Dale E. Martin <dale@xxxxxxxxxxxxxxx> :
> although it looks like the card is only found on cold boots.
Can you send (or put on a webpage) the same dmesg and lspci -vx when
it happens ?
> The irony of this, btw, is that linux/Documentation/Changes still
> recommends gcc 2.95.3 - even in version 2.6.8, possibly in 2.6.10. (I
> think I saw it in there but I don't have it handy.)
You saw it there. No comment.
> Also, when I first started looking at this, I tried to #define
> RTL8169_DEBUG but was getting compile errors. Should I file a bug about
> this ?
I lost this one. Re-added to the queue. Don't bother with a PR for it.
> It looks like the poor 1Ghz C3 in this machine is going to be the
> bottleneck for getting good transfer rates over gigabit ethernet ;-)
It depends on the kind of transfer. A bigger MTU really helps. So does
TSO and checksumming but it is really dependant on the kind of load.
This is the content of the patch against 2.4.28.
> What is the intention of this patch? Is it to handle the 2.95.x issue or
> something else ?
See above. Can you:
- send me discretly the .o built from the r8169 sources with 2.95.3 ?
- pest me until you have a 2.95.3 compiled r8169 module which does not
lock up any more ?