[PATCH] 2.5.41, cciss (3 of 3)

[PATCH] 2.5.41, cciss (3 of 3)

[Stephen Cameron]

Wait up to 20 seconds for polled commands to complete.  A certain multiport
storage box needs this.

diff -urN linux-2.5.41-p/drivers/block/cciss.c linux-2.5.41-q/drivers/block/cciss.c
--- linux-2.5.41-p/drivers/block/cciss.c	Wed Oct  9 15:22:14 2002
+++ linux-2.5.41-q/drivers/block/cciss.c	Wed Oct  9 15:54:35 2002
@@ -1318,9 +1318,9 @@
         unsigned long done;
         int i;
 
-        /* Wait (up to 2 seconds) for a command to complete */
+        /* Wait (up to 20 seconds) for a command to complete */
 
-        for (i = 200000; i > 0; i--) {
+        for (i = 2000000; i > 0; i--) {
                 done = hba[ctlr]->access.command_completed(hba[ctlr]);
                 if (done == FIFO_EMPTY) {
                         udelay(10);     /* a short fixed delay */

Re: [PATCH] 2.5.41, cciss (3 of 3)

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 998 bytes --]

On Fri, 2002-10-11 at 16:10, Stephen Cameron wrote:
> 
> Wait up to 20 seconds for polled commands to complete.  A certain multiport
> storage box needs this.
> 
> diff -urN linux-2.5.41-p/drivers/block/cciss.c linux-2.5.41-q/drivers/block/cciss.c
> --- linux-2.5.41-p/drivers/block/cciss.c	Wed Oct  9 15:22:14 2002
> +++ linux-2.5.41-q/drivers/block/cciss.c	Wed Oct  9 15:54:35 2002
> @@ -1318,9 +1318,9 @@
>          unsigned long done;
>          int i;
>  
> -        /* Wait (up to 2 seconds) for a command to complete */
> +        /* Wait (up to 20 seconds) for a command to complete */
>  
> -        for (i = 200000; i > 0; i--) {
> +        for (i = 2000000; i > 0; i--) {
>                  done = hba[ctlr]->access.command_completed(hba[ctlr]);
>                  if (done == FIFO_EMPTY) {
>                          udelay(10);     /* a short fixed delay */

ugh 20 seconds udelay....

why can't you sleep here ?
(and yes 20 seconds WILL trigger watchdogs!)


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [PATCH] 2.5.41, cciss (3 of 3)

[Stephen Cameron]

Arjan van de Ven wrote:

> On Fri, 2002-10-11 at 16:10, Stephen Cameron wrote:
[... a bogus patch involving a 20 second udelay ...]

> ugh 20 seconds udelay....
>
> why can't you sleep here ? [...]

No reason that I can see.  Thanks for pointing it out.
Is this better?  (I also changed the spaces to tabs 
while I was at it.)

-- steve

diff -urN linux-2.5.41/drivers/block/cciss.c linux-2.5.41-20sec/drivers/block/cciss.c
--- linux-2.5.41/drivers/block/cciss.c	Fri Oct 11 09:17:00 2002
+++ linux-2.5.41-20sec/drivers/block/cciss.c	Fri Oct 11 09:07:17 2002
@@ -1297,24 +1297,25 @@
 /*
  *   Wait polling for a command to complete.
  *   The memory mapped FIFO is polled for the completion.
- *   Used only at init time, interrupts disabled.
+ *   Used only at init time, interrupts from the HBA are disabled.
  */
 static unsigned long pollcomplete(int ctlr)
 {
-        unsigned long done;
-        int i;
+	unsigned long done;
+	int i;
+	DECLARE_WAIT_QUEUE_HEAD(polling_wqh);
 
-        /* Wait (up to 2 seconds) for a command to complete */
+	/* Wait (up to 20 seconds) for a command to complete */
 
-        for (i = 200000; i > 0; i--) {
-                done = hba[ctlr]->access.command_completed(hba[ctlr]);
-                if (done == FIFO_EMPTY) {
-                        udelay(10);     /* a short fixed delay */
-                } else
-                        return (done);
-        }
-        /* Invalid address to tell caller we ran out of time */
-        return 1;
+	for (i = 20 * HZ; i > 0; i--) {
+		done = hba[ctlr]->access.command_completed(hba[ctlr]);
+		if (done == FIFO_EMPTY)
+			interruptible_sleep_on_timeout(&polling_wqh, 1);
+		else
+			return (done);
+	}
+	/* Invalid address to tell caller we ran out of time */
+	return 1;
 }
 /*
  * Send a command to the controller, and wait for it to complete.  

RE: [PATCH] 2.5.41, cciss (3 of 3)

[Cameron, Steve]

> From: Arjan van de Ven wrote:
> On Fri, 2002-10-11 at 16:10, Stephen Cameron wrote:
> > 
> > Wait up to 20 seconds for polled commands to complete.  A 
> certain multiport
> > storage box needs this.
> > 
> > diff -urN linux-2.5.41-p/drivers/block/cciss.c 
> linux-2.5.41-q/drivers/block/cciss.c
> > --- linux-2.5.41-p/drivers/block/cciss.c	Wed Oct  9 15:22:14 2002
> > +++ linux-2.5.41-q/drivers/block/cciss.c	Wed Oct  9 15:54:35 2002
> > @@ -1318,9 +1318,9 @@
> >          unsigned long done;
> >          int i;
> >  
> > -        /* Wait (up to 2 seconds) for a command to complete */
> > +        /* Wait (up to 20 seconds) for a command to complete */
> >  
> > -        for (i = 200000; i > 0; i--) {
> > +        for (i = 2000000; i > 0; i--) {
> >                  done = 
> hba[ctlr]->access.command_completed(hba[ctlr]);
> >                  if (done == FIFO_EMPTY) {
> >                          udelay(10);     /* a short fixed delay */
> 
> ugh 20 seconds udelay....
> 
> why can't you sleep here ?
> (and yes 20 seconds WILL trigger watchdogs!)

Ok.  Nevermind that patch then.  We'll have to rethink it.
-- steve