RE: [RFC 1/7] mac80211: Add check for napi handle before WARN_ON
From: Rakesh Pillai
Date: Sun Jul 26 2020 - 12:23:32 EST
> -----Original Message-----
> From: Rakesh Pillai <pillair@xxxxxxxxxxxxxx>
> Sent: Friday, July 24, 2020 11:51 AM
> To: 'Johannes Berg' <johannes@xxxxxxxxxxxxxxxx>;
> 'ath10k@xxxxxxxxxxxxxxxxxxx' <ath10k@xxxxxxxxxxxxxxxxxxx>
> Cc: 'linux-wireless@xxxxxxxxxxxxxxx' <linux-wireless@xxxxxxxxxxxxxxx>;
> 'linux-kernel@xxxxxxxxxxxxxxx' <linux-kernel@xxxxxxxxxxxxxxx>;
> 'kvalo@xxxxxxxxxxxxxx' <kvalo@xxxxxxxxxxxxxx>; 'davem@xxxxxxxxxxxxx'
> <davem@xxxxxxxxxxxxx>; 'kuba@xxxxxxxxxx' <kuba@xxxxxxxxxx>;
> 'netdev@xxxxxxxxxxxxxxx' <netdev@xxxxxxxxxxxxxxx>;
> 'dianders@xxxxxxxxxxxx' <dianders@xxxxxxxxxxxx>;
> 'evgreen@xxxxxxxxxxxx' <evgreen@xxxxxxxxxxxx>
> Subject: RE: [RFC 1/7] mac80211: Add check for napi handle before
> > -----Original Message-----
> > From: Johannes Berg <johannes@xxxxxxxxxxxxxxxx>
> > Sent: Friday, July 24, 2020 1:37 AM
> > To: Rakesh Pillai <pillair@xxxxxxxxxxxxxx>; ath10k@xxxxxxxxxxxxxxxxxxx
> > Cc: linux-wireless@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > kvalo@xxxxxxxxxxxxxx; davem@xxxxxxxxxxxxx; kuba@xxxxxxxxxx;
> > netdev@xxxxxxxxxxxxxxx; dianders@xxxxxxxxxxxx;
> > Subject: Re: [RFC 1/7] mac80211: Add check for napi handle before
> > WARN_ON
> > On Thu, 2020-07-23 at 23:56 +0530, Rakesh Pillai wrote:
> > > > > - WARN_ON_ONCE(softirq_count() == 0);
> > > > > + WARN_ON_ONCE(napi && softirq_count() == 0);
> > > >
> > > > FWIW, I'm pretty sure this is incorrect - we make assumptions on
> > > > softirqs being disabled in mac80211 for serialization and in place of
> > > > some locking, I believe.
> > > >
> > >
> > > I checked this, but let me double confirm.
> > > But after this change, no packet is submitted from driver in a softirq
> > context.
> > > So ideally this should take care of serialization.
> > I'd guess that we have some reliance on BHs already being disabled, for
> > things like u64 sync updates, or whatnot. I mean, we did "rx_ni()" for a
> > reason ... Maybe lockdep can help catch some of the issues.
> > But couldn't you be in a thread and have BHs disabled too?
> This would ideally beat the purpose and possibly hurt the other subsystems
> running on the same core.
We do have the usage of napi_gro_receive and netif_receive_skb in mac80211.
/* deliver to local stack */
Also all the rx_handlers are called under the " rx->local->rx_path_lock" lock.
Is the BH disable still required ?
> > johannes