[PATCH 3/5] SYSLINK:IPU-PM: solve circular dependencies ipu_pm

[Tim Gardner]

On 07/01/2010 09:53 AM, Vadillo, Miguel wrote:

>>>    /*
>>> +  Function for setup ipu_pm module
>>> + *
>>> + */
>>> +int ipu_pm_setup(void *notify_driver_handle)
>>> +{
>>> +	u32 i = 0;
>>> +	ipu_pm_notifydrv_handle = notify_driver_handle;
>>> +	/* Get the shared RCB */
>>> +	rcb_table = (struct sms *) ioremap(PM_SHM_BASE_ADDR,
>>> +		sizeof(struct sms));
>>> +
>>> +	pm_event = kzalloc(sizeof(struct pm_event) * NUMBER_PM_EVENTS,
>>> +				GFP_KERNEL);
>>> +
>>> +	/* Each event has it own sem */
>>> +	for (i = 0; i<   NUMBER_PM_EVENTS; i++) {
>>> +		pm_event[i].sem_handle = kzalloc(sizeof(struct semaphore),
>>> +						GFP_KERNEL);
>>> +		sema_init(pm_event[i].sem_handle, 0);
>>> +		pm_event[i].event_type = i;
>>> +	}
>>> +	return 0;
>>> +}
>>> +EXPORT_SYMBOL(ipu_pm_setup);
>>> +
>>
>> Why bother returning a status code if its just a constant? How about
>> checking for failures when calling ioremap() and kzalloc() ?
>
> This part is now fixed and will be available for the next release too.
 >We are checking the kzalloc. Also the ioremap is not done anymore 
since we are requesting a heap from the share memory.

So, you're saying this is all different in 2.6.35 (which I assume is the 
'next release') ? That could be awhile.

As far as I can tell from looking at the repositories at 
git://dev.omapzoom.org/pub/scm/integration you wrote these patches just 
for the Ubuntu branch. Since we may have to live with this 2.6.34 based 
kernel for a few weeks more, please take the time to write correct code 
and resend this patch. At least put some BUG_ON()s in it so we catch 
allocation failures.

rtg
-- 
Tim Gardner tim.gardner at canonical.com