phosphor-hwmon bottleneck potential

[Patrick Williams]

Rick,

On Fri, May 05, 2017 at 09:48:01AM -0700, Rick Altherr wrote:
> I've chatted with Patrick V. separately about the driver.  AST2400/2500 fan
> tach hardware measures only one fan at a time.  I think we can adjust the
> driver settings to reduce the measurement time but it will scale with # of
> tachs being read.

I never looked at this driver before but it looks like it is doing an
'msleep' in the hwmon read path after resetting a counter register and
then counting rotations?  Two comments:

1. As it stands, it doesn't appear that this driver is actually
multi-reader safe (either thread or process).  There is no locking or
queueing to prevent one reader from resetting the result register while
another is performing the msleep loop.  Multi-threading might "go
faster" but it will give entirely wrong results by my reading.

2. It seems bad to do a long-running msleep in the hwmon read path to
begin with.  Should this driver be restructured to have a kthread read
the channels in the background on a polling interval instead of
initiating by userspace action?

-- 
Patrick Williams
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Digital signature
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20170505/eabe3e73/attachment.sig>