public inbox for linuxppc-dev@ozlabs.org 
 help / color / mirror / Atom feed
* RE: Power PC 440GX Jumbo Packets EMAC2/3
@ 2005-02-24 13:12 zDL Linux PPC
  2005-02-24 13:45 ` Gerhard Jaeger
  0 siblings, 1 reply; 5+ messages in thread
From: zDL Linux PPC @ 2005-02-24 13:12 UTC (permalink / raw)
  To: 'Gerhard Jaeger', zML Linux PPC Dev; +Cc: zML Linux PPC Embedded

> On Thursday 24 February 2005 12:46, zDL Linux PPC wrote:
> > I have been doing some experiments with Jumbo packets (MTU 
> 9000) and have
> > noticed that once you exceed the 2KB fifo size, I end up 
> with no response
> > from the box.
> > 
> > I can ping the box with no problem at all at sizes greater 
> than 2KB, however
> > upon further investigation I noticed packets of size 2006 
> bytes causes no
> > further packet to be transmitted (adding debug prints to 
> ibm_emac_core.c,
> > the driver believes the packet has been transmitted), after 
> adding a further
> > 42 bytes of header information this comes to the exact size 
> of the Transmit
> > Fifo 2048 (2KB). I have also changed the Transmit Fifo 
> Register to 1024
> > bytes, this solves the problem at 2006 bytes however now it 
> has moved to 982
> > bytes of data (+42bytes of header = 1024)
> > 
> > Has anyone else experienced this problem while either using 
> a smaller
> > transmit fifo (512 or 1024 bytes) and  MTU 1500 or a larger 
> MTU (>2033) and
> > EMAC transmit fifo of 2048. I guess this will really apply 
> to people with
> > larger MTU not many will decrease the Tx fifo, and I am 
> further guessing
> > that not many are using 9000 MTU as not much hardware seems 
> to support it, I
> > can only use it through a direct connection between PC's 
> and not through a
> > switch!
> > 
> > Any suggestions on how to overcome this problem??? My only 
> idea is to either
> > add data to the packet (don't think this will work) or 
> forceably fragment
> > the packet (if it is 2K in size) to ensure that it never 
> uses the exact size
> > of the tx fifo.....
> 
> Have you applied Matt Porters patch? It has been submitted at 
> the beginning
> of this month?

No I haven't seen it on either of these lists, do you know where I can find
it 

Thanks

^ permalink raw reply	[flat|nested] 5+ messages in thread
* RE: Power PC 440GX Jumbo Packets EMAC2/3
@ 2005-02-24 13:55 zDL Linux PPC
  0 siblings, 0 replies; 5+ messages in thread
From: zDL Linux PPC @ 2005-02-24 13:55 UTC (permalink / raw)
  To: 'Gerhard Jaeger'; +Cc: zML Linux PPC Dev, zML Linux PPC Embedded

Gerhard,

> [SNIPSNAP]
> > > > Any suggestions on how to overcome this problem??? My only 
> > > idea is to either
> > > > add data to the packet (don't think this will work) or 
> > > forceably fragment
> > > > the packet (if it is 2K in size) to ensure that it never 
> > > uses the exact size
> > > > of the tx fifo.....
> > > 
> > > Have you applied Matt Porters patch? It has been submitted at 
> > > the beginning
> > > of this month?
> > 
> > No I haven't seen it on either of these lists, do you know 
> where I can find
> > it 
> > 
> > Thanks
> 
> Browse the linuxpp-embedded archive list, I think it was 
> around the 9th
> of february...
> or try this one
> http://ozlabs.org/pipermail/linuxppc-embedded/2005-February/00
1260.html

Thanks for this however it is not the problem I am experiencing at present,
my problem is with the TX and not the RX side of things, I have added debug
and the ibm_emac_core prints out that the packet has been transmitted,
however I see nothing at the other end when you transmit a packet of tx fifo
size bytes from the PPC 440 GX, packets of <tx fifo size or >tx fifo size
are transmitted fine from the PPC 440 GX, so long as you do not transmit a
packet of Tx fifo size.

Thanks

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Power PC 440GX Jumbo Packets EMAC2/3
@ 2005-02-24 11:46 zDL Linux PPC
  2005-02-24 12:19 ` Gerhard Jaeger
  0 siblings, 1 reply; 5+ messages in thread
From: zDL Linux PPC @ 2005-02-24 11:46 UTC (permalink / raw)
  To: zML Linux PPC Dev; +Cc: zML Linux PPC Embedded

I have been doing some experiments with Jumbo packets (MTU 9000) and have
noticed that once you exceed the 2KB fifo size, I end up with no response
from the box.

I can ping the box with no problem at all at sizes greater than 2KB, however
upon further investigation I noticed packets of size 2006 bytes causes no
further packet to be transmitted (adding debug prints to ibm_emac_core.c,
the driver believes the packet has been transmitted), after adding a further
42 bytes of header information this comes to the exact size of the Transmit
Fifo 2048 (2KB). I have also changed the Transmit Fifo Register to 1024
bytes, this solves the problem at 2006 bytes however now it has moved to 982
bytes of data (+42bytes of header = 1024)

Has anyone else experienced this problem while either using a smaller
transmit fifo (512 or 1024 bytes) and  MTU 1500 or a larger MTU (>2033) and
EMAC transmit fifo of 2048. I guess this will really apply to people with
larger MTU not many will decrease the Tx fifo, and I am further guessing
that not many are using 9000 MTU as not much hardware seems to support it, I
can only use it through a direct connection between PC's and not through a
switch!

Any suggestions on how to overcome this problem??? My only idea is to either
add data to the packet (don't think this will work) or forceably fragment
the packet (if it is 2K in size) to ensure that it never uses the exact size
of the tx fifo.....

^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2005-02-24 13:55 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-02-24 13:12 Power PC 440GX Jumbo Packets EMAC2/3 zDL Linux PPC
2005-02-24 13:45 ` Gerhard Jaeger
  -- strict thread matches above, loose matches on Subject: below --
2005-02-24 13:55 zDL Linux PPC
2005-02-24 11:46 zDL Linux PPC
2005-02-24 12:19 ` Gerhard Jaeger

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox