I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
2*10Mbps=20Mbps.
I try to fix the pbm.
kris wrote:
>
> I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> 2*10Mbps=20Mbps.
> I try to fix the pbm.
perhaps a bug exists on the dispatcher when 2 identical cards exist.
Anyone have 2 identical cards for test ?
kris wrote:
>
> kris wrote:
> >
> > I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> > kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> > 2*10Mbps=20Mbps.
> > I try to fix the pbm.
>
> perhaps a bug exists on the dispatcher when 2 identical cards exist.
> Anyone have 2 identical cards for test ?
After other tries, the ne.c file is buggy. Confirmation.
I investigate. Anyone helps me ?
regards
Christophe
On Sun, Aug 25, 2002 at 02:28:18PM +0200, kris wrote:
> kris wrote:
> >
> > I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> > kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> > 2*10Mbps=20Mbps.
> > I try to fix the pbm.
>
> perhaps a bug exists on the dispatcher when 2 identical cards exist.
> Anyone have 2 identical cards for test ?
The ISA bus is a cause of the speed limit.
--
Vojtech Pavlik
SuSE Labs
On Sun, 25 Aug 2002, Christophe Devalquenaire wrote:
> kris wrote:
> > kris wrote:
> > >
> > > I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> > > kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> > > 2*10Mbps=20Mbps.
> > > I try to fix the pbm.
> >
> > perhaps a bug exists on the dispatcher when 2 identical cards exist.
> > Anyone have 2 identical cards for test ?
>
> After other tries, the ne.c file is buggy. Confirmation.
> I investigate. Anyone helps me ?
Do you have evidence of a specific problem?
Or the same hardware running faster with other drivers or kernel versions?
This sounds as if you are just running out of ISA bus bandwidth...
--
Donald Becker [email protected]
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
Hi,
On Sun, 25 Aug 2002, Donald Becker wrote:
> This sounds as if you are just running out of ISA bus bandwidth...
Can you explain why it's unidirectional then?
Thunder
--
--./../...-/. -.--/---/..-/.-./..././.-../..-. .---/..-/.../- .-
--/../-./..-/-/./--..-- ../.----./.-../.-.. --./../...-/. -.--/---/..-
.- -/---/--/---/.-./.-./---/.--/.-.-.-
--./.-/-.../.-./.././.-../.-.-.-
Donald Becker wrote:
>
> On Sun, 25 Aug 2002, Christophe Devalquenaire wrote:
> > kris wrote:
> > > kris wrote:
> > > >
> > > > I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> > > > kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> > > > 2*10Mbps=20Mbps.
> > > > I try to fix the pbm.
> > >
> > > perhaps a bug exists on the dispatcher when 2 identical cards exist.
> > > Anyone have 2 identical cards for test ?
> >
> > After other tries, the ne.c file is buggy. Confirmation.
> > I investigate. Anyone helps me ?
>
> Do you have evidence of a specific problem?
> Or the same hardware running faster with other drivers or kernel versions?
>
> This sounds as if you are just running out of ISA bus bandwidth...
My configuration : a server with 2 cards NE2000(10 Mbps), a machine with
2 cards : 1 NE2000(ISA) and 1 3C905b(PCI).
server : NE2000 <-> workstation : NE2000 in 192.168.1.X
server : NE2000 <-> workstation : 3C905b in 192.168.0.X
I try with a 2.4.6, 2.4.15, and 2.5.31. there is near from no
differences in the code of ne.c, gkrellm tells me that when I download
on both cards big files (for tests), the bandwidth falls to 5Mbps for
each card.
If 1 download stops, the other bandwidth is up to 10Mbps immediatly.
No other ISA card is used on the server.
On Sun, Aug 25, 2002 at 11:00:02AM -0600, Thunder from the hill wrote:
> Hi,
>
> On Sun, 25 Aug 2002, Donald Becker wrote:
> > This sounds as if you are just running out of ISA bus bandwidth...
>
> Can you explain why it's unidirectional then?
Unidirectional what? Sorry, guess I didn't understand your question.
--
Vojtech Pavlik
SuSE Labs
On Sun, Aug 25, 2002 at 07:58:32PM +0200, Christophe Devalquenaire wrote:
> Donald Becker wrote:
> >
> > On Sun, 25 Aug 2002, Christophe Devalquenaire wrote:
> > > kris wrote:
> > > > kris wrote:
> > > > >
> > > > > I have 2 ne2000 isa cards (10Mbps for each) and with this versions of
> > > > > kernel the bandwith is divided by 2. So 2*5Mbps = 10Mbps instead of
> > > > > 2*10Mbps=20Mbps.
> > > > > I try to fix the pbm.
> > > >
> > > > perhaps a bug exists on the dispatcher when 2 identical cards exist.
> > > > Anyone have 2 identical cards for test ?
> > >
> > > After other tries, the ne.c file is buggy. Confirmation.
> > > I investigate. Anyone helps me ?
> >
> > Do you have evidence of a specific problem?
> > Or the same hardware running faster with other drivers or kernel versions?
> >
> > This sounds as if you are just running out of ISA bus bandwidth...
>
> My configuration : a server with 2 cards NE2000(10 Mbps), a machine with
> 2 cards : 1 NE2000(ISA) and 1 3C905b(PCI).
> server : NE2000 <-> workstation : NE2000 in 192.168.1.X
> server : NE2000 <-> workstation : 3C905b in 192.168.0.X
>
> I try with a 2.4.6, 2.4.15, and 2.5.31. there is near from no
> differences in the code of ne.c, gkrellm tells me that when I download
> on both cards big files (for tests), the bandwidth falls to 5Mbps for
> each card.
> If 1 download stops, the other bandwidth is up to 10Mbps immediatly.
>
> No other ISA card is used on the server.
ISA has a maximum bandwidth usually about a megabyte per second in PIO
(io-port, which the NE2k uses) mode. That's 10MBps. With two cards, i
gets evenly distributed. What's so surprising here?
--
Vojtech Pavlik
SuSE Labs
> ISA has a maximum bandwidth usually about a megabyte per second in PIO
> (io-port, which the NE2k uses) mode. That's 10MBps. With two cards, i
> gets evenly distributed. What's so surprising here?
>
> --
> Vojtech Pavlik
> SuSE Labs
ok, thanks