Re: [PATCH 7/7] arch/unicore32/kernel/dma.c: ensure arguments to request_irq and free_irq are compatible

[Julia Lawall <julia.lawall@lip6.fr>]

On Thu, 15 Mar 2012, Guan Xuetao wrote:

> On Wed, 2012-03-14 at 10:23 +0100, Julia Lawall wrote:
>>
>> On Wed, 14 Mar 2012, Guan Xuetao wrote:
>>
>>> On Wed, 2012-03-14 at 11:19 +0300, Dan Carpenter wrote:
>>>> On Wed, Mar 14, 2012 at 04:07:24PM +0800, Guan Xuetao wrote:
>>>>> puv3_init_dma() is called ONCE when initializing.
>>>>> In logical, if request_irq(IRQ_DMAERR, *) failed, free_irq(IRQ_DMA, *)
>>>>> is unnecessary, and dma device/driver can keep on working.
>>>>> The patch could be:
>>>>>   	ret = request_irq(IRQ_DMAERR, dma_err_handler, 0, "DMAERR", NULL);
>>>>>   	if (ret) {
>>>>>   		printk(KERN_CRIT "Can't register IRQ for DMAERR\n");
>>>>>  -		free_irq(IRQ_DMA, "DMA");
>>>>>   		return ret;
>>>>>   	}
>>>>
>>>> It seems like you should remove the error return as well?
>>>>
>>>> regards,
>>>> dan carpenter
>>>>
>>> The error return value will only generate an extra warning message, and
>>> have no side-effect.
>>
>> The whole thing seems a little strange.  I guess your point is that the
>> call site never looks at the return value?  Wouldn't it be better to make
>> there be no return value in that case?  If there is a return value, some
>> calling context in the future might take that into account and then the
>> lack of a free_irq would be a memory leak.  Also if the first request_irq
>> can never fail, perhaps that should be made explicit by not testing the
>> return value?
>>
>> julia
> This function is an init_call, not a probe function, and it is only
> called ONCE.
> The dma device here has two interrupts, one IRQ_DMA, another IRQ_DMAERR.
> And the device could work without IRQ_DMAERR.
> The return value should indicate whether there is something wrong during
> initialization, so the function needs return errno when any request_irq
> is failed.
> For the first request_irq, some code has prepared its resources before
> this call, so I suppose it successful. However, the return value must be
> tested.

OK, thank you for the explanation. I will change the patch.

julia