All of lore.kernel.org
 help / color / mirror / Atom feed
* [parisc-linux] C3000 and 2.6
@ 2003-11-20 18:52 M. Grabert
  2003-11-20 19:02 ` Matthew Wilcox
  2003-11-27  3:07 ` [parisc-linux] C3000 and 2.6 - need config help M. Grabert
  0 siblings, 2 replies; 15+ messages in thread
From: M. Grabert @ 2003-11-20 18:52 UTC (permalink / raw)
  To: parisc-linux

Hi,

before I'm wasting my time compiling the kernel, is the
serial console now working with 2.6?

There used to be problems with the SCSI driver aswell (would complain
that termination problem was present, but AFAIK this was already fixed)?


Right now I'm using a 120GB IDE using the onboard IDE of the C3k, and
I get just about 3MB/s with 2.4.22-paX, which might indicate that DMA is
not supported (I turned DMA/32bit on with hdparm, but no effects). I also
doubt that the driver for the onboard IDE chipset is much better in the
2.6 kernel, so I opted to install a IDE PCI card.

Unfortunately the only new PCI IDE cards you can get in the shops these
days that are ones with the Promise or the SiliconImage chipset;
suppor for the Promise chipset under linux used to be crap, and
SiliconImage is only suppoted with 2.6, but at least it works, so I went
for the latter. Therefore I'm eager to get 2.6 running!


Thanks,
  Max


PS: Building network scheduling (QoS) as modules results in unresolved
    symbols (using 2.4.22-pa13). Building then in the kernel (2.4.22-pa17)
    works.

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6
  2003-11-20 18:52 [parisc-linux] C3000 and 2.6 M. Grabert
@ 2003-11-20 19:02 ` Matthew Wilcox
  2003-11-20 19:24   ` M. Grabert
       [not found]   ` <200311210820.42414.pacman@leovey-szki.sulinet.hu>
  2003-11-27  3:07 ` [parisc-linux] C3000 and 2.6 - need config help M. Grabert
  1 sibling, 2 replies; 15+ messages in thread
From: Matthew Wilcox @ 2003-11-20 19:02 UTC (permalink / raw)
  To: M. Grabert; +Cc: parisc-linux

On Thu, Nov 20, 2003 at 06:52:50PM +0000, M. Grabert wrote:
> before I'm wasting my time compiling the kernel, is the
> serial console now working with 2.6?

Not really.

> There used to be problems with the SCSI driver aswell (would complain
> that termination problem was present, but AFAIK this was already fixed)?

Yes, I found that bug a few weeks ago.

> Right now I'm using a 120GB IDE using the onboard IDE of the C3k, and
> I get just about 3MB/s with 2.4.22-paX, which might indicate that DMA is
> not supported (I turned DMA/32bit on with hdparm, but no effects). I also
> doubt that the driver for the onboard IDE chipset is much better in the
> 2.6 kernel, so I opted to install a IDE PCI card.

Yeah ... I wouldn't recommend using the internal chipset for writable
data ;-)

> PS: Building network scheduling (QoS) as modules results in unresolved
>     symbols (using 2.4.22-pa13). Building then in the kernel (2.4.22-pa17)
>     works.

... which symbols?

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6
  2003-11-20 19:02 ` Matthew Wilcox
@ 2003-11-20 19:24   ` M. Grabert
  2003-11-20 19:45     ` Matthew Wilcox
       [not found]   ` <200311210820.42414.pacman@leovey-szki.sulinet.hu>
  1 sibling, 1 reply; 15+ messages in thread
From: M. Grabert @ 2003-11-20 19:24 UTC (permalink / raw)
  To: Matthew Wilcox; +Cc: parisc-linux

On Thu, 20 Nov 2003, Matthew Wilcox wrote:

> > PS: Building network scheduling (QoS) as modules results in unresolved
> >     symbols (using 2.4.22-pa13). Building then in the kernel (2.4.22-pa17)
> >     works.
>
> ... which symbols?

I compiled -pa17 with the same config as -pa13, and the unresolved symbols
in combination of QoS are gone now (of which there were many).

