Received-SPF: pass (google.com: domain of linux-kernel+bounces-52208-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Message-ID: <c4a6f0aba84c328dc31d7ec40fe3d4c2b35c6c14.camel@gmail.com>
Subject: Re: [PATCH 1/2] driver: core: add dedicated workqueue for devlink
 removal
From: Nuno =?ISO-8859-1?Q?S=E1?= <noname.nuno@gmail.com>
To: "Rafael J. Wysocki" <rafael@kernel.org>, nuno.sa@analog.com
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>, Frank Rowand
 <frowand.list@gmail.com>, Rob Herring <robh+dt@kernel.org>, Andy Shevchenko
 <andriy.shevchenko@linux.intel.com>, Daniel Scally <djrscally@gmail.com>, 
 Heikki Krogerus <heikki.krogerus@linux.intel.com>, Sakari Ailus
 <sakari.ailus@linux.intel.com>, Len Brown <lenb@kernel.org>,
 linux-acpi@vger.kernel.org, devicetree@vger.kernel.org, 
 linux-kernel@vger.kernel.org
Date: Mon, 05 Feb 2024 09:29:44 +0100
In-Reply-To: <CAJZ5v0g5JbstLhCaXcY1kawP8etB5Z4TBuGXHz8_wsrXm3CaQA@mail.gmail.com>
References: 
	<20240202-fix-device-links-overlays-v1-0-f9fd1404c8e2@analog.com>
	 <20240202-fix-device-links-overlays-v1-1-f9fd1404c8e2@analog.com>
	 <CAJZ5v0g5JbstLhCaXcY1kawP8etB5Z4TBuGXHz8_wsrXm3CaQA@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.50.3 (3.50.3-1.fc39) 
Precedence: bulk
MIME-Version: 1.0

On Fri, 2024-02-02 at 16:59 +0100, Rafael J. Wysocki wrote:
> On Fri, Feb 2, 2024 at 1:18=E2=80=AFPM Nuno Sa via B4 Relay
> <devnull+nuno.sa.analog.com@kernel.org> wrote:
> >=20
> > From: Nuno Sa <nuno.sa@analog.com>
> >=20
> > Let's use a dedicated queue for devlinks since releasing a link happens
> > asynchronously but some code paths, like DT overlays, have some
> > expectations regarding the of_node when being removed (the refcount mus=
t
> > be 1). Given how devlinks are released that cannot be assured. Hence, a=
dd a
> > dedicated queue so that it's easy to sync against devlinks removal.
>=20
> Thanks for following my suggestion!
>=20
> > While at it, make sure to explicitly include <linux/workqueue.h>.
> >=20
> > Signed-off-by: Nuno Sa <nuno.sa@analog.com>
> > ---
> > =C2=A0drivers/base/core.c=C2=A0=C2=A0=C2=A0 | 33 ++++++++++++++++++++++=
+++++++----
> > =C2=A0include/linux/fwnode.h |=C2=A0 1 +
> > =C2=A02 files changed, 30 insertions(+), 4 deletions(-)
> >=20
> > diff --git a/drivers/base/core.c b/drivers/base/core.c
> > index 14d46af40f9a..06e7766b5227 100644
> > --- a/drivers/base/core.c
> > +++ b/drivers/base/core.c
> > @@ -31,6 +31,7 @@
> > =C2=A0#include <linux/swiotlb.h>
> > =C2=A0#include <linux/sysfs.h>
> > =C2=A0#include <linux/dma-map-ops.h> /* for dma_default_coherent */
> > +#include <linux/workqueue.h>
> >=20
> > =C2=A0#include "base.h"
> > =C2=A0#include "physical_location.h"
> > @@ -44,6 +45,7 @@ static bool fw_devlink_is_permissive(void);
> > =C2=A0static void __fw_devlink_link_to_consumers(struct device *dev);
> > =C2=A0static bool fw_devlink_drv_reg_done;
> > =C2=A0static bool fw_devlink_best_effort;
> > +static struct workqueue_struct *devlink_release_queue __ro_after_init;
> >=20
> > =C2=A0/**
> > =C2=A0 * __fwnode_link_add - Create a link between two fwnode_handles.
> > @@ -235,6 +237,11 @@ static void __fw_devlink_pickup_dangling_consumers=
(struct
> > fwnode_handle *fwnode,
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0 __fw_devlink_pickup_dangling_consumers(child, new_sup=
);
> > =C2=A0}
> >=20
> > +void fwnode_links_flush_queue(void)
> > +{
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 flush_workqueue(devlink_release_q=
ueue);
> > +}
> > +
> > =C2=A0static DEFINE_MUTEX(device_links_lock);
> > =C2=A0DEFINE_STATIC_SRCU(device_links_srcu);
> >=20
> > @@ -531,9 +538,10 @@ static void devlink_dev_release(struct device *dev=
)
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * It may take a while =
to complete this work because of the SRCU
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * synchronization in d=
evice_link_release_fn() and if the consumer or
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * supplier devices get=
 deleted when it runs, so put it into the "long"
> > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * workqueue.
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * devlink workqueue.
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 *
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 */
> > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 queue_work(system_long_wq, &link-=
>rm_work);
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 queue_work(devlink_release_queue,=
 &link->rm_work);
> > =C2=A0}
> >=20
> > =C2=A0static struct class devlink_class =3D {
> > @@ -636,10 +644,27 @@ static int __init devlink_class_init(void)
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0 return ret;
> >=20
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 ret =3D class_interface_regi=
ster(&devlink_class_intf);
> > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (ret)
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (ret) {
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0 class_unregister(&devlink_class);
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0 return ret;
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 }
> > +
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 /*
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * Using a dedicated queue f=
or devlinks since releasing a link happens
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * asynchronously but some c=
ode paths, like DT overlays, have some
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * expectations regarding th=
e of_node when being removed (the refcount
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * must be 1). Given how dev=
links are released that cannot be assured.
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * Hence, add a dedicated qu=
eue so that it's easy to sync against
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 * devlinks removal.
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 */
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 devlink_release_queue =3D alloc_w=
orkqueue("devlink_release", 0, 0);
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (!devlink_release_queue) {
> > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0 class_interface_unregister(&devlink_class_intf);
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0 class_unregister(&devlink_class);
>=20
> This is a bit drastic.
>=20
> I think that device links can still work if devlink_release_queue is
> NULL, just devlink_dev_release() needs to check it and release
> synchronously if it is NULL.
>=20

Agreed, I'll do that way. It will always synchronously remove the links (wh=
ich is
different than before) but I guess that failing in allocating the queue is =
rather
unlikely.

- Nuno S=C3=A1