[KJ] Re: [PATCH 14/21] net/s2io: replace schedule_timeout() with

[Nishanth Aravamudan <nacc@us.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 1216 bytes --]

On Mon, Jan 17, 2005 at 03:15:05PM -0800, Raghavendra Koushik wrote:
> Correct me if Iam wrong, but isn't HZ always defined as 
> "number of jiffies" on the machine that equals 1sec in duration. 
> So even if duration of jiffy varies, Macro HZ is assured to be 1 sec.
> This way schedule_timeout (HZ / 2) on any machine is supposed to 
> timeout after half a second and so on and so forth.

You are correct that timing out relative to HZ is ok. However, I would like to
remove this dependency wherever it exists. That way requesting a delay of
500 milliseconds can always be done via msleep() and if the behind-the-scenes
workings of schedule_timeout() change (for instance relative to tickless
systems), msleep() should still succeed.

There are some issues with the task states (more applicable to
TASK_INTERRUPTIBLE than TASK_UNINTERRUPTIBLE), but they are not relevant to your
code.

The patch really is more for consistency's sake. I don't want people to see your
driver and then assume they should use schedule_timeout() because someone else
has. msleep() is the preferred interface for longer unconditional delays. It
also makes it clearer that the sleep is for 500 msecs (e.g.) than HZ/2, IMO.

Thanks,
Nish

[-- Attachment #2: Type: text/plain, Size: 167 bytes --]

_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
http://lists.osdl.org/mailman/listinfo/kernel-janitors