However I have another one here (albeit unrelated to QoS, but IPv6 netfilter):

root@beast:~# depmod -a
depmod: *** Unresolved symbols in
/lib/modules/2.4.22-pa17/kernel/net/ipv6/netfilter/ip6_queue.o
depmod:         ip6_route_me_harder



Slan,
  Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6
  2003-11-20 19:24   ` M. Grabert
@ 2003-11-20 19:45     ` Matthew Wilcox
  2003-11-21  0:38       ` M. Grabert
  0 siblings, 1 reply; 15+ messages in thread
From: Matthew Wilcox @ 2003-11-20 19:45 UTC (permalink / raw)
  To: M. Grabert; +Cc: Matthew Wilcox, parisc-linux

On Thu, Nov 20, 2003 at 07:24:49PM +0000, M. Grabert wrote:
> On Thu, 20 Nov 2003, Matthew Wilcox wrote:
> 
> > > PS: Building network scheduling (QoS) as modules results in unresolved
> > >     symbols (using 2.4.22-pa13). Building then in the kernel (2.4.22-pa17)
> > >     works.
> >
> > ... which symbols?
> 
> I compiled -pa17 with the same config as -pa13, and the unresolved symbols
> in combination of QoS are gone now (of which there were many).
> 
> However I have another one here (albeit unrelated to QoS, but IPv6 netfilter):
> 
> root@beast:~# depmod -a
> depmod: *** Unresolved symbols in
> /lib/modules/2.4.22-pa17/kernel/net/ipv6/netfilter/ip6_queue.o
> depmod:         ip6_route_me_harder

How strange.

$ find -type f |xargs grep ip6_route_me_harder
./include/net/ip6_route.h:extern int                    ip6_route_me_harder(struct sk_buff *skb);
./net/netsyms.c:EXPORT_SYMBOL(ip6_route_me_harder);

Are you sure the kernel you are currently running was compiled with
CONFIG_NETFILTER and CONFIG_IPV6 turned on?

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6
  2003-11-20 19:45     ` Matthew Wilcox
@ 2003-11-21  0:38       ` M. Grabert
  0 siblings, 0 replies; 15+ messages in thread
From: M. Grabert @ 2003-11-21  0:38 UTC (permalink / raw)
  To: Matthew Wilcox; +Cc: parisc-linux

On Thu, 20 Nov 2003, Matthew Wilcox wrote:

> > However I have another one here (albeit unrelated to QoS, but IPv6 netfilter):
> >
> > root@beast:~# depmod -a
> > depmod: *** Unresolved symbols in
> > /lib/modules/2.4.22-pa17/kernel/net/ipv6/netfilter/ip6_queue.o
> > depmod:         ip6_route_me_harder
>
> How strange.
>
> $ find -type f |xargs grep ip6_route_me_harder
> ./include/net/ip6_route.h:extern int                    ip6_route_me_harder(struct sk_buff *skb);
> ./net/netsyms.c:EXPORT_SYMBOL(ip6_route_me_harder);
>
> Are you sure the kernel you are currently running was compiled with
> CONFIG_NETFILTER and CONFIG_IPV6 turned on?

Definitely.


xam@sal:~$ ping6 -c3 beast.ipv6.damnyou.org
PING beast.ipv6.damnyou.org(beast.ipv6.damnyou.org) 56 data bytes
64 bytes from beast.ipv6.damnyou.org: icmp_seq=1 ttl=63 time=60.1 ms
64 bytes from beast.ipv6.damnyou.org: icmp_seq=2 ttl=63 time=39.0 ms
64 bytes from beast.ipv6.damnyou.org: icmp_seq=3 ttl=63 time=39.6 ms

--- beast.ipv6.damnyou.org ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2018ms
rtt min/avg/max/mdev = 39.076/46.302/60.183/9.819 ms


(beast.ipv6.damnyou.org is the parisc box running 2.4.22-pa17)
Proof that IPv6 netfilter is working:


root@beast:/home/xam# ip6tables -L FORWARD
Chain FORWARD (policy DROP)
target     prot opt source               destination
DROP       udp      anywhere             anywhere           multiport
dports 135,1025
DROP       tcp      anywhere             anywhere           multiport
dports 135,1025
[...]


and it works as expected ...


greetings, Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* [parisc-linux] Re: HP9000-K460,K420,Dx70/2
       [not found]   ` <200311210820.42414.pacman@leovey-szki.sulinet.hu>
@ 2003-11-21 12:47     ` Matthew Wilcox
  0 siblings, 0 replies; 15+ messages in thread
From: Matthew Wilcox @ 2003-11-21 12:47 UTC (permalink / raw)
  To: pacman; +Cc: Matthew Wilcox, parisc-linux

Please don't send me personal mail about installation problems; others
are equally capable of answering these questions and they may respond
when I can't.

