fix compile-time divide-by-zero in scsi_ioctl

fix compile-time divide-by-zero in scsi_ioctl

[David Brownell]

This fixes a compiler-reported divide-by-zero when HZ < 100.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>

Index: at91/block/scsi_ioctl.c
===================================================================
--- at91.orig/block/scsi_ioctl.c	2006-07-30 01:17:34.000000000 -0700
+++ at91/block/scsi_ioctl.c	2006-07-30 01:44:17.000000000 -0700
@@ -63,7 +63,7 @@ static int scsi_get_bus(request_queue_t 
 
 static int sg_get_timeout(request_queue_t *q)
 {
-	return q->sg_timeout / (HZ / USER_HZ);
+	return q->sg_timeout / max(1,(HZ / USER_HZ));
 }
 
 static int sg_set_timeout(request_queue_t *q, int __user *p)

Re: fix compile-time divide-by-zero in scsi_ioctl

[Matthew Wilcox]

On Tue, Sep 26, 2006 at 08:48:48AM -0700, David Brownell wrote:
> This fixes a compiler-reported divide-by-zero when HZ < 100.

Shouldn't we instead be using jiffies_to_clock_t?  Thus:

Index: at91/block/scsi_ioctl.c
===================================================================
--- at91.orig/block/scsi_ioctl.c	2006-07-30 01:17:34.000000000 -0700
+++ at91/block/scsi_ioctl.c	2006-07-30 01:44:17.000000000 -0700
@@ -63,7 +63,7 @@ static int scsi_get_bus(request_queue_t 
 
 static int sg_get_timeout(request_queue_t *q)
 {
-	return q->sg_timeout / (HZ / USER_HZ);
+	return jiffies_to_clock_t(q->sg_timeout);
 }
 
 static int sg_set_timeout(request_queue_t *q, int __user *p)

Re: fix compile-time divide-by-zero in scsi_ioctl

[David Brownell]

On Wednesday 27 September 2006 1:27 pm, Matthew Wilcox wrote:
> On Tue, Sep 26, 2006 at 08:48:48AM -0700, David Brownell wrote:
> > This fixes a compiler-reported divide-by-zero when HZ < 100.
> 
> Shouldn't we instead be using jiffies_to_clock_t?  Thus:

Could be ... the units in that function unclear to me though.

What I care about is not needing to carry a patch around in my tree ...
I got no response when I posted a fix to LKML several months back,
your suggestion could well be better!

FWIW HZ=64 can be convenient for various embedded platforms with
32768/sec system clocks; HZ=100 can be impractical.

- Dave



> Index: at91/block/scsi_ioctl.c
> ===================================================================
> --- at91.orig/block/scsi_ioctl.c	2006-07-30 01:17:34.000000000 -0700
> +++ at91/block/scsi_ioctl.c	2006-07-30 01:44:17.000000000 -0700
> @@ -63,7 +63,7 @@ static int scsi_get_bus(request_queue_t 
>  
>  static int sg_get_timeout(request_queue_t *q)
>  {
> -	return q->sg_timeout / (HZ / USER_HZ);
> +	return jiffies_to_clock_t(q->sg_timeout);
>  }
>  
>  static int sg_set_timeout(request_queue_t *q, int __user *p)
> 

Re: fix compile-time divide-by-zero in scsi_ioctl

[Matthew Wilcox]

On Wed, Sep 27, 2006 at 04:28:26PM -0700, David Brownell wrote:
> On Wednesday 27 September 2006 1:27 pm, Matthew Wilcox wrote:
> > On Tue, Sep 26, 2006 at 08:48:48AM -0700, David Brownell wrote:
> > > This fixes a compiler-reported divide-by-zero when HZ < 100.
> > 
> > Shouldn't we instead be using jiffies_to_clock_t?  Thus:
> 
> Could be ... the units in that function unclear to me though.

Me too, tbh.  But the function looks right.

static inline clock_t jiffies_to_clock_t(long x)
{
#if (TICK_NSEC % (NSEC_PER_SEC / USER_HZ)) == 0
        return x / (HZ / USER_HZ);
#else
        u64 tmp = (u64)x * TICK_NSEC;
        do_div(tmp, (NSEC_PER_SEC / USER_HZ));
        return (long)tmp;
#endif
}