[Click] Click and e1000e
Roman Chertov
rchertov at cs.ucsb.edu
Tue Aug 19 19:22:49 EDT 2008
Adam Greenhalgh wrote:
> Hi,
>
> I've been playing with the ixgbe driver, I have reception for a single
> queue working but am currently trying to get multiqueue support
> working.
>
> Does any one have any thoughts on how click should handle NICs which
> support multi queuing ? Our current thinking is that each queue needs
> a separate click "poll" element to enable multithreading to work.
> However this breaks the current poll element device ownership model.
I have not checked the manual extensively, but how does the card
schedule which queue should be drained? Or do the queue owners
cooperate in creating a schedule?
My progress so far:
I have managed to get the igb poll_on/off, tx_pqueue and tx_clean to
work. I changed those functions to heavily rely on the already existing
igb code. As a result, tx_eob and tx_start are just empty shells now.
With the new code, I have managed to send 1.4 Mpps 64-byte frames.
I still need to do the RX code before the driver becomes useful.
Roman
>
> Adam
>
> 2008/8/18 Roman Chertov <rchertov at cs.ucsb.edu>:
>> Hello,
>> I am not sure how many people follow e1000, list but they have made
>> a split between e1000 to e1000 and e1000e. There is also an igb driver
>> which is pretty similar to e1000e code. We got 6 quad-port Intel which
>> require the use of the igb driver. I am currently trying to port the
>> igb driver to Click, and I am facing a few issues. The main issue is
>> when the machine crashes, it has a tendency to damage the file system,
>> so that even fsck cannot fix all the issues. I am curious if anybody
>> had issues with crashes causing fs corruption and how the minimized the
>> effects.
>>
>> Roman
>> _______________________________________________
>> click mailing list
>> click at amsterdam.lcs.mit.edu
>> https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>>
>
More information about the click
mailing list