On Fri, Nov 21, 2003 at 08:20:42AM +0100, Paczkó András wrote:
> My company have three HP9000, a K460, a K420, a Dx70/2.
> I have to install Linux (or *BSD) on this machine.
> I probe on K460,K420 becuse the Dx have a fan problem.

As far as I know, BSD does not support the K class.

> I download the debianhppa_NONUS.iso-t and the latest (PA-13) netinst iso from 
> pateam.
> 
> I got an idea: "You need a boot param, if your box have an ncr53c8xx, 
> ncr53c8xx=burst:3..."
> 
> My boxes have ncr53c710, a tried it, but not works, (burst:1..2..3)

That's obsolete advice; the problem was fixed in 2.4.16-pa14.

> the boot fail:
> 
> serial .... with MANY PORTS....
> bla bla
> 
> dump....
> 
> Is anyone who are able to install linux these boxes?

Yes, I and others have successfully installed Linux on these boxes.  It
sounds like you need to interact with IPL and change console=ttyS0 to
console=ttyB0, but you haven't provided enough information for me to
be certain.

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

^ permalink raw reply	[flat|nested] 15+ messages in thread

* [parisc-linux] C3000 and 2.6 - need config help
  2003-11-20 18:52 [parisc-linux] C3000 and 2.6 M. Grabert
  2003-11-20 19:02 ` Matthew Wilcox
@ 2003-11-27  3:07 ` M. Grabert
  2003-11-27  4:21   ` Grant Grundler
  1 sibling, 1 reply; 15+ messages in thread
From: M. Grabert @ 2003-11-27  3:07 UTC (permalink / raw)
  To: parisc-linux

On Thu, 20 Nov 2003, M. Grabert wrote:

> Right now I'm using a 120GB IDE using the onboard IDE of the C3k, and
> I get just about 3MB/s with 2.4.22-paX, which might indicate that DMA is
> not supported (I turned DMA/32bit on with hdparm, but no effects). I also
> doubt that the driver for the onboard IDE chipset is much better in the
> 2.6 kernel, so I opted to install a IDE PCI card.

The IDE harddisk is running for the last couple of months, and I had no
errors whatsoever (using the onboard IDE chipset of the C3k).

However it appears that writing and reading from/to the harddisk slows
down the access time quite a bit (ie. streaming a 180kbit/s movie and
writing a file with 1000kbit/s makes the movie to appear/play jerky);
whereas multiple reads doesn't affect it (ie. streaming serveral movies at
a time with 180kbit/s each).

> Unfortunately the only new PCI IDE cards you can get in the shops these
> days that are ones with the Promise or the SiliconImage chipset;
> suppor for the Promise chipset under linux used to be crap, and
> SiliconImage is only suppoted with 2.6, but at least it works, so I went
> for the latter. Therefore I'm eager to get 2.6 running!

Can anybody tell me what config options I need to boot a 2.6 kernel
without serial console support (since it's broken) and without
STI (since I don't have a graphics card in my C3k)?
I've seen a config option 'DUMMY_CONSOLE' in .config, but I don't find
how to select it in 'make menuconfig' or any description in Documentation/.

I have several machines that run headless - I use serial console then.
But I never had a situation where I didn't even have serial console ...


