/ubi_ctrl -m 1

================================

[3] conclusion
    It runs well in my mx23 board, and no DMA-TIMEOUT occur.
I think there is some conflict in your board.

BTW:
    I split the MTD to 20M and (the rest size).
The flash_eraseall works well.
If i do not split the MTD, the flash_eraseall will not work.
I guest overflow occurs in some place.

Best Regards
Huang Shijie


> On Tue, Aug 2, 2011 at 8:37 AM, Huang Shijie<b32955@freescale.com>  wrote:
>> Hi,
>>> Hi,
>>>
>>> Don't think it's a hardware issue. After all it is working on the
>>> mx23evk using the Freescale BSP based on 2.6.35.3.
>> I doubt the Freescale BSP has fix the conflict, while the community kernel
>> does not
>> fix it.
>> Anyway, I will do more test on the mx23 board myself.
> Strange indeed. I suspect that the BSP used another way to mark bad blocks.
>
>>
>>
>>> Now I tested using this kernel command line:
>>> mtdparts=gpmi-nand:20m(boot),-(user) gpmi_debug_init
>>> (So I removed the ubi rootfs stuff, see previous mails).
>>>
>>> Its booting fine now without complaining about dma timeouts.
>> ok.
> That's mainly because the ubi stuff is not done at startup.
>
>>> Have tried flash_eraseall /dev/mtd1. Now it's complaining and thinking
>>> all blocks are bad. So a "flash_erase: Skipping bad block at 1db00000"
>>> message for every erase block.
>>> Writing 1 to /sys/devices/platform/imx23-gpmi-nand/ignorebad, makes to
>> This is a very dangerous interface.
> I know but it solved an issue here. Now flash_eraseall seems to be
> working again without that ignorebad turned on (only finds one bad
> erase blcok).
> Again: I suspect that the BSP used another way to mark bad blocks.
>
>>> go a little further giving 'only one' error but I doubt it's working
>>> at all. See attached log file.
>>>
>>> When the trying ubiformat /dev/mtd1 its stopping with a dma timeout.
>> echo 0x20>  /sys/devices/platform/imx23-gpmi-nand/gpmi_debug
>>
>> to show the GPMI register when the DMA timeout occurs.
> Don't have this sysfs interface here. I also don't see where it is
> registered in the driver code.
> Anyway I have hardcoded GPMI_DEBUG_CRAZY in gpmi_debug_setup.
>
> log file attached.
>
> Regards,
> Koen
>
>>
>> Best Regards
>> Huang Shijie
>>> Every subsequent action (e.g. using ubiformat or flash_eraseall) using
>>> that dma channel is failing.
>>> I have put a printk in mxs-dma.c around line 387:
>>>
>>>      if (mxs_chan->status == DMA_IN_PROGRESS&&    !append)
>>>      {
>>>          pr_err("dma already in progess!\n");
>>>          return NULL;
>>>      }
>>>
>>> I see that printk also in the log file in attach. Seems like the
>>> mxs-dma is failing at some point and for the rest of the time always
>>> keeps it's status at DMA_IN_PROGRESS
>>>
>>> Any input or feedback is welcome.
>>>
>>> Regards,
>>> Koen
>>>
>>>
>>> On Sun, Jul 31, 2011 at 3:51 PM, Marek Vasut<marek.vasut@gmail.com>
>>>   wrote:
>>>> On Friday, July 29, 2011 05:00:48 PM Koen Beel wrote:
>>>>> On Fri, Jul 29, 2011 at 2:41 PM, Lothar Waßmann<LW@karo-electronics.de>
>>>> wrote:
>>>>>> Hi,
>>>>>>
>>>>>> Koen Beel writes:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On Fri, Jul 29, 2011 at 9:20 AM, Lothar
>>>>>>> Waßmann<LW@karo-electronics.de>
>>>> wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Koen Beel writes:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> I have test on mx23evk board. Still see the same issues.
>>>>>>>>>
>>>>>>>>> On Wed, Jul 27, 2011 at 3:53 AM, Huang Shijie<b32955@freescale.com>
>>>> wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> Thanks for your test.
>>>>>>>>>>
>>>>>>>>>>> Hi,
>>>>>>>>>>>
>>>>>>>>>>> It's not really working for me.
>>>>>>>>>>> I've applied all gpmi-nand driver patches and the dma driver
>>>>>>>>>>> patches.
>>>>>>>>>>>
>>>>>>>>>>> I have added following kernel parameters:
>>>>>>>>>>> mtdparts=gpmi-nand:20m(boot),-(user) ubi.mtd=1 root=ubi0:rootfs0
>>>>>>>>>>> rootfstype=ubifs gpmi_debug_init
>>>>>>>>>>>
>>>>>>>>>>> During boot I get already get DMA timeout:
>>>>>>>>>>> [    2.560000] [ start_dma_without_bch_irq : 392 ] DMA timeout,
>>>>>>>>>>> last DMA
>>>>>>>>>>>
>>>>>>>>>>> :1
>>>>>>>>>>>
>>>>>>>>>>> [    3.560000] [ start_dma_with_bch_irq : 427 ] bch timeout!!!
>>>>>>>>>>> ...
>>>>>>>>>>> (see log file in attach line 89)
>>>>>>>>> I still see this DMA timeout when booting.
>>>>>>>>> What kernel parameters do you use? I still use same as above.
>>>>>>>> Do you have an SD card in the system? I'm getting the same error when
>>>>>>>> accessing an SD card. Without SD card I can use the flash without any
>>>>>>>> errors.
>>>>>>> No SD card in the system. At least not on my mx23evk board.
>>>>>>> My real target hardware has a SDIO wifi chip.
>>>>>>>
>>>>>>> Are you also testing on the mx23evk board? Really strange it is not
>>>>>>> reproducible here then.
>>>>>>> Anyone using mx23evk please give:
>>>>>>> - revision number of board (I have tested on rev C1)
>>>>>>> - kernel parameter used (see previous mail for mine)
>>>>>> It's getting ever more strange. I only have problems with concurrent
>>>>>> access to the flash and SD-card when I do file system based accessed
>>>>>> to the SD card.
>>>>>> A 'dd if=/dev/mmcblk0 of=/dev/null' while excessively using the flash
>>>>>> works well, while mounting the SD card and doing an 'ls' produces
>>>>>> immediate BCH DMA timeouts.
>>>>>>
>>>>>> Perhaps someone can make anything out of that.
>>>>> I have checked and the function dma_irq_callback is never called
>>>>> during my tests. This means the DMA transfer does not even works once.
>>>>>
>>>>> Also no clue here ...
>>>> Aren't you getting some undervolt on the powerrail for example ? btw the
>>>> driver
>>>> looks a bit crappy to me, but let's believe it works ;-)
>>>>
>>>>>> Lothar Waßmann
>>>>>> --
>>>>>> ___________________________________________________________
>>>>>>
>>>>>> Ka-Ro electronics GmbH | Pascalstraße 22 | D - 52076 Aachen
>>>>>> Phone: +49 2408 1402-0 | Fax: +49 2408 1402-10
>>>>>> Geschäftsführer: Matthias Kaussen
>>>>>> Handelsregistereintrag: Amtsgericht Aachen, HRB 4996
>>>>>>
>>>>>> www.karo-electronics.de | info@karo-electronics.de
>>>>>> ___________________________________________________________
>>>>> _______________________________________________
>>>>> linux-arm-kernel mailing list
>>>>> linux-arm-kernel@lists.infradead.org
>>>>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>
>>