[PATCH 0/4] airo fixes

[PATCH 0/4] airo fixes

[Michal Schmidt]

Hello,

this is a series of four patches for the airo driver:

1/4: airo: don't use the interface name so much before registration
2/4: airo: delay parts of initialization until the netdev is up
3/4: airo: disable the PCI device when unloading module
4/4: airo: start with radio off

The first two fix the race condition Matteo Croce was experiencing.
The third one fixes a bug I noticed during testing.
The last one saves some power when the card is not used.

Michal

Re: [PATCH 0/4] airo fixes

[Dan Williams]

On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
> Hello,
> 
> this is a series of four patches for the airo driver:
> 
> 1/4: airo: don't use the interface name so much before registration
> 2/4: airo: delay parts of initialization until the netdev is up
> 3/4: airo: disable the PCI device when unloading module
> 4/4: airo: start with radio off
> 
> The first two fix the race condition Matteo Croce was experiencing.
> The third one fixes a bug I noticed during testing.
> The last one saves some power when the card is not used.

These all look good to me, I might be able to find some time to
runtime-test them too if you want more testing.

Dan

Re: [PATCH 0/4] airo fixes

[Michal Schmidt]

Dan Williams skrev:
> On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
>> Hello,
>>
>> this is a series of four patches for the airo driver:
>>
>> 1/4: airo: don't use the interface name so much before registration
>> 2/4: airo: delay parts of initialization until the netdev is up
>> 3/4: airo: disable the PCI device when unloading module
>> 4/4: airo: start with radio off
>>
>> The first two fix the race condition Matteo Croce was experiencing.
>> The third one fixes a bug I noticed during testing.
>> The last one saves some power when the card is not used.
> 
> These all look good to me, I might be able to find some time to
> runtime-test them too if you want more testing.
> 
> Dan

That would be nice. Thank you.
Michal

Re: [PATCH 0/4] airo fixes

[Matteo Croce]

> On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
> Hello,
>
> this is a series of four patches for the airo driver:
>
> 1/4: airo: don't use the interface name so much before registration
> 2/4: airo: delay parts of initialization until the netdev is up
> 3/4: airo: disable the PCI device when unloading module
> 4/4: airo: start with radio off
>
> The first two fix the race condition Matteo Croce was experiencing.
> The third one fixes a bug I noticed during testing.
> The last one saves some power when the card is not used.

I have applied your 4 patches on both a vanilla 2.6.21.5 kernel and
a 2.6.22-rc6 wireless-dev one, and i found that only patches 1/4 and 2/4 are 
good.
the 3/4 one gives this bogus output:

airo(): Probing for PCI adapters
PCI: Setting latency timer of device 0000:05:08.0 to 64
airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
airo(eth%d): Doing fast bap_reads
airo(): WPA is supported.
airo(eth0): MAC enabled 0:d:29:4f:c:be
airo(): Finished probing for PCI adapters

and the 4/4 definitely broke all:

airo(): Probing for PCI adapters
PCI: Setting latency timer of device 0000:05:08.0 to 64
airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
airo(eth%d): Doing fast bap_reads
airo(eth%d): Bad MAC enable reason = 2, rid = 0, offset = 0
airo(eth%d): MAC could not be enabled
airo(): Finished probing for PCI adapters

Re: [PATCH 0/4] airo fixes

[Michal Schmidt]

Matteo Croce skrev:
>> On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
>> Hello,
>>
>> this is a series of four patches for the airo driver:
>>
>> 1/4: airo: don't use the interface name so much before registration
>> 2/4: airo: delay parts of initialization until the netdev is up
>> 3/4: airo: disable the PCI device when unloading module
>> 4/4: airo: start with radio off
>>
>> The first two fix the race condition Matteo Croce was experiencing.
>> The third one fixes a bug I noticed during testing.
>> The last one saves some power when the card is not used.
> 
> I have applied your 4 patches on both a vanilla 2.6.21.5 kernel and
> a 2.6.22-rc6 wireless-dev one, and i found that only patches 1/4 and 2/4 are 
> good.

