[patch] cfi: prevent kernel panic with cfi flash

[Jörn Engel]

On Tue, 4 October 2005 22:47:29 +0200, Jan Pedersen wrote:
> 
> When using the cfi (common flash interface) driver, every word written to
> the flash chips is followed by a operation complete poll. This poll is
> intended to have a timeout of 1 ms. However this timeout is calculated by
> HZ/1000, which happends to be 0 because HZ < 1000. The result of this is
> that there will be just one poll for operation complete. If this single poll
> fails, the kernel panics. This patch increases this timeout to HZ (1
> second). This is far more than needed, but is preferred over a panic. This
> fix is well tested and completely avoids the panic.

Hmm.  Is there a good reason to pick (HZ) instead of (HZ/1000 + 1)?
The latter would be one jiffy more than 1ms, much closer to intended
value and still always nonzero.

> Signed-off-by: Jan Pedersen <jp at jp-embedded.com>
> ---
> diff -Naur linux-2.4.31.org/drivers/mtd/chips/cfi_cmdset_0002.c
> linux-2.4.31/drivers/mtd/chips/cfi_cmdset_0002.c
> --- linux-2.4.31.org/drivers/mtd/chips/cfi_cmdset_0002.c	2004-11-17
> 06:54:21.000000000 -0500
> +++ linux-2.4.31/drivers/mtd/chips/cfi_cmdset_0002.c	2005-08-22
> 12:14:17.000000000 -0400
> @@ -510,7 +510,7 @@
>  	   or tells us why it failed. */        
>  	dq6 = CMD(1<<6);
>  	dq5 = CMD(1<<5);
> -	timeo = jiffies + (HZ/1000); /* setting timeout to 1ms for now */
> +	timeo = jiffies + (HZ); /* setting timeout to 1s for now */
>  		
>  	oldstatus = cfi_read(map, adr);
>  	status = cfi_read(map, adr);
> 

Jörn

-- 
Measure. Don't tune for speed until you've measured, and even then
don't unless one part of the code overwhelms the rest.
-- Rob Pike