Thanks for any hints,
  Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27  3:07 ` [parisc-linux] C3000 and 2.6 - need config help M. Grabert
@ 2003-11-27  4:21   ` Grant Grundler
  2003-11-27  4:33     ` M. Grabert
  0 siblings, 1 reply; 15+ messages in thread
From: Grant Grundler @ 2003-11-27  4:21 UTC (permalink / raw)
  To: M. Grabert; +Cc: parisc-linux

On Thu, Nov 27, 2003 at 03:07:18AM +0000, M. Grabert wrote:
...
> The IDE harddisk is running for the last couple of months, and I had no
> errors whatsoever (using the onboard IDE chipset of the C3k).

Really! I'm amazed since last time I played with suckyio IDE it
barely worked for CD-ROM installs.

> However it appears that writing and reading from/to the harddisk slows
> down the access time quite a bit (ie. streaming a 180kbit/s movie and
> writing a file with 1000kbit/s makes the movie to appear/play jerky);
> whereas multiple reads doesn't affect it (ie. streaming serveral movies at
> a time with 180kbit/s each).

DMA is probably not enabled. This means all data is "manually" (ie by
the CPU) read or written to the IDE controller 16-bits at a time.
With interrupts turned off, mind you.

Enabling DMA would make a big difference in performance.

> Can anybody tell me what config options I need to boot a 2.6 kernel
> without serial console support (since it's broken) and without
> STI (since I don't have a graphics card in my C3k)?

Willy, tausq, and I hacked around on suckyio serial support.
I posted a preliminary patch to
	 ftp://ftp.parisc-linux.org/patches/diff-2.6.0-suckyio-serial-02

ISTR we have to reset the USB/LAN device via a firmware call *before*
messing with IO MMU (spurious DMA by USB or LAN will MCA the box).
The problem is this firmware call also blows away the serial console.

Work remaining:
1) determine when USB or LAN was left active by firmware.
2) only call firmware to reset all IO when USB or LAN was left active.
3) a) make an early call to init suckyio and it's serial port.
   b) or save/restore suckyio config space registers before/after
      the PDC_IO_RESET calls so PDC console continues to work.

> I've seen a config option 'DUMMY_CONSOLE' in .config, but I don't find
> how to select it in 'make menuconfig' or any description in Documentation/.
> 
> I have several machines that run headless - I use serial console then.
> But I never had a situation where I didn't even have serial console ...

yeah - the real problem is init sequence completely changed in 2.6 (vs 2.4).
We haven't reworked all the same quirks from 2.4 into 2.6 yet.

hth,
grant

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27  4:21   ` Grant Grundler
@ 2003-11-27  4:33     ` M. Grabert
       [not found]       ` <Pine.LNX.4.58.0311270444110.9783@sal.ucc.ie>
  2003-11-27 18:29       ` Joel Soete
  0 siblings, 2 replies; 15+ messages in thread
From: M. Grabert @ 2003-11-27  4:33 UTC (permalink / raw)
  To: Grant Grundler; +Cc: parisc-linux

On Wed, 26 Nov 2003, Grant Grundler wrote:

> On Thu, Nov 27, 2003 at 03:07:18AM +0000, M. Grabert wrote:
>
> > However it appears that writing and reading from/to the harddisk slows
> > down the access time quite a bit (ie. streaming a 180kbit/s movie and
> > writing a file with 1000kbit/s makes the movie to appear/play jerky);
> > whereas multiple reads doesn't affect it (ie. streaming serveral movies at
> > a time with 180kbit/s each).
>
> DMA is probably not enabled. This means all data is "manually" (ie by
> the CPU) read or written to the IDE controller 16-bits at a time.
> With interrupts turned off, mind you.
>
> Enabling DMA would make a big difference in performance.

See the old thread:

http://lists.parisc-linux.org/pipermail/parisc-linux/2003-August/020851.html
http://lists.parisc-linux.org/pipermail/parisc-linux/2003-August/020886.html

I've tried to enable DMA and unmask interrupts, but obviously it doesn't
work; at least it doesn't make any difference in performance.

> > I've seen a config option 'DUMMY_CONSOLE' in .config, but I don't find
> > how to select it in 'make menuconfig' or any description in Documentation/.
> >
> > I have several machines that run headless - I use serial console then.
> > But I never had a situation where I didn't even have serial console ...
>
> yeah - the real problem is init sequence completely changed in 2.6 (vs 2.4).
> We haven't reworked all the same quirks from 2.4 into 2.6 yet.

