Re: SPI read problem

[Jonathan Cameron]

On 04/21/11 12:30, Murali K. Vemuri wrote:
> 2011/4/21 Murali K. Vemuri <vemuri.muralikrishna@xxxxxxxxx>:
>> 2011/4/21 Pei Lin <telent997@xxxxxxxxx>:
>>> 2011/4/21 Murali K. Vemuri <vemuri.muralikrishna@xxxxxxxxx>:
>>>> On Wed, Apr 20, 2011 at 10:48 PM, Jonathan Cameron <jic23@xxxxxxxxx> wrote:
>>>>> On 04/20/11 05:40, Murali K. Vemuri wrote:
>>>>>> On Wed, Apr 20, 2011 at 12:19 PM, 徐建辉 <xujianhui168@xxxxxxxxx> wrote:
>>>>>>> you can use tasklet.
>>>>>>>
>>>>>>> 2011/4/19 Murali K. Vemuri <vemuri.muralikrishna@xxxxxxxxx>
>>>>>>>>
>>>>>>>> Hello there,
>>>>>>>> I have a device which is Written_to/Read_from using SPI Bus.
>>>>>>>>
>>>>>>>> I initialize the device like this:
>>>>>>>>
>>>>>>>> static int __init myDEV_init(void)
>>>>>>>> {
>>>>>>>>    return spi_register_driver(&myDEV_driver);
>>>>>>>> }
>>>>>>>> module_init(myDEV_init);
>>>>>>>>
>>>>>>>> My myDEV_probe() function looks like this:
>>>>>>>>
>>>>>>>> static int __devinit myDEV_probe(struct spi_device *spi)
>>>>>>>> {
>>>>>>>> .....
>>>>>>>> .....
>>>>>>>> /*when my CPU receives a particular IRQ, I have to do a spi_read() and
>>>>>>>> do couple of spi_write(), I am using like this:*/
>>>>>>>> err = request_threaded_irq (MY_DEV_IRQ, NULL, myDEV_irq_thread,
>>>>>>>>            IRQF_TRIGGER_RISING, "MYDEV", NULL);
>>>>> This should be fine.  The thread can sleep just fine. We do this
>>>>> all over the place.  Could be something to do with the irq
>>>>> handling on the device.
>>>>>>>>
>>>>>>>> return 0;
>>>>>>>> }
>>>>>>>>
>>>>>>>> When the Interrupt is received by the CPU, myDEV_irq_thread is being
>>>>>>>> called.
>>>>>>>> However, in the ISR I have to do  spi_read(). This is always returning
>>>>>>>> "Timed out".
>>>>> That's very suspicious.  What spi controller are we dealing with here?
>>>>>>>>
>>>>>>>> To verify if the device is not at all working, along with registering
>>>>>>>> for the IRQ, I also created a char device through I did a IOCTL.
>>>>>>>> The IOCTL command in turn does spi_read(). In this case spi_read() is
>>>>>>>> successful.
>>>>>>>>
>>>>>>>> Since it is necessary to do spi_read() in my ISR, I searched around
>>>>>>>> and found some information that
>>>>>>>> spi_read() is callable only from contexts that can sleep.
>>>>>>>>
>>>>>>>> How can I make my ISR sleep? or is there any other way of doing this?
>>>>>>>>
>>>>>>>> Any help is highly appreciated.
>>>>>>>> Thanks & regards
>>>>>>>> Murali
>>>>>>>> --
>>>>>>>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>>>>>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>>>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>>>>>> Please read the FAQ at  http://www.tux.org/lkml/
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> Hi there,
>>>>>>
>>>>>> I tried using "tasklet", "workqueue" and semaphore ... all my attempts failed.
>>>>>> if there is any other method I should try, please let me know.
>>>>> Something nastier is happening here.  It's not a problem with what you've
>>>>> presented in this email.  Perhaps a minimum example of the full code surrounding
>>>>> the problem may help us diagnose it.
>>>>>
>>>>
>>>> Thanks a lot for the inputs. this is the read function I have:
>>>> (What  I am trying to do is read register E6 of my device, the
>>>> register width is 16 bits.).
>>>>
>>>> static struct spi_device * my_SPI_DEV; (this is populated in the probe function)
>>>>
>>>> static int myDEV_read(u8 * buf, u16 len)
>>>> {
>>>>    int err;
>>>>    err = spi_read(my_SPI_DEV, buf, len);
>>>>    return err;
>>>> }
>>>> static int read_e6()
>>>> {
>>>>    u8 buf[3];
>>>>    int err;
>>>>    buf[0] = 0xe6;
>>>>    buf[1] = 0x00;
>>>>    buf[2] = 0x00;
>>>>    err = myDEV_read(buf, 3);
>>>>    printk(KERN_ERR "myDEV read returned: %d\n",err);
>>>>    return err;
>>>> }
>>>>
>>>> Now, this function read_e6() is called from the ISR I mentioned above.
>>>>
>>>
>>>
>>> spi_read will call spi_sync and this call may only be used from a
>>> context that may sleep.  The sleep
>>>  is non-interruptible, and has no timeout. should use in
>>> non-interruptable context.
>>> you should put the spi_read on context which can sleep, as bottom half
>>> "work queue"
>>> Or use spi_async interface.
>>>
>> I added "workqueue", and got the same result. Also, I used "tasklet"
>> and got the same result.
>> spi_async does not seem to be working for me, returning EINVAL (-22).
>> spi_sync() seem to be returning 0 (supposed to be success) but no data
>> coming out ... still investigating why ZEROS are returned instead of
>> valid data.
>>
>> if you have any suggestions, please let me know.
>>
>> Murali
>>>
>>>> 90% of time, the read returns "-5" (I/O Error) and about 10% time it
>>>> gets "0".  However I added a little hexdump to check the content in
>>>> case the return value is "0". The register content is always "0".
>>>>
>>>> Please let me know if I am doing any mistake.
>>>> Thanks & regards
>>>> Murali
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>> Please read the FAQ at  http://www.tux.org/lkml/
>>>>
>>>
>>>
>>>
>>> --
>>> Best Regards
>>> Lin
>>>
>>
> Okay, finally dug about and found that My chip was expecting the read
> command in different style and I am sending in different style.
> my spi_read() has to send only one byte (the register address ) on the
> SIMO line and expect 2 bytes on the SOMI line.
> I used a load of debug prints to find that my code (as below) is
> providing the two bytes of space on SIMO line itself. So, what changes
> are needed in this code to read correctly?
> 
> static struct spi_device * my_SPI_DEV; (this is populated in the probe function)
> 
> static int myDEV_read(u8 * buf, u16 len)
> {
>    int err;
>    err = spi_read(my_SPI_DEV, buf, len);
>    return err;
> }
> static int read_e6()
> {
>    u8 buf[3];
>    int err;
>    buf[0] = 0xe6;
>    buf[1] = 0x00;
>    buf[2] = 0x00;
>    err = myDEV_read(buf, 3);
>    printk(KERN_ERR "myDEV read returned: %d\n",err);
>    return err;
> }

int a = spi_w8r16(my_sPI_DEV, 0xe6);

a should contain your result, or a negative if an
error occurs.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/