media: dvbdev.h: do some kernel-doc cleanups

Some kernel-doc warnings in <media/dvbdev.h> were introduced. A fixup
patch addressed them was already merged, but Randy's approach from:
https://lore.kernel.org/linux-media/20221203060931.19953-1-rdunlap@infradead.org

Had some advantages, as it moves the @dvbdev to the right place inside
dvb_remove_device() documentation and it makes clearer about what
refcounter struct dvb_device refers to.

So, apply the changes suggested by Randy.

Suggested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

media: dvb-core: Fix use-after-free due to race at dvb_register_device()

dvb_register_device() dynamically allocates fops with kmemdup()
to set the fops->owner.
And these fops are registered in 'file->f_ops' using replace_fops()
in the dvb_device_open() process, and kfree()d in dvb_free_device().

However, it is not common to use dynamically allocated fops instead
of 'static const' fops as an argument of replace_fops(),
and UAF may occur.
These UAFs can occur on any dvb type using dvb_register_device(),
such as dvb_dvr, dvb_demux, dvb_frontend, dvb_net, etc.

So, instead of kfree() the fops dynamically allocated in
dvb_register_device() in dvb_free_device() called during the
.disconnect() process, kfree() it collectively in exit_dvbdev()
called when the dvbdev.c module is removed.

Link: https://lore.kernel.org/linux-media/20221117045925.14297-4-imv4bel@gmail.com
Signed-off-by: Hyunwoo Kim <imv4bel@gmail.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

media: dvbdev: fix build warning due to comments

Previous commit that introduces reference counter does not add proper
comments, which will lead to warning when building htmldocs. Fix them.

Reported-by: "Stephen Rothwell" <sfr@canb.auug.org.au>
Fixes: 0fc044b2b5e2 ("media: dvbdev: adopts refcnt to avoid UAF")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

media: dvbdev: adopts refcnt to avoid UAF

dvb_unregister_device() is known that prone to use-after-free.
That is, the cleanup from dvb_unregister_device() releases the dvb_device
even if there are pointers stored in file->private_data still refer to it.

This patch adds a reference counter into struct dvb_device and delays its
deallocation until no pointer refers to the object.

Link: https://lore.kernel.org/linux-media/20220807145952.10368-1-linma@zju.edu.cn
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

media: dvb-core: Enhance shared multi-frontend support

Drivers for devices with multiple frontends which cannot be used
concurrently due to hardware limitations which enforce that restriction
by setting the mfe_shared field to 1 exhibit rather unfriendly behavior
towards applications: The unavailable frontend devices cannot be opened
at all, not even for read-only access to query information. Even worse,
any open call is blocked for 5 seconds by default.

Allow drivers for such devices to behave like regular busy frontend
devices instead, i.e. still allowing concurrent read access to the
unavailable frontend and denying concurrent write access with -EBUSY
without delay.

This patch does not alter the behavior of any existing driver to avoid
regressions. Driver developers who wish to take advantage of this must
ensure their driver can handle all read-only accesses to the unavailable
frontend, and indicate the capability by setting the mfe_shared field to
2 instead of 1.

Add a check to dvb-usb-init.c when automatically setting the mfe_shared
field that when a driver has already set the field to 2, it is not
overwritten.

Document the additional capability in the code comment about mfe_shared.

Link: https://lore.kernel.org/linux-media/trinity-22c77578-26b0-4867-9ff7-2668e5d22c64-1642799929896@3c-app-gmx-bap04
Signed-off-by: Robert Schlabbach <robert_s@gmx.net>
Tested-by: Robert Schlabbach <robert_s@gmx.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

media: dvbdev: Switch to new kerneldoc syntax for named variable macro argument

The syntax without dots is available since commit 43756e347f21
("scripts/kernel-doc: Add support for named variable macro arguments").

The same HTML output is produced with and without this patch.

Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: fix kernel-doc markups

Some identifiers have different names between their prototypes
and the kernel-doc markup. Seome seems to be due to cut-and-paste
related issues.

Others need to be fixed, as kernel-doc markups should use this format:
identifier - description

Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com> # IPU3 and V4L2
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dvbdev.h: keep * together with the type

Having the '*' in the next line separated from the type makes it
hard to see that these functions return a pointer to that type.