So is there current a way to get 2.6 working *without* serial console or STI ?

Thanks,
  Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
       [not found]       ` <Pine.LNX.4.58.0311270444110.9783@sal.ucc.ie>
@ 2003-11-27  6:15         ` Grant Grundler
  2003-11-27 15:16           ` M. Grabert
  0 siblings, 1 reply; 15+ messages in thread
From: Grant Grundler @ 2003-11-27  6:15 UTC (permalink / raw)
  To: M. Grabert; +Cc: parisc-linux

On Thu, Nov 27, 2003 at 04:45:49AM +0000, M. Grabert wrote:
> I forgot to mention my current IDE settings (unchanged since August):
> 
> root@beast:/home/xam# hdparm -c /dev/hda
> 
> /dev/hda:
>  IO_support   =  1 (32-bit)

The above info is either wrong or ignored.
"data" register for PC87560 is 16-bits only.
(IDE function is NS87415 compatible)

> root@beast:/home/xam# hdparm -d /dev/hda
> 
> /dev/hda:
>  using_dma    =  1 (on)

Given only 3MB/s, I find it hard to believe.
SuckyIO (PC87560UBD) Data sheet claims 16MB/s per channel. 

However, I just now wonder if the limiting factor is not the IDE if the
kernel is copying data to/from kernel to user space.  Our copy_to/from_user
could be alot more efficient (2x guesstimate) for the PA8500 CPU.
I see max bandwidth of 4-5MB/s on a PA8600 (552Mhz) over 100BT.

With a 32X CD-ROM drive in my c3k, "time dd" is only getting 925 KB/s.
I'll guess a few more factors could be muddying the waters here.

root@gggc3k:~# hdparm -i /dev/cdrom

/dev/cdrom:

 Model=CD-532E-B, FwRev=2.0A, SerialNo=
 Config={ Fixed Removeable DTR<=5Mbs DTR>10Mbs nonMagnetic }
 RawCHS=0/0/0, TrkSize=0, SectSize=0, ECCbytes=0
 BuffType=DualPortCache, BuffSize=128kB, MaxMultSect=0
 (maybe): CurCHS=0/0/0, CurSects=0, LBA=yes, LBAsects=0
 IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  sdma0 sdma1 sdma2 mdma0 mdma1 *mdma2 
 AdvancedPM=no

 * signifies the current active mode


root@gggc3k:~# hdparm /dev/cdrom

/dev/cdrom:
 HDIO_GET_MULTCOUNT failed: Invalid argument
 IO_support   =  0 (default 16-bit)
 unmaskirq    =  0 (off)
 using_dma    =  0 (off)
 keepsettings =  0 (off)
 readonly     =  1 (on)
 readahead    =  8 (on)
 HDIO_GETGEO failed: Invalid argument

Note the mdma2 and using_dma settings don't agree.
I tried to set using_dma:

root@gggc3k:~# hdparm -d 1 /dev/cdrom

/dev/cdrom:
 setting using_dma to 1 (on)
 HDIO_SET_DMA failed: Operation not permitted
 using_dma    =  0 (off)
root@gggc3k:~# uname -a
Linux gggc3k 2.4.22-pa17 #18 Fri Nov 21 21:16:31 PST 2003 parisc GNU/Linux


BTW, I checked that CONFIG_IDEDMA_ONLYDISK is off.

grant

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27  6:15         ` Grant Grundler
@ 2003-11-27 15:16           ` M. Grabert
  2003-11-27 18:15             ` Grant Grundler
  0 siblings, 1 reply; 15+ messages in thread
From: M. Grabert @ 2003-11-27 15:16 UTC (permalink / raw)
  To: Grant Grundler; +Cc: parisc-linux

On Wed, 26 Nov 2003, Grant Grundler wrote:

> However, I just now wonder if the limiting factor is not the IDE if the
> kernel is copying data to/from kernel to user space.  Our copy_to/from_user
> could be alot more efficient (2x guesstimate) for the PA8500 CPU.
> I see max bandwidth of 4-5MB/s on a PA8600 (552Mhz) over 100BT.

