On 06/29/2005 11:45 PM Jan Kiszka wrote:
> Wolfgang Grandegger wrote:
>> ...
>> I was rigt, The problem is with "frags" in rt_3c59.c":
>
> Oh, yes, our old friend again. :)
>
>>
>> for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
>> skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
>>
>> vp->tx_ring[entry].frag[i+1].addr =
>> // *** RTnet: page mapping correct? Or is this code never used?
>> cpu_to_le32(pci_map_single(vp->pdev,
>>
>> (void*)page_address(frag->page) +
>> frag->page_offset,
>> frag->size, PCI_DMA_TODEVICE));
>>
>> What memory is actually used for the RTnet skb's?
>>
>
> The usage of skb_shinfo is illegal with real-time skbs (rtskbs). It
> assumes that there is some skb_shared_info struct at the end of the
> payload buffer. But rtskbs are not shared as normal skbs between several
> users. Thus, the assumption for RTnet should be that always only a
> single fragment exists and that the rtskb at hand already describes it.
> But I don't know yet how to apply this to the 3c59x code.
OK, I realized as well, that I stumbeled over untested code. For some
reason the kernel of FC 2 uses "zero-copy". I disabled DD_ZEROCOPY and
now it works. Is the rtskb memory contiguous? How is it allocated?
Actually what the DMA engine of the 3c59 needs is a chain of memory
chunk address and size. I will look into this later on.
Thank.
Wolfgang.
-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
_______________________________________________
RTnet-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/rtnet-users
