2010-01-21 02:55:47

by Luis Chamberlain

[permalink] [raw]
Subject: RFC/RFT: compat-wireless for 2.6.32.4 - MQ backport support

The 2.6.32.4 kernel got added MQ support back. I've tried to backport
this and this is what I have so far:

http://bombadil.infradead.org/~mcgrof/tmp/compat-wireless-2.6.32.4-rc1.tar.bz2

Not sure about the select queue stuff yet though. Reports of uses of
ath9k (./scripts/driver-select ath9k) on 2.6.23..2.6.26 would be
appreciated.

Luis

--- 18-multiqueue.patch ---

The 2.6.23 kernel added multiqueue support. That release
relied on the on the notion of struct net_device_subqueue
attached to the netdevice struct as an array. The 2.6.27
renamed these to struct netdev_queue, amogst other changes.

For kernels 2.6.23..2.6.26 then we backport MQ support by
using the equivalent calls on the struct netdev_queue to
the struct net_device_subqueue.

For older kernels than 2.6.23 we can only stop all the
queues then.

--- a/net/mac80211/util.c 2010-01-20 18:35:04.000000000 -0800
+++ a/net/mac80211/util.c 2010-01-20 18:34:11.000000000 -0800
@@ -287,7 +287,13 @@

rcu_read_lock();
list_for_each_entry_rcu(sdata, &local->interfaces, list)
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
netif_tx_wake_queue(netdev_get_tx_queue(sdata->dev, queue));
+#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
+ netif_start_subqueue(sdata->dev, queue);
+#else
+ netif_wake_queue(sdata->dev);
+#endif
rcu_read_unlock();
}

@@ -322,7 +328,13 @@

rcu_read_lock();
list_for_each_entry_rcu(sdata, &local->interfaces, list)
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
netif_tx_stop_queue(netdev_get_tx_queue(sdata->dev, queue));
+#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
+ netif_stop_subqueue(sdata->dev, queue);
+#else
+ netif_stop_queue(sdata->dev);
+#endif
rcu_read_unlock();
}



2010-01-21 18:56:37

by Luis R. Rodriguez

[permalink] [raw]
Subject: Re: RFC/RFT: compat-wireless for 2.6.32.4 - MQ backport support

On Thu, Jan 21, 2010 at 10:38 AM, Johannes Berg
<[email protected]> wrote:
> On Thu, 2010-01-21 at 10:16 -0800, Luis R. Rodriguez wrote:
>> On Thu, Jan 21, 2010 at 1:44 AM, Johannes Berg
>> <[email protected]> wrote:
>> > On Wed, 2010-01-20 at 21:55 -0500, Luis R. Rodriguez wrote:
>> >
>> >> @@ -287,7 +287,13 @@
>> >>
>> >>       rcu_read_lock();
>> >>       list_for_each_entry_rcu(sdata, &local->interfaces, list)
>> >> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
>> >>               netif_tx_wake_queue(netdev_get_tx_queue(sdata->dev, queue));
>> >> +#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
>> >> +             netif_start_subqueue(sdata->dev, queue);
>> >> +#else
>> >> +             netif_wake_queue(sdata->dev);
>> >> +#endif
>> >
>> > That's incorrect, you need to check for all hardware queues being awake.
>>
>> Can you elaborate? I'm not following.
>
> Well something like
>
> stopped = 0;
> for (i = 0; i < hw_queues; i++)
>        if (queue_stop_reason[i])
>                stopped++;
>
> if (stopped == 0)
>        netif_wake_queue()

Got it, thanks.

Luis

2010-01-21 09:44:33

by Johannes Berg

[permalink] [raw]
Subject: Re: RFC/RFT: compat-wireless for 2.6.32.4 - MQ backport support

On Wed, 2010-01-20 at 21:55 -0500, Luis R. Rodriguez wrote:

> @@ -287,7 +287,13 @@
>
> rcu_read_lock();
> list_for_each_entry_rcu(sdata, &local->interfaces, list)
> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
> netif_tx_wake_queue(netdev_get_tx_queue(sdata->dev, queue));
> +#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
> + netif_start_subqueue(sdata->dev, queue);
> +#else
> + netif_wake_queue(sdata->dev);
> +#endif

That's incorrect, you need to check for all hardware queues being awake.

johannes


Attachments:
signature.asc (801.00 B)
This is a digitally signed message part

2010-01-21 18:16:56

by Luis R. Rodriguez

[permalink] [raw]
Subject: Re: RFC/RFT: compat-wireless for 2.6.32.4 - MQ backport support

On Thu, Jan 21, 2010 at 1:44 AM, Johannes Berg
<[email protected]> wrote:
> On Wed, 2010-01-20 at 21:55 -0500, Luis R. Rodriguez wrote:
>
>> @@ -287,7 +287,13 @@
>>
>>       rcu_read_lock();
>>       list_for_each_entry_rcu(sdata, &local->interfaces, list)
>> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
>>               netif_tx_wake_queue(netdev_get_tx_queue(sdata->dev, queue));
>> +#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
>> +             netif_start_subqueue(sdata->dev, queue);
>> +#else
>> +             netif_wake_queue(sdata->dev);
>> +#endif
>
> That's incorrect, you need to check for all hardware queues being awake.

Can you elaborate? I'm not following.

Luis

2010-01-21 18:38:17

by Johannes Berg

[permalink] [raw]
Subject: Re: RFC/RFT: compat-wireless for 2.6.32.4 - MQ backport support

On Thu, 2010-01-21 at 10:16 -0800, Luis R. Rodriguez wrote:
> On Thu, Jan 21, 2010 at 1:44 AM, Johannes Berg
> <[email protected]> wrote:
> > On Wed, 2010-01-20 at 21:55 -0500, Luis R. Rodriguez wrote:
> >
> >> @@ -287,7 +287,13 @@
> >>
> >> rcu_read_lock();
> >> list_for_each_entry_rcu(sdata, &local->interfaces, list)
> >> +#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,27))
> >> netif_tx_wake_queue(netdev_get_tx_queue(sdata->dev, queue));
> >> +#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23))
> >> + netif_start_subqueue(sdata->dev, queue);
> >> +#else
> >> + netif_wake_queue(sdata->dev);
> >> +#endif
> >
> > That's incorrect, you need to check for all hardware queues being awake.
>
> Can you elaborate? I'm not following.

Well something like

stopped = 0;
for (i = 0; i < hw_queues; i++)
if (queue_stop_reason[i])
stopped++;

if (stopped == 0)
netif_wake_queue()

johannes


Attachments:
signature.asc (801.00 B)
This is a digitally signed message part