OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

virtio-comment message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: Re: [PATCH v2] virtio-net: Fix and update VIRTIO_NET_F_NOTF_COAL feature


On Mon, Feb 13, 2023 at 03:43:55PM +0200, Alvaro Karsz wrote:
> I'll add an intro explaining how the entire coalescing thing works,
> without relying on the examples.
> 
> > > How about adding the following line in the main paragraph:
> > >
> > > Coalescing parameters are met when the number of sent/received packets
> > > reaches \field{tx_max_packets} since the last used buffer
> > > notification, or when \field{usecs} elapses since the last used buffer
> > > notification, whichever comes first.
> >
> > But this does mean we are tying things to used buffer
> > notifications which have their own suppression thing.
> > which yes is documented there but I would prefer to
> > make things easier to grok without jumping back and forth.
> > maybe we need a chapter explaining this idea.
> > E.g. add a concept of "a packet notification".
> > explain how they are suppressed and they may or may
> > not trigger a used buffer notification depending on whether
> > driver requested them.
> >
> 
> Honestly, having a "packet notification" concept that may evaluate to
> nothing seems a bit confusing to me.
> At the end of the day, the device sends used buffer notifications.
> 
> Eventually, we send a notification only if the notifications are not
> suppressed by the driver, so we'll need to refer to the section
> explaining how the notifications suppression works (and reader will
> need to jump to another paragraph), or we'll need to re-explain it in
> the NOTF_COAL paragraph (which seems redundant).
> 
> Maybe we can explain in the intro how the notifications coalescing
> works, and then mention that even if the coalescing parameters are met
> (this will be explained in the intro as well), we send a notification
> only if the notifications are not suppressed (then refer to the
> relevant part).
> 
> Seems ok to you?

ok so the event we are looking for is "coalescing parameters are met".
It is ok but I would like something that works even if
there is no coalescing. How about "notification conditions are met" then?
The coalescing parameters specify the conditions then



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]