Instead, keep it next to the type name so it is clear that it is
a pointer to that type.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: dvbdev.h: fix duplicated word in comment

Change the doubled word "the" in a comment to "to the".

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

media: docs: fix minor typos

Fix minor typos in the DVB demux documentation.

Signed-off-by: Marc Gonzalez <marc.w.gonzalez@free.fr>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: dvbdev: add a mutex protecting the "mdev" pointer

During destruction, a race condition in
dvb_media_controller_disable_source() can cause a kernel crash,
because the "mdev" pointer has been read successfully while another
task executes dvb_usb_media_device_unregister(), which destroys the
object. Example for such a crash:

general protection fault: 0000 [#1] SMP
CPU: 1 PID: 301 Comm: vdr Not tainted 4.8.1-nuc+ #102
[142B blob data]
task: ffff8802301f2040 task.stack: ffff880233728000
RIP: 0010:[<ffffffff816c296b>] [<ffffffff816c296b>] dvb_frontend_release+0xcb/0x120
RSP: 0018:ffff88023372bdd8 EFLAGS: 00010202
RAX: 001fd55c000000da RBX: ffff880236bad810 RCX: 0000000000000000
RDX: ffff880235bd81f0 RSI: 0000000000000246 RDI: ffff880235bd81e8
RBP: ffff88023372be00 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: ffff88022f009910 R12: 0000000000000000
R13: ffff880235a21a80 R14: ffff880235bd8000 R15: ffff880235bb8a78
FS: 0000000000000000(0000) GS:ffff88023fd00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f96edd69818 CR3: 0000000002406000 CR4: 00000000001006e0
Stack:
ffff88022f009900 0000000000000008 ffff880235bb8a78 ffff8802344fbb20
ffff880236437b40 ffff88023372be48 ffffffff8117a81e ffff880235bb8a78
ffff88022f009910 ffff8802335a7400 ffff8802301f2040 ffff88022f009900
Call Trace:
[<ffffffff8117a81e>] __fput+0xde/0x1d0
[<ffffffff8117a949>] ____fput+0x9/0x10
[<ffffffff810a9fce>] task_work_run+0x7e/0xa0
[<ffffffff81094bab>] do_exit+0x27b/0xa50
[<ffffffff810407e3>] ? __do_page_fault+0x1c3/0x430
[<ffffffff81095402>] do_group_exit+0x42/0xb0
[<ffffffff8109547f>] SyS_exit_group+0xf/0x10
[<ffffffff8108bedb>] entry_SYSCALL_64_fastpath+0x13/0x8f
Code: 31 c9 49 8d be e8 01 00 00 ba 01 00 00 00 be 03 00 00 00 e8 68 2d a0 ff 48 8b 83 10 03 00 00 48 8b 80 88 00 00 00 48 85 c0 74 12 <48> 8b 80 88 02 00 00 48 85 c0 74 06 49 8b 7d
RIP [<ffffffff816c296b>] dvb_frontend_release+0xcb/0x120

[mchehab+samsung@kernel.org: fix a Coding Style issue]
Signed-off-by: Max Kellermann <max.kellermann@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>

media: dvbdev: fix building on ia64

Not sure why, but, on ia64, with Linaro's gcc 7.3 compiler,
using #ifdef (CONFIG_I2C) is not OK.

So, replace it by IS_ENABLED(CONFIG_I2C), in order to fix the
builds there.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>

media: dvb-core: add helper functions for I2C binding

The dvb_attach()/dvb_detach() methods are ugly hacks designed
to keep using the I2C low-level API. The proper way is to
do I2C bus bindings instead.

Several modules were already converted to use it. Yet,
it is painful to use it, as lots of code need to be
duplicated.

Make it easier by providing two new helper functions:
- dvb_module_probe()
- dvb_module_release()

Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>

media: fix usage of whitespaces and on indentation

On several places, whitespaces are being used for indentation,
or even at the end of the line.

Fix them.

Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>

media: move dvb kAPI headers to include/media

Except for DVB, all media kAPI headers are at include/media.

Move the headers to it.

Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>