clang-16 points out a control flow integrity (kcfi) issue when event
callbacks get converted to incompatible types:
sound/core/seq/seq_midi.c:135:30: error: cast from 'int (*)(struct snd_rawmidi_substream *, const char *, int)' to 'snd_seq_dump_func_t' (aka 'int (*)(void *, void *, int)') converts to incompatible function type [-Werror,-Wcast-function-type-strict]
135 | snd_seq_dump_var_event(ev, (snd_seq_dump_func_t)dump_midi, substream);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sound/core/seq/seq_virmidi.c:83:31: error: cast from 'int (*)(struct snd_rawmidi_substream *, const unsigned char *, int)' to 'snd_seq_dump_func_t' (aka 'int (*)(void *, void *, int)') converts to incompatible function type [-Werror,-Wcast-function-type-strict]
83 | snd_seq_dump_var_event(ev, (snd_seq_dump_func_t)snd_rawmidi_receive, vmidi->substream);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
For addressing those errors, introduce wrapper functions that are used
for callbacks and bridge to the actual function call with pointer
cast.
The code was originally added with the initial ALSA merge in linux-2.5.4.
[ the patch description shamelessly copied from Arnd's original patch
-- tiwai ]
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reported-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20240213101020.459183-1-arnd@kernel.org
Link: https://lore.kernel.org/r/20240213135343.16411-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
When compiling with gcc version 14.0.0 20231220 (experimental)
and W=1, I've noticed the following warning:
sound/core/seq/seq_memory.c: In function 'snd_seq_pool_init':
sound/core/seq/seq_memory.c:445:41: warning: 'kvmalloc_array' sizes specified with
'sizeof' in the earlier argument and not in the later argument [-Wcalloc-transposed-args]
445 | cellptr = kvmalloc_array(sizeof(struct snd_seq_event_cell), pool->size,
| ^~~~~~
Since 'n' and 'size' arguments of 'kvmalloc_array()' are multiplied
to calculate the final size, their actual order doesn't affect the
result and so this is not a bug. But it's still worth to fix it.
Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
Link: https://lore.kernel.org/r/20231221091605.14660-1-dmantipov@yandex.ru
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The compile warnings with -Wformat-truncation appearing at
snd_seq_midisynth_probe() in seq_midi.c are false-positive; those must
fit within the given string size.
For suppressing the warning, replace snprintf() with scnprintf().
As stated in the above, truncation doesn't matter.
Link: https://lore.kernel.org/r/20230915082802.28684-3-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The filling of a port name string got a warning with W=1 due to the
potentially too long group name. Add the string precision to limit
the size.
Fixes: 81fd444aa3 ("ALSA: seq: Bind UMP device")
Link: https://lore.kernel.org/r/20230915082802.28684-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
ALSA sequencer core still delivers events to the disabled UMP Group,
leaving this handling to the device. But it's rather risky and it's
easy to imagine that such an unexpected event may screw up the device
firmware.
This patch avoids the superfluous event deliveries by setting the
group_filter of the UMP client as default, and evaluate the
group_filter properly at delivery from non-UMP clients.
The grouop_filter is updated upon the dynamic UMP Function Block
updates, so that it follows the change of the disabled UMP Groups,
too.
Fixes: d2b7060777 ("ALSA: seq: Add UMP group filter")
Link: https://lore.kernel.org/r/20230912085144.32534-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Align with the other components, and use snd_device_alloc() for the
new sound device for sequencer, too. No functional changes.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Signed-off-by: Curtis Malainey <cujomalainey@chromium.org>
Tested-by: Curtis Malainey <cujomalainey@chromium.org>
Link: https://lore.kernel.org/r/20230816160252.23396-9-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
UMP v1.1 supports the protocol switch via a UMP Stream message. When
it's received, we need to take care of the midi_version field in the
corresponding sequencer client, too.
This patch introduces a new ops to notify the protocol change to
snd_seq_ump_ops for handling it.
Link: https://lore.kernel.org/r/20230612081054.17200-9-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
For allowing applications to track the FB active changes, this patch
adds the notification from the system port at each time a FB change is
handled and the active flag or re-grouping happens.
