> > As per my earlier email, the proposal is not trying to support arbitrary
 > > transforms.  If the admin has e.g. a SPEED-DUPLEX value of 100g-f,10m-h
 > > and they want to change the 10m-h value to 10m-f then they would need to
 > > enter 100g-f,10m-f, or first do -=10m-h followed by += of 10m-f.  If they
 > > just want to add or remove a specific value then they can use += or -=.
 > > For instance, if they wanted to add CPU 12 to the list of cpus, they'd do
 > > set-linkprop -p cpus+=12.
 > 
 > In all of these cases, I presume that the datalink.conf will have something
 > (imo gross) like
 > 
 > en_10gfdx_cap=1,en_10ghdx_cap=..,en_1000fdx_cap=..,en_1000hdx_cap=..,en_100fdx_cap=..,en_100hdx_cap=..,
 >  en_10fdx_cap=..,en_10hdx_cap=..

Which case?  Certainly not in the model I'm suggesting, as those
properties no longer would exist.

 > >  > Also in today's model, only the modified setting (en_10fdx_cap = 0, 
 > >  > in my example) is saved in datalink.conf.  As I understand, in your
 > >  > proposal, the whole set of enabled props would have to get stored in
 > >  > datalink.conf (and then, if the NIC is DR'ed or otherwise changes its
 > >  > defaults, the stored set in datalink.conf may be out of sync).
 > > 
 > > There are pros and cons with both approaches in the case of DR'd hardware.
 > > Are they overriding the setting because of a switch problem?  A card
 > > problem?  Just general confusion?  Unclear, so it's hard to say what ideal
 > > behavior would be.
 > 
 > Imposing defaults from user-space for untuned values (instead of just
 > letting the driver do its thing) is not a good idea- reset-linkprop 
 > carefully avoids this by actually removing the property setting itself.

I don't know why you think this is being proposed.  If someone tunes
en-mii, then there will be a recorded value for it that will be set when
the link is created.  If there is not a setting for en-mii, then no value
will be set when the link is created.

-- 
meem

Reply via email to