On 2012-04-04 15:23, Arnd Bergmann wrote:
> On Wednesday 04 April 2012, Rafal Prylowski wrote:
>> On 2012-04-03 22:08, Arnd Bergmann wrote:
>>>>> +static void ep93xx_pata_wait_for_iordy(void __iomem *base)
>>>>> +{
>>>>> + unsigned long deadline = jiffies + msecs_to_jiffies(1000);
>>>>> + while (!ep93xx_pata_check_iordy(base) &&
>>>>> + time_is_before_jiffies(deadline))
>>>>> + cpu_relax();
>>>>> +}
>>>
>>> Much better for a delay than the previous version. However, it's
>>> still a busy wait, which is bad for realtime behavior, especially
>>> since this can potentially take many milliseconds. If possible,
>>> it should have an msleep() or at least cond_resched() instead of
>>> the cpu_relax(). Obviously that will only work when no spinlocks
>>> are held.
>>>
>>
>> Unfortunately, we can't use msleep() or cond_resched() here
>> - ep93xx_pata_wait_for_iordy() is called from interrupt handler.
>
> Well, it also seems inappropriate to have a one second delay timeout
> inside of the interrupt handler.
>
> I suppose if you are emulating what a regular read of the status
> register does, there isn't much to improve here though.
>
> Arnd
According to ATA specification, maximum IORDY pulse width is 1250ns,
so I'll set timeout to 2ms (I should check this before posting the driver..).
RP
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