> the 3/4 one gives this bogus output:
> 
> airo(): Probing for PCI adapters
> PCI: Setting latency timer of device 0000:05:08.0 to 64
> airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
> airo(eth%d): Doing fast bap_reads
> airo(): WPA is supported.
> airo(eth0): MAC enabled 0:d:29:4f:c:be
> airo(): Finished probing for PCI adapters

I must have missed some pieces still printing the device name before 
registration. This should be just cosmetic. I understand the card still 
works in this case?

> and the 4/4 definitely broke all:
> 
> airo(): Probing for PCI adapters
> PCI: Setting latency timer of device 0000:05:08.0 to 64
> airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
> airo(eth%d): Doing fast bap_reads
> airo(eth%d): Bad MAC enable reason = 2, rid = 0, offset = 0
> airo(eth%d): MAC could not be enabled
> airo(): Finished probing for PCI adapters

That's worse. I'll see how it is possible.
Thanks for testing!

Michal

Re: [PATCH 0/4] airo fixes

[Matteo Croce]

> On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
> Hello,
>
> this is a series of four patches for the airo driver:
>
> 1/4: airo: don't use the interface name so much before registration
> 2/4: airo: delay parts of initialization until the netdev is up
> 3/4: airo: disable the PCI device when unloading module
> 4/4: airo: start with radio off
>
> The first two fix the race condition Matteo Croce was experiencing.
> The third one fixes a bug I noticed during testing.
> The last one saves some power when the card is not used.

I have applied your 4 patches on both a vanilla 2.6.21.5 kernel and
a 2.6.22-rc6 wireless-dev one, and i found that only patches 1/4 and 2/4 are 
good.
the 3/4 one gives this bogus output:

airo(): Probing for PCI adapters
PCI: Setting latency timer of device 0000:05:08.0 to 64
airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
airo(eth%d): Doing fast bap_reads
airo(): WPA is supported.
airo(eth0): MAC enabled 0:d:29:4f:c:be
airo(): Finished probing for PCI adapters

and the 4/4 definitely broke all:

airo(): Probing for PCI adapters
PCI: Setting latency timer of device 0000:05:08.0 to 64
airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
airo(eth%d): Doing fast bap_reads
airo(eth%d): Bad MAC enable reason = 2, rid = 0, offset = 0
airo(eth%d): MAC could not be enabled
airo(): Finished probing for PCI adapters

Re: [PATCH 0/4] airo fixes

[Matteo Croce]

Sorry, previous mail was repeated....

> > On Wed, 2007-06-27 at 23:14 +0200, Michal Schmidt wrote:
> > Hello,
> >
> > this is a series of four patches for the airo driver:
> >
> > 1/4: airo: don't use the interface name so much before registration
> > 2/4: airo: delay parts of initialization until the netdev is up
> > 3/4: airo: disable the PCI device when unloading module
> > 4/4: airo: start with radio off
> >
> > The first two fix the race condition Matteo Croce was experiencing.
> > The third one fixes a bug I noticed during testing.
> > The last one saves some power when the card is not used.
>
> I have applied your 4 patches on both a vanilla 2.6.21.5 kernel and
> a 2.6.22-rc6 wireless-dev one, and i found that only patches 1/4 and 2/4
> are good.
> the 3/4 one gives this bogus output:
>
> airo(): Probing for PCI adapters
> PCI: Setting latency timer of device 0000:05:08.0 to 64
> airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
> airo(eth%d): Doing fast bap_reads
> airo(): WPA is supported.
> airo(eth0): MAC enabled 0:d:29:4f:c:be
> airo(): Finished probing for PCI adapters
>
> and the 4/4 definitely broke all:
>
> airo(): Probing for PCI adapters
> PCI: Setting latency timer of device 0000:05:08.0 to 64
> airo(eth%d): cmd:111 status:7f11 rsp0:2 rsp1:0 rsp2:0
> airo(eth%d): Doing fast bap_reads
> airo(eth%d): Bad MAC enable reason = 2, rid = 0, offset = 0
> airo(eth%d): MAC could not be enabled
> airo(): Finished probing for PCI adapters

I investigated further:
1/4 and 3/4 are OK
2/4 makes the ugly output
4/4 broke everything

also one of the 4 patches causes a random kernel panic on vanilla 2.6.21.5 but
I don't care since they are meant to be applied on another kernel tree

Bye
Matteo Croce