The limiting factor is definitely not the IDE hard disk, it's a new
Seagate 120GB which has a sustained data transfer rate of >30MB/s.

The max. transfer rate of any 100BT network is around 5MByte/s.

Personally I also get something between 4-5MB/s from any host to any
hosy in my home network. If I do a network transfer to the C3k (IDE hard disk),
I get about 4.5MB/s, but after a while the transfer stalls for a short
time, and then continues. However I see sustained 4.5MB/s to my C3k if I
write to any of the SCSI hard disks - without any stalls.

I suppose if it tries to write the data to the IDE hard disk (3.3MB/s),
it can't keep up with the speed of the network (4.5MB/s), so it buffers
the data in the RAM. At some point the buffer is full and the kernel
delays the networks transfer until the buffer content is written to the
IDE disk.


> With a 32X CD-ROM drive in my c3k, "time dd" is only getting 925 KB/s.
> I'll guess a few more factors could be muddying the waters here.

That's equivalent to a 5-6X CD-ROM. Ouch.

[...]

Quite a lot old IDE cdroms drives don't support dma transfers.
Hower you'll get the same message if you try to enable DMA if you have a
kernel that doesn't have a driver with DMA support for your IDE chipset.

I honestly don't thing DMA works ...


 Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27 15:16           ` M. Grabert
@ 2003-11-27 18:15             ` Grant Grundler
  2003-11-27 18:21               ` M. Grabert
  0 siblings, 1 reply; 15+ messages in thread
From: Grant Grundler @ 2003-11-27 18:15 UTC (permalink / raw)
  To: M. Grabert; +Cc: parisc-linux

On Thu, Nov 27, 2003 at 03:16:08PM +0000, M. Grabert wrote:
> The max. transfer rate of any 100BT network is around 5MByte/s.

100BT is capable of 90-95Mb/s ~= 11MB/s.
There's a website someplace that has 100BT performance numbers
for all sorts of card/OS combinations.

I've measured ftp transfers > 10MB/s between ia64 and ia32 machines
over 100BT link.

I'll take another look at why DMA isn't working.
It looks like NS87415 driver has the support in place.
The key bit seems to be init_hwif_ns87415().

grant

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27 18:15             ` Grant Grundler
@ 2003-11-27 18:21               ` M. Grabert
  2003-11-27 18:49                 ` Grant Grundler
  0 siblings, 1 reply; 15+ messages in thread
From: M. Grabert @ 2003-11-27 18:21 UTC (permalink / raw)
  To: Grant Grundler; +Cc: parisc-linux

On Thu, 27 Nov 2003, Grant Grundler wrote:

> On Thu, Nov 27, 2003 at 03:16:08PM +0000, M. Grabert wrote:
> > The max. transfer rate of any 100BT network is around 5MByte/s.
>
> 100BT is capable of 90-95Mb/s ~= 11MB/s.

That's theoretical, not really practically possible/feasible :-/

> There's a website someplace that has 100BT performance numbers
> for all sorts of card/OS combinations.
>
> I've measured ftp transfers > 10MB/s between ia64 and ia32 machines
> over 100BT link.

ftp? uncompressed? Well I'm impressed ...
The fastest 100MBit link I ever had was 7MBytes/s.

> I'll take another look at why DMA isn't working.
> It looks like NS87415 driver has the support in place.
> The key bit seems to be init_hwif_ns87415().

Thanks alot,

 Max

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27  4:33     ` M. Grabert
       [not found]       ` <Pine.LNX.4.58.0311270444110.9783@sal.ucc.ie>
@ 2003-11-27 18:29       ` Joel Soete
  1 sibling, 0 replies; 15+ messages in thread
From: Joel Soete @ 2003-11-27 18:29 UTC (permalink / raw)
  To: M. Grabert, Grant Grundler; +Cc: parisc-linux

Hi all,

> > I've seen a config option 'DUMMY_CONSOLE' in .config, but I don't find
> > how to select it in 'make menuconfig' or any description in Documentation/.
> >
> > I have several machines that run headless - I use serial console then.
> >
>But I never had a situation where I didn't even have serial console ...
>
> yeah - the real problem is init sequence completely changed in 2.6 (vs
2.4).
> We haven't reworked all the same quirks from 2.4 into 2.6 yet.

