Re: [PATCH][2.6] de4x5 cleanup

Re: [PATCH][2.6] de4x5 cleanup

[Jeff Garzik]

Marc Zyngier wrote:
> Jeff, Andrew,
> 
> The enclosed patch tries to clean the tulip de4x5 driver, and update
> it to some of the 2.6 APIs :
> 
> - Use PCI and EISA probing APIs
> - Use generic DMA API
> - Fix DE425 init on the Jensen
> - Remove de4x5 from Space.c
> 
> It's been tested on x86 and Alpha, with a DE425 (21040, EISA), a DE435
> (21040, PCI) and a quad DLink (4*21143, PCI).
> 
> The major problem with this patch is that, because of the Space.c
> removal, interfaces get potentially renumbered. This has been
> discussed to death on lkml, without any obvious solution. IMHO, we'd
> better remove as many drivers as we can from Space.c before
> 2.6.0-final ships...


It looks pretty good.  My big objection isn't the Space.c stuff but 
de4x5 itself.  I had hoped it would be gone by now :)  Oh well, not your 
fault.

A quick review of the patch looks good, and this is work that has long 
wanted doing (if de4x5 wasn't simply erased).  I'll apply after a more 
thorough review.

	Jeff

Re: [PATCH][2.6] de4x5 cleanup

[Marc Zyngier]

>>>>> "Jeff" == Jeff Garzik <jgarzik@pobox.com> writes:

Jeff> It looks pretty good.  My big objection isn't the Space.c stuff
Jeff> but de4x5 itself.  I had hoped it would be gone by now :) Oh
Jeff> well, not your fault.

Well... I too, had hoped the tulip (and its dc21040 derivative)
driver would be the only one. While it works nicely on modern stuff
(2114[023]), it has some problems on a few of my test boxes (mostly
Alphas with onboard 21040).

And most of all, tulip doesn't support EISA :-).

Thanks,

        M.
-- 
Places change, faces change. Life is so very strange.

Re: [PATCH][2.6] de4x5 cleanup

[Jeff Garzik]

Marc Zyngier wrote:
>>>>>>"Jeff" == Jeff Garzik <jgarzik@pobox.com> writes:
> 
> 
> Jeff> It looks pretty good.  My big objection isn't the Space.c stuff
> Jeff> but de4x5 itself.  I had hoped it would be gone by now :) Oh
> Jeff> well, not your fault.
> 
> Well... I too, had hoped the tulip (and its dc21040 derivative)
> driver would be the only one. While it works nicely on modern stuff
> (2114[023]), it has some problems on a few of my test boxes (mostly
> Alphas with onboard 21040).

Well, 2.6 tulip intentionally doesn't work at all on 21040 and 21041 :) 
  That's what de2104x driver is for.  Can you give that a beating, and 
verify that it works for you?

I know that 2114x chips with sym phys still have some issues; jamal has 
created a patch that fixes that (mostly updates tulip with the latest 
Becker phy code from Donald's tulip.c), which in turn fixes some 2114x 
issues on alpha.


> And most of all, tulip doesn't support EISA :-).

If it meant I could kill de4x5, then I would be all for adding EISA 
support to tulip... :)

	Jeff

Re: [PATCH][2.6] de4x5 cleanup

[Marc Zyngier]

>>>>> "Jeff" == Jeff Garzik <jgarzik@pobox.com> writes:

Jeff> Well, 2.6 tulip intentionally doesn't work at all on 21040 and
Jeff> 21041 :) That's what de2104x driver is for.  Can you give that a
Jeff> beating, and verify that it works for you?

Last time I checked, I had ugly Oopses on one of my alpha boxes
(AS-255), but that whas a few month ago, and I had other stuff to fix
at the moment. I'll give it a go tomorrow.

Jeff> If it meant I could kill de4x5, then I would be all for adding
Jeff> EISA support to tulip... :)

I'm afraid this would introduce a lot of the ugly crap that is
currently in de4x5 (21040 is seen through an EISA to PCI bridge, with
a sparse mapping)... I'll have a look at what we can do without
introduce too many hacks.

If de2104x behaves correctly, another solution would be to turn de4x5
to be EISA only. That would be a good way to finally kill it ;-).

Thanks,

        M.
-- 
Places change, faces change. Life is so very strange.

de2104x (was Re: [PATCH][2.6] de4x5 cleanup)

[Marc Zyngier]

>>>>> "Jeff" == Jeff Garzik <jgarzik@pobox.com> writes:

Jeff> Well, 2.6 tulip intentionally doesn't work at all on 21040 and
Jeff> 21041 :) That's what de2104x driver is for.  Can you give that a
Jeff> beating, and verify that it works for you?

