[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240820074102.52c7c43a@kernel.org>
Date: Tue, 20 Aug 2024 07:41:02 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Larry Chiu <larry.chiu@...ltek.com>
Cc: Justin Lai <justinlai0215@...ltek.com>, "davem@...emloft.net"
<davem@...emloft.net>, "edumazet@...gle.com" <edumazet@...gle.com>,
"pabeni@...hat.com" <pabeni@...hat.com>, "linux-kernel@...r.kernel.org"
<linux-kernel@...r.kernel.org>, "netdev@...r.kernel.org"
<netdev@...r.kernel.org>, "andrew@...n.ch" <andrew@...n.ch>,
"jiri@...nulli.us" <jiri@...nulli.us>, "horms@...nel.org"
<horms@...nel.org>, "rkannoth@...vell.com" <rkannoth@...vell.com>,
"jdamato@...tly.com" <jdamato@...tly.com>, Ping-Ke Shih
<pkshih@...ltek.com>
Subject: Re: [PATCH net-next v27 07/13] rtase: Implement a function to
receive packets
On Tue, 20 Aug 2024 05:13:32 +0000 Larry Chiu wrote:
> > Memory allocation failures happen, we shouldn't risk spamming the logs.
> > I mean these two messages and the one in rtase_alloc_rx_data_buf(),
> > the should be removed.
> >
> > There is a alloc_fail statistic defined in include/net/netdev_queues.h
> > that's the correct way to report buffer allocation failures.
>
> Hi, Jakub,
> Can we just count the rx_alloc_fail here?
> If we implement the "netdev_stat_ops", we can report this counter.
I think so.
> > And you should have a periodic service task / work which checks for
> > buffers being exhausted, and if they are schedule NAPI so that it tries
> > to allocate.
>
> We will redefine the rtase_rx_ring_fill() to check the buffers and
> try to get page from the pool.
> Should we return the budget to schedule this NAPI if there are some
> empty buffers?
I wouldn't recommend that. If system is under memory stress
we shouldn't be adding extra load by rescheduling NAPI.
Powered by blists - more mailing lists