> So is there current a way to get 2.6 working *without* serial console or
STI ?

I rebuild a k 2.6.0-test11 with my gcc-3.3.2 64bits without SUPERIO and PDC
console. No pb to rebuild but it failled to boot on the b2k :( (and no console
meana no mean to analyse :( ). 

Sorry to have no more help, but for such kind of pb I read (some time ago
on lwn) that some one was working on a 'network' console for 2.6 (suporting
at this time tulipe's nic among others iirc). I try to contact him but never
get any feedback. Is somebody known the status of this patch in 2.6?

Thanks,
    Joel

-------------------------------------------------------------------------
Tiscali ADSL: 12 mois à 29,50 €/mois! L'Internet rapide, c'est pour tout
le monde.
http://reg.tiscali.be/default.asp?lg=fr

^ permalink raw reply	[flat|nested] 15+ messages in thread

* Re: [parisc-linux] C3000 and 2.6 - need config help
  2003-11-27 18:21               ` M. Grabert
@ 2003-11-27 18:49                 ` Grant Grundler
  0 siblings, 0 replies; 15+ messages in thread
From: Grant Grundler @ 2003-11-27 18:49 UTC (permalink / raw)
  To: M. Grabert; +Cc: parisc-linux

On Thu, Nov 27, 2003 at 06:21:26PM +0000, M. Grabert wrote:
> On Thu, 27 Nov 2003, Grant Grundler wrote:
> 
> > On Thu, Nov 27, 2003 at 03:16:08PM +0000, M. Grabert wrote:
> > > The max. transfer rate of any 100BT network is around 5MByte/s.
> >
> > 100BT is capable of 90-95Mb/s ~= 11MB/s.
> 
> That's theoretical, not really practically possible/feasible :-/

it's possible with a switched network. It's not possible
with regular hub where one has to deal with collisions.
Full duplex switched networks obsolete token passing as a method to
avoid packet collisions.

> > There's a website someplace that has 100BT performance numbers
> > for all sorts of card/OS combinations.
> >
> > I've measured ftp transfers > 10MB/s between ia64 and ia32 machines
> > over 100BT link.
> 
> ftp? uncompressed? Well I'm impressed ...
> The fastest 100MBit link I ever had was 7MBytes/s.

This is an excellent paper on routing:
	http://robur.slu.se/Linux/net-development/jamal/FF-html/

Peak throughput (in one port of a 4-port tulip card and out another)
is about 50Mb/s because the single CPU is saturated.  He does better
with an SMP machine. Jamal later got 20% perf improvement (measured in
packets per second) using MMIO instead of IO port space . But still
saturating the CPU. Jamal was confident at the time he could route at
link rate (95Mb/s) with 64byte packets on a faster SMP machine.

grant

^ permalink raw reply	[flat|nested] 15+ messages in thread

end of thread, other threads:[~2003-11-27 18:49 UTC | newest]

Thread overview: 15+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-11-20 18:52 [parisc-linux] C3000 and 2.6 M. Grabert
2003-11-20 19:02 ` Matthew Wilcox
2003-11-20 19:24   ` M. Grabert
2003-11-20 19:45     ` Matthew Wilcox
2003-11-21  0:38       ` M. Grabert
     [not found]   ` <200311210820.42414.pacman@leovey-szki.sulinet.hu>
2003-11-21 12:47     ` [parisc-linux] Re: HP9000-K460,K420,Dx70/2 Matthew Wilcox
2003-11-27  3:07 ` [parisc-linux] C3000 and 2.6 - need config help M. Grabert
2003-11-27  4:21   ` Grant Grundler
2003-11-27  4:33     ` M. Grabert
     [not found]       ` <Pine.LNX.4.58.0311270444110.9783@sal.ucc.ie>
2003-11-27  6:15         ` Grant Grundler
2003-11-27 15:16           ` M. Grabert
2003-11-27 18:15             ` Grant Grundler
2003-11-27 18:21               ` M. Grabert
2003-11-27 18:49                 ` Grant Grundler
2003-11-27 18:29       ` Joel Soete

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.