So, as promised, I tried de2104x on the AS-255 (which was problematic
a few months ago). I must say it works nicely, as long as I don't
unplug the cable...

Here are the details :

otherside:~# lspci
00:06.0 SCSI storage controller: LSI Logic / Symbios Logic (formerly NCR) 53c810 (rev 11)
00:07.0 Non-VGA unclassified device: Intel Corp. 82378IB [SIO ISA Bridge] (rev 43)
00:0d.0 VGA compatible controller: Digital Equipment Corporation PBXGB [TGA2] (rev 22)
00:0e.0 Ethernet controller: Digital Equipment Corporation DECchip 21040 [Tulip] (rev 26)
otherside:~# ethtool eth0
Settings for eth0:
        Supported ports: [ TP AUI ]
        Supported link modes:   10baseT/Half 10baseT/Full 
        Supports auto-negotiation: No
        Advertised link modes:  10baseT/Half 10baseT/Full 
        Advertised auto-negotiation: No
        Speed: Unknown! (5)
        Duplex: Half
        Port: AUI
        PHYAD: 0
        Transceiver: internal
        Auto-negotiation: on
        Current message level: 0x000000f7 (247)
        Link detected: yes

Note that an AUI port is reported, while the system really has only a
10baseT port.

while booting, everything looks OK :
[...]
de2104x PCI Ethernet driver v0.5.4 (Jan 1, 2002)
eth0: 21040 at 0xfffffc0301209000, 00:00:25:4a:21:3a, IRQ 15
[...]
eth0: enabling interface
eth0: set link 10baseT-HD
eth0:    mode 0xfffc0040, sia 0xffffffcd,0xffff8f01,0xffffffff,0xffff0000
eth0:    set mode 0xfffc0000, set sia 0x8f01,0xffff,0x0
eth0: link up, media 10baseT-HD

It works perfectly, until I unplug the cable (I know, I'm a bad
boy... ;-), wait 10 seconds, then plug it back :

eth0: link down
eth0: set link AUI
eth0:    mode 0xfffc0000, sia 0xffffffc1,0xffff8f09,0xffff0705,0xffff0006
eth0:    set mode 0xfffc0000, set sia 0x8f09,0x705,0x6
eth0: link up, media AUI

So link is now up, on a non-existant AUI port ?!?!
Let's inform the driver that we want the TP port instead :

otherside:~# ethtool -s eth0 port tp
Cannot set new settings: Invalid argument
  not setting port

So 'tp' is invalid ? Interesting... Let's try the old ifconfig down/up
method :

eth0: disabling interface
eth0: enabling interface
eth0: set link AUI
eth0:    mode 0xfffc0040, sia 0xffffffc1,0xffff8f09,0xffff0705,0xffff0006
eth0:    set mode 0xfffc0000, set sia 0x8f09,0x705,0x6
eth0: link up, media AUI

No effect... Let's try it again, for fun :

eth0: disabling interface
eth0: timeout expired stopping DMA
eth0: enabling interface
eth0: set link AUI
eth0:    mode 0xfffc0040, sia 0xffffffc1,0xffff8f09,0xffff0705,0xffff0006
eth0:    set mode 0xfffc0000, set sia 0x8f09,0x705,0x6
eth0: link up, media AUI

Problem stopping DMA ? Frightening ! :-) Finally, I had to reboot the
system to get the the network back up (since modules are not really
useable currently on alpha, but that's another story).

Now, the same thing with the ugly de4x5 :

[...]
0000:00:0e.0: DE434/5 at 0x8400, h/w address 00:00:f8:25:4a:21,
      and requires IRQ15 (provided by PCI BIOS).
de4x5.c:V0.546 2001/02/22 davies@maniac.ultranet.com
[...]
eth0: media is TP.

Unplug the cable, wait 10 seconds, plug it back :

eth0: media is unconnected, link down or incompatible connection.
eth0: media is TP.

Works as expected.

While this problem is a corner case (the de2104x driver itself works
fine), it would be nice if we could solve this issue (loosing network
connectivity because someone rebooted the switch is a bad
behaviour). I promise this machine wont run de4x5 ever again once it
is solved :-).

Thanks,

        M.
-- 
Places change, faces change. Life is so very strange.