2019-10-01 11:22:01

by Johannes Berg

Subject: [PATCH v2] mac80211: keep BHs disabled while calling drv_tx_wake_queue()

From: Johannes Berg <[email protected]>

Drivers typically expect this, as it's the case for almost all cases
where this is called (i.e. from the TX path). Also, the code in mac80211
itself (if the driver calls ieee80211_tx_dequeue()) expects this as it
uses this_cpu_ptr() without additional protection.

This should fix various reports of the problem:
https://bugzilla.kernel.org/show_bug.cgi?id=204127
https://lore.kernel.org/linux-wireless/CAN5HydrWb3o_FE6A1XDnP1E+xS66d5kiEuhHfiGKkLNQokx13Q@mail.gmail.com/
https://lore.kernel.org/lkml/[email protected]/

Reported-by: Jiri Kosina <[email protected]>
Reported-by: Aaron Hill <[email protected]>
Reported-by: Lukas Redlinger <[email protected]>
Reported-by: Oleksii Shevchuk <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
---
v2:
* use local_bh_enable/disable to capture the last occurrence
* split spin_lock_bh() into local_bh_disable()/spin_lock() to
make it clearer while we unlock them separately
---
net/mac80211/util.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/net/mac80211/util.c b/net/mac80211/util.c
index 051a02ddcb85..32a7a53833c0 100644
--- a/net/mac80211/util.c
+++ b/net/mac80211/util.c
@@ -247,7 +247,8 @@ static void __ieee80211_wake_txqs(struct ieee80211_sub_if_data *sdata, int ac)
struct sta_info *sta;
int i;

- spin_lock_bh(&fq->lock);
+ local_bh_disable();
+ spin_lock(&fq->lock);

if (sdata->vif.type == NL80211_IFTYPE_AP)
ps = &sdata->bss->ps;
@@ -273,9 +274,9 @@ static void __ieee80211_wake_txqs(struct ieee80211_sub_if_data *sdata, int ac)
&txqi->flags))
continue;

- spin_unlock_bh(&fq->lock);
+ spin_unlock(&fq->lock);
drv_wake_tx_queue(local, txqi);
- spin_lock_bh(&fq->lock);
+ spin_lock(&fq->lock);
}
}

@@ -288,12 +289,14 @@ static void __ieee80211_wake_txqs(struct ieee80211_sub_if_data *sdata, int ac)
(ps && atomic_read(&ps->num_sta_ps)) || ac != vif->txq->ac)
goto out;

- spin_unlock_bh(&fq->lock);
+ spin_unlock(&fq->lock);

drv_wake_tx_queue(local, txqi);
+ local_bh_enable();
return;
out:
- spin_unlock_bh(&fq->lock);
+ spin_unlock(&fq->lock);
+ local_bh_enable();
}

static void
--
2.20.1

2019-10-01 11:48:32

by Toke Høiland-Jørgensen

Subject: Re: [PATCH v2] mac80211: keep BHs disabled while calling drv_tx_wake_queue()

Johannes Berg <[email protected]> writes:

> From: Johannes Berg <[email protected]>
>
> Drivers typically expect this, as it's the case for almost all cases
> where this is called (i.e. from the TX path). Also, the code in mac80211
> itself (if the driver calls ieee80211_tx_dequeue()) expects this as it
> uses this_cpu_ptr() without additional protection.
>
> This should fix various reports of the problem:
> https://bugzilla.kernel.org/show_bug.cgi?id=204127
> https://lore.kernel.org/linux-wireless/CAN5HydrWb3o_FE6A1XDnP1E+xS66d5kiEuhHfiGKkLNQokx13Q@mail.gmail.com/
> https://lore.kernel.org/lkml/[email protected]/
>
> Reported-by: Jiri Kosina <[email protected]>
> Reported-by: Aaron Hill <[email protected]>
> Reported-by: Lukas Redlinger <[email protected]>
> Reported-by: Oleksii Shevchuk <[email protected]>
> Signed-off-by: Johannes Berg <[email protected]>

Reviewed-by: Toke Høiland-Jørgensen <[email protected]>

2019-10-01 12:01:05

by Jiri Kosina

Subject: Re: [PATCH v2] mac80211: keep BHs disabled while calling drv_tx_wake_queue()

On Tue, 1 Oct 2019, Toke Høiland-Jørgensen wrote:

> Johannes Berg <[email protected]> writes:
>
> > From: Johannes Berg <[email protected]>
> >
> > Drivers typically expect this, as it's the case for almost all cases
> > where this is called (i.e. from the TX path). Also, the code in mac80211
> > itself (if the driver calls ieee80211_tx_dequeue()) expects this as it
> > uses this_cpu_ptr() without additional protection.
> >
> > This should fix various reports of the problem:
> > https://bugzilla.kernel.org/show_bug.cgi?id=204127
> > https://lore.kernel.org/linux-wireless/CAN5HydrWb3o_FE6A1XDnP1E+xS66d5kiEuhHfiGKkLNQokx13Q@mail.gmail.com/
> > https://lore.kernel.org/lkml/[email protected]/
> >
> > Reported-by: Jiri Kosina <[email protected]>

Reported-and-tested-by: Jiri Kosina <[email protected]>

> > Reported-by: Aaron Hill <[email protected]>
> > Reported-by: Lukas Redlinger <[email protected]>
> > Reported-by: Oleksii Shevchuk <[email protected]>
> > Signed-off-by: Johannes Berg <[email protected]>
>
> Reviewed-by: Toke Høiland-Jørgensen <[email protected]>

Thanks,

--
Jiri Kosina
SUSE Labs