Re: [PATCH] Use 64bit timer values while calibrating BogoMips

[Arjan van de Ven]

On Wed, 31 Oct 2007 16:52:16 +0100
Andi Kleen <ak@xxxxxxx> wrote:

> 
> When calibrating BogoMips 32bit architectures using 
> read_current_timer will truncate the timer value to 32bit.
> This is a problem on x86 when the TSC timer is already beyond
> 2^32-1 or wraps during the calibration. The current code
> has no wrap handling.  
> 
> This could potentially lead to wrong BogoMips, causing 
> incorrect delays e.g. in hardware drivers and then difficult
> to track down bugs.

bogomips tends to not be used for that nowadays
> 
> Change read_current_timer() to always return a 64bit
> value even on 32bit and do the computation in 64bit.
> 
> Only found by code inspection; i haven't seen a failure 
> in the wild.
> 
> avr32/sparc64 read_current_timer() updated, but I haven't 
> compiled/tested them.

one general comment, if you really want 64 bit... you're using
"unsigned long long" all over, just use u64 instead.


-- 
If you want to reach me at my work email, use arjan@xxxxxxxxxxxxxxx
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/