Received-SPF: pass (google.com: domain of linux-kernel+bounces-63626-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
References: <20240202165315.2506384-1-leitao@debian.org>
In-Reply-To: <20240202165315.2506384-1-leitao@debian.org>
From: Eric Dumazet <edumazet@google.com>
Date: Tue, 13 Feb 2024 14:57:49 +0100
Message-ID: <CANn89iLWWDjp71R7zttfTcEvZEdmcA1qo47oXkAX5DuciYvOtQ@mail.gmail.com>
Subject: Re: [PATCH net-next v3] net: dqs: add NIC stall detector based on BQL
To: Breno Leitao <leitao@debian.org>
Cc: kuba@kernel.org, davem@davemloft.net, pabeni@redhat.com, 
	Steven Rostedt <rostedt@goodmis.org>, Masami Hiramatsu <mhiramat@kernel.org>, 
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>, Andrew Morton <akpm@linux-foundation.org>, 
	weiwan@google.com, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, 
	horms@kernel.org, Jonathan Corbet <corbet@lwn.net>, Randy Dunlap <rdunlap@infradead.org>, 
	Bjorn Helgaas <bhelgaas@google.com>, Johannes Berg <johannes.berg@intel.com>, 
	=?UTF-8?Q?Thomas_Wei=C3=9Fschuh?= <linux@weissschuh.net>, 
	"open list:TRACING" <linux-trace-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Feb 2, 2024 at 5:55=E2=80=AFPM Breno Leitao <leitao@debian.org> wro=
te:
>
> From: Jakub Kicinski <kuba@kernel.org>
>
> softnet_data->time_squeeze is sometimes used as a proxy for
> host overload or indication of scheduling problems. In practice
> this statistic is very noisy and has hard to grasp units -
> e.g. is 10 squeezes a second to be expected, or high?
>
> Delaying network (NAPI) processing leads to drops on NIC queues
> but also RTT bloat, impacting pacing and CA decisions.
> Stalls are a little hard to detect on the Rx side, because
> there may simply have not been any packets received in given
> period of time. Packet timestamps help a little bit, but
> again we don't know if packets are stale because we're
> not keeping up or because someone (*cough* cgroups)
> disabled IRQs for a long time.

Please note that adding other sysfs entries is expensive for workloads
creating/deleting netdev and netns often.

I _think_ we should find a way for not creating
/sys/class/net/<interface>/queues/tx-{Q}/byte_queue_limits  directory
and files
for non BQL enabled devices (like loopback !)