[PATCH] rules: avoid reading DRIVER for devices without modalias

[PATCH] rules: avoid reading DRIVER for devices without modalias

[Alan Jenkins]

DRIVER requires a readlink() in sysfs.  The MODALIAS key is provided by
uevent and requires no additional read.  Test MODALIAS first.

Signed-off-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>

diff --git a/rules/rules.d/80-drivers.rules b/rules/rules.d/80-drivers.rules
index 7e13e80..2b68ddc 100644
--- a/rules/rules.d/80-drivers.rules
+++ b/rules/rules.d/80-drivers.rules
@@ -2,7 +2,7 @@
 
 ACTION!="add", GOTO="drivers_end"
 
-DRIVER!="?*", ENV{MODALIAS}="?*", RUN{ignore_error}+="/sbin/modprobe -b $env{MODALIAS}"
+ENV{MODALIAS}="?*", DRIVER!="?*", RUN{ignore_error}+="/sbin/modprobe -b $env{MODALIAS}"
 SUBSYSTEM="tifm", ENV{TIFM_CARD_TYPE}="SD", RUN+="/sbin/modprobe -b tifm_sd"
 SUBSYSTEM="tifm", ENV{TIFM_CARD_TYPE}="MS", RUN+="/sbin/modprobe -b tifm_ms"
 SUBSYSTEM="memstick", RUN+="/sbin/modprobe -b --all ms_block mspro_block"

Re: [PATCH] rules: avoid reading DRIVER for devices without modalias

[Kay Sievers]

On Wed, Jun 3, 2009 at 16:51, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:
> DRIVER requires a readlink() in sysfs.  The MODALIAS key is provided by
> uevent and requires no additional read.  Test MODALIAS first.

Are you sure that makes a difference? We *should* sort the keys while
parsing, and run the cheaper ones before the more expensive ones. At
least that's how it was before the rewrite of the parser, and it
*should* be still the same today.

Thanks,
Kay

Re: [PATCH] rules: avoid reading DRIVER for devices without modalias

[Alan Jenkins]

Kay Sievers wrote:
> On Wed, Jun 3, 2009 at 16:51, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:
>   
>> DRIVER requires a readlink() in sysfs.  The MODALIAS key is provided by
>> uevent and requires no additional read.  Test MODALIAS first.
>>     
>
> Are you sure that makes a difference? We *should* sort the keys while
> parsing, and run the cheaper ones before the more expensive ones. At
> least that's how it was before the rewrite of the parser, and it
> *should* be still the same today.
>
> Thanks,
> Kay
>   


Sorry, I'd forgotten that.  You're right it makes no difference.

Re: [PATCH] rules: avoid reading DRIVER for devices without

[Marco d'Itri]

On Jun 03, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:

> DRIVER requires a readlink() in sysfs.  The MODALIAS key is provided by
> uevent and requires no additional read.  Test MODALIAS first.
What is the point of checking both DRIVER and ENV{MODALIAS}?
The Debian rules files always checked only ENV{MODALIAS}.

-- 
ciao,
Marco

Re: [PATCH] rules: avoid reading DRIVER for devices without modalias

[Alan Jenkins]

Marco d'Itri wrote:
> On Jun 03, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:
>
>   
>> DRIVER requires a readlink() in sysfs.  The MODALIAS key is provided by
>> uevent and requires no additional read.  Test MODALIAS first.
>>     
> What is the point of checking both DRIVER and ENV{MODALIAS}?
> The Debian rules files always checked only ENV{MODALIAS}.
>   
I think the idea is to avoid calling modprobe if the driver was built 
into the kernel (or the module is already loaded).

Alan

Re: [PATCH] rules: avoid reading DRIVER for devices without

[Marco d'Itri]

On Aug 16, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:

>> What is the point of checking both DRIVER and ENV{MODALIAS}?
>> The Debian rules files always checked only ENV{MODALIAS}.
Nevermind, this is not true...

-- 
ciao,
Marco