2009-04-13 14:36:37

by Peter Horton

Subject: [PATCH] mtd: fix timeout in M25P80 driver

Extend erase timeout in M25P80 SPI Flash driver.

The M25P80 drivers fails erasing sectors on a M25P128 because the ready
wait timeout is too short. Change the timeout from a simple loop count to a
suitable number of seconds.

Signed-off-by: Peter Horton <[email protected]>
---
Index: linux-2.6.29-git12/drivers/mtd/devices/m25p80.c
===================================================================
--- linux-2.6.29-git12.orig/drivers/mtd/devices/m25p80.c 2009-04-12 21:41:16.000000000 +0000
+++ linux-2.6.29-git12/drivers/mtd/devices/m25p80.c 2009-04-12 21:43:01.000000000 +0000
@@ -54,7 +54,7 @@
#define SR_SRWD 0x80 /* SR write protect */

/* Define max times to check status register before we give up. */
-#define MAX_READY_WAIT_COUNT 100000
+#define MAX_READY_WAIT_JIFFIES (10 * HZ) /* eg. M25P128 specs 6s max sector erase */
#define CMD_SIZE 4

#ifdef CONFIG_M25PXX_USE_FAST_READ
@@ -145,20 +145,20 @@
*/
static int wait_till_ready(struct m25p *flash)
{
- int count;
+ unsigned long deadline;
int sr;

- /* one chip guarantees max 5 msec wait here after page writes,
- * but potentially three seconds (!) after page erase.
- */
- for (count = 0; count < MAX_READY_WAIT_COUNT; count++) {
+ deadline = jiffies + MAX_READY_WAIT_JIFFIES;
+
+ do {
if ((sr = read_sr(flash)) < 0)
break;
else if (!(sr & SR_WIP))
return 0;

- /* REVISIT sometimes sleeping would be best */
- }
+ cond_resched();
+
+ } while (!time_after_eq(jiffies, deadline));

return 1;
}

2009-04-13 19:41:17

by Martin Michlmayr

Subject: Re: [PATCH] mtd: fix timeout in M25P80 driver

* Peter Horton <[email protected]> [2009-04-13 15:26]:
> Extend erase timeout in M25P80 SPI Flash driver.
>
> The M25P80 drivers fails erasing sectors on a M25P128 because the ready
> wait timeout is too short. Change the timeout from a simple loop count to a
> suitable number of seconds.
>
> Signed-off-by: Peter Horton <[email protected]>

Tested-by: Martin Michlmayr <[email protected]>

This solves the problem I've seen on the QNAP TS-219.

--
Martin Michlmayr
http://www.cyrius.com/

2009-04-15 23:14:48

by Andrew Morton

Subject: Re: [PATCH] mtd: fix timeout in M25P80 driver

On Mon, 13 Apr 2009 15:26:33 +0100
Peter Horton <[email protected]> wrote:

> Extend erase timeout in M25P80 SPI Flash driver.
>
> The M25P80 drivers fails erasing sectors on a M25P128 because the ready
> wait timeout is too short. Change the timeout from a simple loop count to a
> suitable number of seconds.
>
> Signed-off-by: Peter Horton <[email protected]>
> ---
> Index: linux-2.6.29-git12/drivers/mtd/devices/m25p80.c
> ===================================================================
> --- linux-2.6.29-git12.orig/drivers/mtd/devices/m25p80.c 2009-04-12 21:41:16.000000000 +0000
> +++ linux-2.6.29-git12/drivers/mtd/devices/m25p80.c 2009-04-12 21:43:01.000000000 +0000
> @@ -54,7 +54,7 @@
> #define SR_SRWD 0x80 /* SR write protect */
>
> /* Define max times to check status register before we give up. */
> -#define MAX_READY_WAIT_COUNT 100000
> +#define MAX_READY_WAIT_JIFFIES (10 * HZ) /* eg. M25P128 specs 6s max sector erase */
> #define CMD_SIZE 4
>
> #ifdef CONFIG_M25PXX_USE_FAST_READ
> @@ -145,20 +145,20 @@
> */
> static int wait_till_ready(struct m25p *flash)
> {
> - int count;
> + unsigned long deadline;
> int sr;
>
> - /* one chip guarantees max 5 msec wait here after page writes,
> - * but potentially three seconds (!) after page erase.
> - */
> - for (count = 0; count < MAX_READY_WAIT_COUNT; count++) {
> + deadline = jiffies + MAX_READY_WAIT_JIFFIES;
> +
> + do {
> if ((sr = read_sr(flash)) < 0)
> break;
> else if (!(sr & SR_WIP))
> return 0;
>
> - /* REVISIT sometimes sleeping would be best */
> - }
> + cond_resched();
> +
> + } while (!time_after_eq(jiffies, deadline));
>
> return 1;

