Skip to content

Commit

Permalink
powerpc/rtas: Only sleep in rtas_busy_delay if we have useful work to do
Browse files Browse the repository at this point in the history
RTAS returns extended error codes as a hint of how long the
OS might want to wait before retrying a call. If we have nothing
else useful to do we may as well call back straight away.

This was found when testing the new dynamic dma window feature.
Firmware split the zeroing of the TCE table into 32k chunks but
returned 9901 (which is a suggested wait of 10ms). All up this took
about 10 minutes to complete since msleep is jiffies based and will
round 10ms up to 20ms.

With the patch below we take 3 seconds to complete the same test.
The hint firmware is returning in the RTAS call should definitely
be decreased, but even if we slept 1ms each iteration this would
take 32s.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
  • Loading branch information
Anton Blanchard authored and Benjamin Herrenschmidt committed Apr 27, 2011
1 parent a7b8ad4 commit eca590f
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion arch/powerpc/kernel/rtas.c
Original file line number Diff line number Diff line change
Expand Up @@ -494,7 +494,7 @@ unsigned int rtas_busy_delay(int status)

might_sleep();
ms = rtas_busy_delay_time(status);
if (ms)
if (ms && need_resched())
msleep(ms);

return ms;
Expand Down

0 comments on commit eca590f

Please sign in to comment.