Next-in-Thread Next Message

Warning Change in kernel (2.3.99-pre3) causing problem with cardmgr. 

Forum: PCMCIA driver development
Date: 2000, Apr 06
From: Brian Mays brian

There is a change in the kernel somewhere that is causing a problem: if 2.3.99-pre1 is used things work correctly, but when 2.3.99-pre3 is used cardmgr does things wrong.

Looking at the 2.3.99-pre3 patch, it is apparent that driver has been modified to use the new PCI scheme. I think that this is causing the problem.

Here are the relevant parts of the syslog. In this example, the user is trying to use a network card. Note that "network start 3c575_cb" is run by cargmgr instead of "network start eth0".

  Mar 30 14:28:37 fog kernel: cs: IO port probe 0x0c00-0x0cff: excluding 0xcf8-0xcff
  Mar 30 14:28:37 fog kernel: cs: IO port probe 0x0800-0x08ff: clean.
  Mar 30 14:28:37 fog kernel: cs: IO port probe 0x0100-0x04ff: excluding 0x280-0x287 0x378-0x37f 0x4d0-0x4d7
  Mar 30 14:28:37 fog kernel: cs: IO port probe 0x0a00-0x0aff: clean.
  Mar 30 14:28:38 fog cardmgr[158]: module /lib/modules/2.3.99-pre3/pcmcia/cb_enabler.o not available
  Mar 30 14:28:38 fog cardmgr[158]: executing: 'insmod /lib/modules/2.3.99-pre3/pcmcia/3c575_cb.o'
  Mar 30 14:28:38 fog kernel: 3c575_cb.c:v0.99L+LK1.1.2  3/19/2000  Donald Becker and others  http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html
  Mar 30 14:28:38 fog kernel: eth0: 3Com PCI 3CCFE575CT Cyclone CardBus at 0x1000, PCI: Increasing latency timer of device 02:00.0 to 64
  Mar 30 14:28:38 fog kernel:  00:50:da:85:19:47, IRQ 11
  Mar 30 14:28:38 fog kernel: eth0: CardBus functions mapped 10800080->c8824080
  Mar 30 14:28:38 fog kernel:   8K byte-wide RAM 5:3 Rx:Tx split, MII interface.
  Mar 30 14:28:38 fog kernel:   MII transceiver found at address 0, status 7809.
  Mar 30 14:28:38 fog kernel:   Enabling bus-master transmits and whole-frame receives.
  Mar 30 14:28:38 fog cardmgr[158]: executing: './network start 3c575_cb'
                                                                ^^^^^^^^
  Mar 30 14:28:38 fog cardmgr[158]: + 3c575_cb: error fetching interface information: Device not found
  Mar 30 14:28:38 fog cardmgr[158]: + 3c575_cb: unknown interface.
  Mar 30 14:28:38 fog pumpd[185]: starting at Thu Mar 30 14:28:38 2000  
  Mar 30 14:28:38 fog cardmgr[158]: + Operation failed.

Next-in-Thread Next Message

Messages Inline: 0 1

None Unfortunately there is no easy fix; the API is broken

Re: Warning Change in kernel (2.3.99-pre3) causing problem with cardmgr. (Brian Mays)
Date: 2000, Apr 06
From: David Hinds <dhinds@pcmcia.sourceforge.org>

This is a "feature" of the new hot plug API.  There is no workaround
at this time, because there is no mechanism in the new API to pass the
correct device name to cardmgr.

-- Dave

Sad Re: Unfortunately there is no easy fix; the API is broken

Re: None Unfortunately there is no easy fix; the API is broken (David Hinds)
Date: 2000, Apr 20
From: Steven A. DuChene <sad@valinux.com>

Hmm, so does that mean that unless things change drastically one way or the other, laptop users with PCMCIA cards will not be able to migrate to 2.4 when it is released?

This reasoning sounds remarkably similar to the bad feelings surrounding the network driver changes in the late 2.3 kernels and the Donald Becker crafted generic network drivers (like the eepro100 for instance).

I hope this does not degenerate into a similar situation.

Isn't there some sort of compromise that could be reached?

None I don't think the problem is as bad as all that

Re: Sad Re: Unfortunately there is no easy fix; the API is broken (Steven A. DuChene)
Date: 2000, Apr 20
From: David Hinds <dhinds@pcmcia.sourceforge.org>

No bad feelings are involved, as far as I'm concerned.  And there is
nothing preventing laptop users from migrating to 2.4 whenever it is
released.

I still distribute PCMCIA drivers that build correctly on 2.3.99* and
work correctly.  Linus and others made a choice to not include some of
that code in the kernel; I pointed out what the issues were and he
made a decision that I disagreed with, but I respect his decision.  I
think it would be very unfair to criticize me for those choices, and
Linus will tell you the same thing.

Compared to the situation with Donald Becker, I think there are a few
differences.  I'm not fighting a "turf war" here, and I'm not blowing
off PCMCIA because I didn't get my way.  I continue to release timely
updates and continue to give feedback and send contributions to Linus
and other kernel developers.

I said there was no easy fix, because there really is no easy fix, not
because I'm holding out on anyone.  It is not unsolvable, but it also
is not something I can fix on my own (and my solution has already been
explicitly rejected).

-- Dave
Change in kernel (2.3.99-pre3) causing problem with cardmgr.


Add Message to: "Change in kernel (2.3.99-pre3) causing problem with cardmgr."

Members Subscribe Admin Mode Show Frames Help for HyperNews at pcmcia-cs.sourceforge.net 1.10