2006-10-16 08:43:34

by Cornelia Huck

[permalink] [raw]
Subject: [Patch 0/3] Driver core: Some probing changes

Hi,

the following patches attempt to fix some issues in the current device
probing code:

[1/3] Don't stop probing on ->probe errors.
[2/3] Change function call order in device_bind_driver().
[3/3] Per-subsystem multithreaded probing.

Patches are against -gkh tree. Works for me on s390 and on i386 with
pci multithreaded probing enabled. (I also enabled multithreaded
probing on the css and ccw busses in order to test the code on s390,
but this doesn't make much sense since we already do async device
recognition, so I'm not sending a patch.)


2006-10-16 14:39:30

by Alan Stern

[permalink] [raw]
Subject: Re: [Patch 0/3] Driver core: Some probing changes

On Mon, 16 Oct 2006, Cornelia Huck wrote:

> Hi,
>
> the following patches attempt to fix some issues in the current device
> probing code:
>
> [1/3] Don't stop probing on ->probe errors.
> [2/3] Change function call order in device_bind_driver().
> [3/3] Per-subsystem multithreaded probing.
>
> Patches are against -gkh tree. Works for me on s390 and on i386 with
> pci multithreaded probing enabled. (I also enabled multithreaded
> probing on the css and ccw busses in order to test the code on s390,
> but this doesn't make much sense since we already do async device
> recognition, so I'm not sending a patch.)

Your patch 2 looks fine.

Patch 1 is somewhat questionable. Certainly the log message reporting the
error should be left in. The other issue is whether to continue with
probing other drivers. I guess there's no reason not to; stopping short
was merely an optimization.

I'll discuss patch 3 in a separate message.

Alan Stern

2006-10-16 14:58:45

by Cornelia Huck

[permalink] [raw]
Subject: Re: [Patch 0/3] Driver core: Some probing changes

On Mon, 16 Oct 2006 10:39:28 -0400 (EDT),
Alan Stern <[email protected]> wrote:

> Patch 1 is somewhat questionable. Certainly the log message reporting the
> error should be left in.

OK, that makes sense. I'll do an updated patch.

> The other issue is whether to continue with
> probing other drivers. I guess there's no reason not to; stopping short
> was merely an optimization.

And not stopping also solves the "broken driver prevents any other
driver from matching" problem.

--
Cornelia Huck
Linux for zSeries Developer
Tel.: +49-7031-16-4837, Mail: [email protected]