Messages in this thread | | | Subject | Re: Gigabit Linux Server Bottlenecks | From | Jes Sorensen <> | Date | 11 Feb 2000 18:56:45 +0100 |
| |
>>>>> "Bjorn" == Bjorn Wesen <bjorn@sparta.lu.se> writes:
Bjorn> On Thu, 10 Feb 2000, Zachary Amsden wrote: >> > > Well in order to use it and really gain on it for transmit, you >> want > > scatter/gather skbuffs, something which I really want to >> look at > > (provided I can find the time). If you are doing the >> data copy anyway, > > there checksum is almost free.
Bjorn> Is the idea that the skb data pointer instead will be a kiovec Bjorn> ? So that the kiovec can point to the locked down COW-marked Bjorn> pages in memory where the data came from, released when TCP Bjorn> finally acks it ?
Something along those lines, and have the first element in the kiobuf be the data area containing the headers.
Bjorn> But the tail and head of the skb, that will still be in Bjorn> skb-local memory right ?
Most of that relates to where the headers are currently located, it's something we need to look closer at how to deal with when designing this.
Bjorn> I'd love to see that work - we run linux on embedded HW and Bjorn> every less copying of data can be a pretty performance win. And Bjorn> our ethernet HW is already scatter-gather.
Well, I should sure help on Gigabit Ethernet and HIPPI as well.
Jes
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |