Received-SPF: pass (google.com: domain of linux-kernel+bounces-86293-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <2057fd477ad404c9adf603eac1ad933c04ecf293.camel@redhat.com>
Subject: Re: [PATCH v3 00/10] Make PCI's devres API more consistent
From: Philipp Stanner <pstanner@redhat.com>
To: Hans de Goede <hdegoede@redhat.com>, Maarten Lankhorst
 <maarten.lankhorst@linux.intel.com>, Maxime Ripard <mripard@kernel.org>, 
 Thomas Zimmermann <tzimmermann@suse.de>, David Airlie <airlied@gmail.com>,
 Daniel Vetter <daniel@ffwll.ch>, Bjorn Helgaas <bhelgaas@google.com>, Sam
 Ravnborg <sam@ravnborg.org>, dakr@redhat.com
Cc: dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org, 
	linux-pci@vger.kernel.org
Date: Thu, 29 Feb 2024 09:31:20 +0100
In-Reply-To: <20240206134000.23561-2-pstanner@redhat.com>
References: <20240206134000.23561-2-pstanner@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.48.4 (3.48.4-1.fc38) 
Precedence: bulk
MIME-Version: 1.0

@Bjorn:
Hey Bjorn, are we good with this series? Any more wishes or
suggestions?

P.


On Tue, 2024-02-06 at 14:39 +0100, Philipp Stanner wrote:
> Changes in v3:
> =C2=A0 - Use the term "PCI devres API" in some forgotten places.
> =C2=A0 - Fix more grammar errors in patch #3.
> =C2=A0 - Remove the comment advising to call (the outdated) pcim_intx() i=
n
> pci.c
> =C2=A0 - Rename __pcim_request_region_range() flags-field "exclusive" to
> =C2=A0=C2=A0=C2=A0 "req_flags", since this is what the int actually repre=
sents.
> =C2=A0 - Remove the call to pcim_region_request() from patch #10. (Hans)
>=20
> Changes in v2:
> =C2=A0 - Make commit head lines congruent with PCI's style (Bjorn)
> =C2=A0 - Add missing error checks for devm_add_action(). (Andy)
> =C2=A0 - Repair the "Returns: " marks for docu generation (Andy)
> =C2=A0 - Initialize the addr_devres struct with memset(). (Andy)
> =C2=A0 - Make pcim_intx() a PCI-internal function so that new drivers
> won't
> =C2=A0=C2=A0=C2=A0 be encouraged to use the outdated pci_intx() mechanism=
.
> =C2=A0=C2=A0=C2=A0 (Andy / Philipp)
> =C2=A0 - Fix grammar and spelling (Bjorn)
> =C2=A0 - Be more precise on why pcim_iomap_table() is problematic (Bjorn)
> =C2=A0 - Provide the actual structs' and functions' names in the commit
> =C2=A0=C2=A0=C2=A0 messages (Bjorn)
> =C2=A0 - Remove redundant variable initializers (Andy)
> =C2=A0 - Regroup PM bitfield members in struct pci_dev (Andy)
> =C2=A0 - Make pcim_intx() visible only for the PCI subsystem so that
> new=C2=A0=C2=A0=C2=A0=20
> =C2=A0=C2=A0=C2=A0 drivers won't use this outdated API (Andy, Myself)
> =C2=A0 - Add a NOTE to pcim_iomap() to warn about this function being
> the=C2=A0=C2=A0=C2=A0 onee
> =C2=A0=C2=A0=C2=A0 xception that does just return NULL.
> =C2=A0 - Consistently use the term "PCI devres API"; also in Patch #10
> (Bjorn)
>=20
>=20
> =C2=A1Hola!
>=20
> PCI's devres API suffers several weaknesses:
>=20
> 1. There are functions prefixed with pcim_. Those are always managed
> =C2=A0=C2=A0 counterparts to never-managed functions prefixed with pci_ =
=E2=80=93 or so
> one
> =C2=A0=C2=A0 would like to think. There are some apparently unmanaged fun=
ctions
> =C2=A0=C2=A0 (all region-request / release functions, and pci_intx()) whi=
ch
> =C2=A0=C2=A0 suddenly become managed once the user has initialized the de=
vice
> with
> =C2=A0=C2=A0 pcim_enable_device() instead of pci_enable_device(). This
> "sometimes
> =C2=A0=C2=A0 yes, sometimes no" nature of those functions is confusing an=
d
> =C2=A0=C2=A0 therefore bug-provoking. In fact, it has already caused a bu=
g in
> DRM.
> =C2=A0=C2=A0 The last patch in this series fixes that bug.
> 2. iomappings: Instead of giving each mapping its own callback, the
> =C2=A0=C2=A0 existing API uses a statically allocated struct tracking one
> mapping
> =C2=A0=C2=A0 per bar. This is not extensible. Especially, you can't creat=
e
> =C2=A0=C2=A0 _ranged_ managed mappings that way, which many drivers want.
> 3. Managed request functions only exist as "plural versions" with a
> =C2=A0=C2=A0 bit-mask as a parameter. That's quite over-engineered consid=
ering
> =C2=A0=C2=A0 that each user only ever mapps one, maybe two bars.
>=20
> This series:
> - add a set of new "singular" devres functions that use devres the
> way
> =C2=A0 its intended, with one callback per resource.
> - deprecates the existing iomap-table mechanism.
> - deprecates the hybrid nature of pci_ functions.
> - preserves backwards compatibility so that drivers using the
> existing
> =C2=A0 API won't notice any changes.
> - adds documentation, especially some warning users about the
> =C2=A0 complicated nature of PCI's devres.
>=20
>=20
> Note that this series is based on my "unify pci_iounmap"-series from
> a
> few weeks ago. [1]
>=20
> I tested this on a x86 VM with a simple pci test-device with two
> regions. Operates and reserves resources as intended on my system.
> Kasan and kmemleak didn't find any problems.
>=20
> I believe this series cleans the API up as much as possible without
> having to port all existing drivers to the new API. Especially, I
> think
> that this implementation is easy to extend if the need for new
> managed
> functions arises :)
>=20
> Greetings,
> P.
>=20
> Philipp Stanner (10):
> =C2=A0 PCI: Add new set of devres functions
> =C2=A0 PCI: Deprecate iomap-table functions
> =C2=A0 PCI: Warn users about complicated devres nature
> =C2=A0 PCI: Make devres region requests consistent
> =C2=A0 PCI: Move dev-enabled status bit to struct pci_dev
> =C2=A0 PCI: Move pinned status bit to struct pci_dev
> =C2=A0 PCI: Give pcim_set_mwi() its own devres callback
> =C2=A0 PCI: Give pci(m)_intx its own devres callback
> =C2=A0 PCI: Remove legacy pcim_release()
> =C2=A0 drm/vboxvideo: fix mapping leaks
>=20
> =C2=A0drivers/gpu/drm/vboxvideo/vbox_main.c |=C2=A0=C2=A0 20 +-
> =C2=A0drivers/pci/devres.c=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 | 1015 ++++++++++=
+++++++++++--
> --
> =C2=A0drivers/pci/iomap.c=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 =
18 +
> =C2=A0drivers/pci/pci.c=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 123 ++-
> =C2=A0drivers/pci/pci.h=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=C2=A0 25 +-
> =C2=A0include/linux/pci.h=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 =
18 +-
> =C2=A06 files changed, 1004 insertions(+), 215 deletions(-)
>=20