Re: re0 watchdog timeouts

[Patrick Welche <prlw1%cam.ac.uk@localhost>]

On Wed, May 17, 2017 at 11:49:38PM -0700, Soren Jacobsen wrote:
> On 05/17 14:21, Patrick Welche wrote:
> > On Wed, May 17, 2017 at 01:12:14PM +0000, Jared McNeill wrote:
> > > Are you able to get performance numbers before/after this set of
> > > changes? It has a noticeable impact on my arm board.
> > 
> > No - the change is from broken to working (I had a long "file
> > corruption" thread that was inconclusive, but which involved cvs
> > protocol ending up inside cvs checkout files - just did a few
> > checkouts, and all is well.  Just to make it non-deterministic, I did
> > a BIOS update at some point during the troubles.)
> 
> What Jared means is: With rtl8169.c 1.151, do you see a performance
> difference between 1.46 and 1.47 of if_re_pci.c?  Both cases (interrupt
> moderation enabled in 1.46, interrupt moderation disabled in 1.47)
> should now be functional.
> 
> For me there's no significant difference.

Just ftp'd file from wm0 server to re0 client:

without RTKQ_IM_HW:
331805736 bytes received in 00:08 (39.16 MiB/s)
331805736 bytes received in 00:08 (39.15 MiB/s)
331805736 bytes received in 00:08 (39.16 MiB/s)
331805736 bytes received in 00:08 (39.19 MiB/s)
331805736 bytes received in 00:08 (39.19 MiB/s)
 
with RTKQ_IM_HW:
331805736 bytes received in 00:08 (39.01 MiB/s)
331805736 bytes received in 00:08 (38.62 MiB/s)
331805736 bytes received in 00:08 (38.53 MiB/s)
331805736 bytes received in 00:08 (38.53 MiB/s)
331805736 bytes received in 00:08 (38.52 MiB/s)
331805736 bytes received in 00:08 (38.51 MiB/s)

Cheers,

Patrick