Subject: Re: optimizations
To: Jonathan Stone <>
From: Dennis Ferguson <>
List: current-users
Date: 09/06/1997 21:48:43
> >I'm pretty sure that the Tulip driver could be made to detect the buggy
> >chips and only do the copy on those revisions, otherwise doing a 2-byte
> >fudge, much like I implemented in the "fxp" driver when porting it from
> >FreeBSD.
> My understanding is that _all_ tulip chips are `buggy' in this
> respect, and the first non-buggy chip will be the 21143.  Supposed to
> be released earlier this summer, it's late, due to bugs.)
> But it might be only `all 100Mbit chips', or I could be completely
> misunderstanding.

My reading of the 21143 hardware reference manual (the one from the DEC
web site, dated May 1997) is that it is impaired in precisely the same
way that all the other 21x4x's are.  Receive descriptor buffers must be
4-byte-aligned and a multiple of 4 bytes in length, and I can't find any
magic CSR or descriptor bits anywhere to tell it to offset the packet two
bytes into the receive buffer.

If the 21143 is really supposed to fix this, could someone point me at how
it does it?  I like these chips alot, but the alignment trouble is problematic
enough for the application we have that we don't use them.

Dennis Ferguson