Re: Re: How to use IP/UDP/TCP checksum offload HW?
|[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]|
On 07.06.2011 16:41, Grant Edwards wrote: > On 2011-06-07, Ilija Kocho <ilijak@xxxxxxxxxxx> wrote: > >>>>> Must we look at the latest BSD code to see how to implement this? >>>> Maybe. Or there might be a simpler way to do it. I would guess that >>>> the current FreeBSD has an API to configure this on a per-interface >>>> basis at run-time. >>>> >>>> If we can assume that there is only one external interface, then we >>>> don't need any API or runtime checks, just a few strategic #if/#endif >>>> pairs and some CDL to control them. >>> So how do we progress this? Should we offline exchange the mods we've >>> each tried (code and CDL)? > I won't have any hardware available until Q4. So I probably won't be > ble to spend much time on it for a few months. I have just started with the driver. My first step is to make plain driver without acceleration so it will take some time. But before we start coding there are many questions to answer... >> I am going to need the same soon only with lwIP. Recent on-chip >> Ethernet controllers provide for some IP processing so IMO we should >> address this issue generally. Is there place for some community >> project? > There are a number of svn/git/hg/whatever hosting sites that could be > used. IIRC, somebody had set up a site to host eCos-related stuff, I > guess this is it: > > http://ecosx.sourceforge.net/ > > It doesn't look like it's been touched for 5 years, but the guy who > set it up is still active on other projects on sourceforge. I guess this is fresher regarding lwIP: http://sourceware.org/ml/ecos-devel/2011-05/msg00003.html Only it might be stalled. John, is this an active project? > One question in my mind is whether it needs to be run-time > configurable on a per-interface basis, or can it be a global, > build-time option. We need to raise and resolve a number of questions indeed. What parts of stack are going to be affected, etc. Needless to say, device code must work with all (both) stacks that shall support new features. It could be that different devices support different (subsets of) IP features, so I guess some per device configuration shall be necessary. Regarding run-time configureability I doubt it. It will add complexity + both memory and performance penalty. Run-time user is (most often) the end user or her admin).Shall (s)he benefit? Or want to deal with it? Ilija -- Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss