PLEASE HELP! *** DANGER *** unable to remove socket power (***PART1***)

Russell King rmk+pcmcia at
Fri Nov 17 18:48:09 EST 2006

On Fri, Nov 17, 2006 at 02:53:47PM -0500, EX500R wrote:
>   /*
>    * Some laptops (IBM T22) do not like us putting the Cardbus
>    * bridge into D3.  At a guess, some other laptop will
>    * probably require this, so leave it commented out for now.
>    */
>    /* pci_set_power_state(dev, 3); */
>  }
>  return ret;
> }
> Apparently, my laptop DOES need to set the bridge into D3!!! I uncommented
> that line and everything starts working.

That's interesting.

> Can you explain how does it work in your community?

We generate patches which are peer reviewed by the subsystem maintainer.
They are then put out for testing (normally in the -mm tree but may be
elsewhere, eg via a mailing list such as this one), and after a time
where no negative feedback has been received, they're submitted to

> It seems to me, that someone found that putting cardbus bridge into D3 on
> his particular machine does not work, commented out that line (at the same
> time, making a lot of other laptops stop working) and committed the changes
> into the main source branch. Is it how that should work? Then, I think that
> there is something really-really wrong.

No.  Someone reported a bug and investigated.  It was found that removing
that line fixed the problem, and a patch was created, and tested.  No
regressions were reported, so eventually it found its way into the -rc
kernels where it was tested more widely.  Again, no regressions were
reported, so it remains in the kernel tree.

So your assertion that it breaks "a lot of other laptops" is apparantly

However, since you've now found a problem with it, there needs to be a
view taken upon which is the least problematical of the two problems.
Or to put it another way - if we enable the D3 state, do we have more
breakage than if we keep it disabled.

Since you're the first to report this I suggest that it's far _safer_ to
keep the code as-is.

However, I no longer maintain PCMCIA, so it's really up to Dominik to
work out what he wants to do about this problem.

Russell King
 Linux kernel    2.6 ARM Linux   -
 maintainer of:  2.6 Serial core

More information about the linux-pcmcia mailing list