[Kernel-janitors] [patch] use schedule() instead of yield()

[Kernel-janitors] [patch] use schedule() instead of yield()

[maximilian attems]

[-- Attachment #1: Type: text/plain, Size: 716 bytes --]

patch is on top of 2.6.6-rc1.

yield() removes process from active array into expired array,
better just preempt with schedule().
as belows timeout is set to 5 seconds, 
schedule() may be better than to use a schedule_timeout(1).

tested on my laptop, but maybe my card never reach this path.


a++ maks


diff -ruN a/drivers/net/sis900.c b/drivers/net/sis900.c
--- a/drivers/net/sis900.c	2004-03-16 08:47:07.000000000 +0100
+++ b/drivers/net/sis900.c	2004-03-12 11:19:29.000000000 +0100
@@ -600,7 +600,7 @@
 
 	if(status & MII_STAT_LINK){
 		while (poll_bit) {
-			yield();
+			schedule();
 
 			poll_bit ^= (mdio_read(net_dev, sis_priv->cur_phy, MII_STATUS) & poll_bit);
 			if (time_after_eq(jiffies, timeout)) {

[-- 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

Re: [Kernel-janitors] [patch] use schedule() instead of

[Felipe W Damasio]

[-- Attachment #1: Type: text/plain, Size: 458 bytes --]

	Max,

maximilian attems wrote:
> yield() removes process from active array into expired array,
> better just preempt with schedule().

	What we wanted to do here is just yield the CPU for a bit (not to give 
it up for a long time, like the 2.6 yield).

> as belows timeout is set to 5 seconds, 

	Eh?

> schedule() may be better than to use a schedule_timeout(1).

	Perhaps:

set_current_state (TASK_UNINTERRUPTIBLE);
schedule_timeout (HZ/100);

	?

Felipe

[-- 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

Re: [Kernel-janitors] [patch] use schedule() instead of

[maximilian attems]

[-- Attachment #1: Type: text/plain, Size: 570 bytes --]


On Mon, 19 Apr 2004, Felipe W Damasio wrote:

> >yield() removes process from active array into expired array,
> >better just preempt with schedule().
> 
> 	What we wanted to do here is just yield the CPU for a bit (not to 
> 	give it up for a long time, like the 2.6 yield).

ok i'll rediff and test tomorrow morning.

> 
> >as belows timeout is set to 5 seconds, 
> 
> 	Eh?

well sis900.c sets 
	unsigned long timeout = jiffies + 5 * HZ;

there for i thought that this hole issue "while(poll_bit)"
wasn't that timing sensitive.
thanks for your propososal.

a++ maks


[-- 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