ow, so it sits there chewing electricity for up to ten seconds.

How much time does this really take, in the real world?

It would be better to fall back to (say) msleep(1) if we find out that
the device is being slow to respond.

2009-04-17 07:25:05

by Artem Bityutskiy

Subject: Re: [PATCH] mtd: fix timeout in M25P80 driver

On Mon, 2009-04-13 at 15:26 +0100, Peter Horton wrote:
> Extend erase timeout in M25P80 SPI Flash driver.
>
> The M25P80 drivers fails erasing sectors on a M25P128 because the ready
> wait timeout is too short. Change the timeout from a simple loop count to a
> suitable number of seconds.
>
> Signed-off-by: Peter Horton <[email protected]>
> ---
> Index: linux-2.6.29-git12/drivers/mtd/devices/m25p80.c
> ===================================================================
> --- linux-2.6.29-git12.orig/drivers/mtd/devices/m25p80.c 2009-04-12 21:41:16.000000000 +0000
> +++ linux-2.6.29-git12/drivers/mtd/devices/m25p80.c 2009-04-12 21:43:01.000000000 +0000
> @@ -54,7 +54,7 @@
> #define SR_SRWD 0x80 /* SR write protect */
>
> /* Define max times to check status register before we give up. */
> -#define MAX_READY_WAIT_COUNT 100000
> +#define MAX_READY_WAIT_JIFFIES (10 * HZ) /* eg. M25P128 specs 6s max sector erase */
> #define CMD_SIZE 4
>
> #ifdef CONFIG_M25PXX_USE_FAST_READ
> @@ -145,20 +145,20 @@
> */
> static int wait_till_ready(struct m25p *flash)
> {
> - int count;
> + unsigned long deadline;
> int sr;
>
> - /* one chip guarantees max 5 msec wait here after page writes,
> - * but potentially three seconds (!) after page erase.
> - */
Why did you remove the comment? Is it wrong or useless?

> - for (count = 0; count < MAX_READY_WAIT_COUNT; count++) {
> + deadline = jiffies + MAX_READY_WAIT_JIFFIES;
> +
> + do {
> if ((sr = read_sr(flash)) < 0)
> break;
> else if (!(sr & SR_WIP))
> return 0;
>
> - /* REVISIT sometimes sleeping would be best */
> - }
> + cond_resched();
> +
> + } while (!time_after_eq(jiffies, deadline));
>
> return 1;
> }

--
Best regards,
Artem Bityutskiy (Битюцкий Артём)

