LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* Network device driver Question
@ 2003-11-26 16:57 John Newlin
  0 siblings, 0 replies; only message in thread
From: John Newlin @ 2003-11-26 16:57 UTC (permalink / raw)
  To: linux-kernel

I'm working on a driver for a new device.  This NIC is on-chip, and 
memory mapped.
It can DMA to/from any memory location.

I am trying to optimize the driver such that it there is no need to copy 
to/from sk_buff
in the send_packet and on packet receive.

The problem I have is with a CPU that has a writeback cache.  When the 
send_packet
is called, the data in the sk_buff may be cached.  In the driver I can 
allocate a page as
uncached and copy to that page, however I am trying to avoid unneeded 
copies.

For receive there is a similar problem.  When I allocate the sk_buff I 
need to invalidate
the caches.  That way after a new packet arrives a read from that memory 
location will
not hit in the cache and return incorrect data.

Is there some sanctioned way that will work on any architecture to cause 
a writeback
of the dcached based on a range of virtual addresses, and similiarly 
cause an invalidate?

Thanks,

-John Newlin
 jnewlin@rawbw.com



^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2003-11-26 16:57 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-11-26 16:57 Network device driver Question John Newlin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).