2009-04-17 08:45:42

by Peter Horton

Subject: Re: [PATCH] mtd: fix timeout in M25P80 driver

Artem Bityutskiy wrote:
> On Mon, 2009-04-13 at 15:26 +0100, Peter Horton wrote:
>> Extend erase timeout in M25P80 SPI Flash driver.
>>
>> The M25P80 drivers fails erasing sectors on a M25P128 because the ready
>> wait timeout is too short. Change the timeout from a simple loop count to a
>> suitable number of seconds.
>>
>> Signed-off-by: Peter Horton <[email protected]>
>> ---
>> Index: linux-2.6.29-git12/drivers/mtd/devices/m25p80.c
>> ===================================================================
>> --- linux-2.6.29-git12.orig/drivers/mtd/devices/m25p80.c 2009-04-12 21:41:16.000000000 +0000
>> +++ linux-2.6.29-git12/drivers/mtd/devices/m25p80.c 2009-04-12 21:43:01.000000000 +0000
>> @@ -54,7 +54,7 @@
>> #define SR_SRWD 0x80 /* SR write protect */
>>
>> /* Define max times to check status register before we give up. */
>> -#define MAX_READY_WAIT_COUNT 100000
>> +#define MAX_READY_WAIT_JIFFIES (10 * HZ) /* eg. M25P128 specs 6s max sector erase */
>> #define CMD_SIZE 4
>>
>> #ifdef CONFIG_M25PXX_USE_FAST_READ
>> @@ -145,20 +145,20 @@
>> */
>> static int wait_till_ready(struct m25p *flash)
>> {
>> - int count;
>> + unsigned long deadline;
>> int sr;
>>
>> - /* one chip guarantees max 5 msec wait here after page writes,
>> - * but potentially three seconds (!) after page erase.
>> - */
> Why did you remove the comment? Is it wrong or useless?
>

I moved the comment up to the definition of MAX_READY_WAIT_JIFFIES.
Looking through the code I missed the fact the driver can generate "chip
erase", on the M25P128 the maximum ready timeout for this is 250s !!!

P.

2009-04-17 09:22:13

by Artem Bityutskiy

Subject: Re: [PATCH] mtd: fix timeout in M25P80 driver

On Fri, 2009-04-17 at 09:22 +0100, Peter Horton wrote:
> Artem Bityutskiy wrote:
> > On Mon, 2009-04-13 at 15:26 +0100, Peter Horton wrote:
> >> Extend erase timeout in M25P80 SPI Flash driver.
> >>
> >> The M25P80 drivers fails erasing sectors on a M25P128 because the ready
> >> wait timeout is too short. Change the timeout from a simple loop count to a
> >> suitable number of seconds.
> >>
> >> Signed-off-by: Peter Horton <[email protected]>
> >> ---
> >> Index: linux-2.6.29-git12/drivers/mtd/devices/m25p80.c
> >> ===================================================================
> >> --- linux-2.6.29-git12.orig/drivers/mtd/devices/m25p80.c 2009-04-12 21:41:16.000000000 +0000
> >> +++ linux-2.6.29-git12/drivers/mtd/devices/m25p80.c 2009-04-12 21:43:01.000000000 +0000
> >> @@ -54,7 +54,7 @@
> >> #define SR_SRWD 0x80 /* SR write protect */
> >>
> >> /* Define max times to check status register before we give up. */
> >> -#define MAX_READY_WAIT_COUNT 100000
> >> +#define MAX_READY_WAIT_JIFFIES (10 * HZ) /* eg. M25P128 specs 6s max sector erase */
> >> #define CMD_SIZE 4
> >>
> >> #ifdef CONFIG_M25PXX_USE_FAST_READ
> >> @@ -145,20 +145,20 @@
> >> */
> >> static int wait_till_ready(struct m25p *flash)
> >> {
> >> - int count;
> >> + unsigned long deadline;
> >> int sr;
> >>
> >> - /* one chip guarantees max 5 msec wait here after page writes,
> >> - * but potentially three seconds (!) after page erase.
> >> - */
> > Why did you remove the comment? Is it wrong or useless?
> >
>
> I moved the comment up to the definition of MAX_READY_WAIT_JIFFIES.
> Looking through the code I missed the fact the driver can generate "chip
> erase", on the M25P128 the maximum ready timeout for this is 250s !!!

So I assume you are going to send another patch? Also, how about
Andrew's suggestion not to hog CPU?

--
Best regards,
Artem Bityutskiy (Битюцкий Артём)