Link: https://lore.kernel.org/r/20230612081054.17200-8-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch implements the handling of the dynamic update of FB info.
When the FB info update is received after the initial parsing, it
means the dynamic FB info update. We compare the result, and if the
actual update is detected, it's notified via a new ops,
notify_fb_change, to the sequencer client, and the corresponding
sequencer ports are updated accordingly.
Link: https://lore.kernel.org/r/20230612081054.17200-7-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The UMP Utility and Stream messages are "groupless", i.e. an incoming
groupless packet should be sent only to the UMP EP port, and the event
with the groupless message is sent to UMP EP as is without the group
translation per port.
Also, the former reserved bit 0 for the client group filter is now
used for groupless events. When the bit 0 is set, the groupless
events are filtered out and skipped.
Link: https://lore.kernel.org/r/20230612081054.17200-6-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Currently the read event packet size in snd_seq_read() is defined by
client->midi_version value that is guaranteed to be zero if UMP isn't
enabled. But the static analyzer doesn't know of the fact, and it
still suspects as if it were leading to a potential overflow.
Add the more explicit check of CONFIG_SND_SEQ_UMP to determine the
aligned_size value for avoiding the confusion.
Fixes: 46397622a3 ("ALSA: seq: Add UMP support")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <error27@gmail.com>
Closes: https://lore.kernel.org/r/202305261415.NY0vapZK-lkp@intel.com/
Link: https://lore.kernel.org/r/20230605144758.6677-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add a new filter bitmap for UMP groups for reducing the unnecessary
read/write when the client is connected to UMP EP seq port.
The new group_filter field contains the bitmap for the groups, i.e.
when the bit is set, the corresponding group is filtered out and
the messages to that group won't be delivered.
The filter bitmap consists of each bit of 1-based UMP Group number.
The bit 0 is reserved for the future use.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-37-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch enhances the /proc/asound/seq/clients output to show a few
more information about the assigned UMP Endpoint and Blocks.
The "Groups" are shown in 1-based group number to align with the
sequencer client name and port number.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-36-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add new ioctls for sequencer clients to query and set the UMP endpoint
and block information.
As a sequencer client corresponds to a UMP Endpoint, one UMP Endpoint
information can be assigned at most to a single sequencer client while
multiple UMP block infos can be assigned by passing the type with the
offset of block id (i.e. type = block_id + 1).
For the kernel client, only SNDRV_SEQ_IOCTL_GET_CLIENT_UMP_INFO is
allowed.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-35-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Create a sequencer port for broadcasting the all group inputs at the
port number 0. This corresponds to a UMP Endpoint connection;
application can read all UMP events from this port no matter which
group the UMP packet belongs to.
Unlike seq ports for other UMP groups, a UMP Endpoint port has no
SND_SEQ_PORT_TYPE_MIDI_GENERIC bit, so that it won't be treated as a
normal MIDI 1.0 device from legacy applications.
The port is named as "MIDI 2.0" to align with representations on other
operation systems.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-34-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch introduces a new ALSA sequencer client for the kernel UMP
object, snd-seq-ump-client. It's a UMP version of snd-seq-midi
driver, while this driver creates a sequencer client per UMP endpoint
which contains (fixed) 16 ports.
The UMP rawmidi device is opened in APPEND mode for output, so that
multiple sequencer clients can share the same UMP endpoint, as well as
the legacy UMP rawmidi devices that are opened in APPEND mode, too.
For input, on the other hand, the incoming data is processed on the
fly in the dedicated hook, hence it doesn't open a rawmidi device.
The UMP packet group is updated upon delivery depending on the target
sequencer port (which corresponds to the actual UMP group).
Each sequencer port sets a new port type bit,
SNDRV_SEQ_PORT_TYPE_MIDI_UMP, in addition to the other standard
types for MIDI.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-33-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
A sequencer client like seq_dummy rather doesn't want to convert UMP
events but receives / sends as is. Add a new event filter flag to
suppress the automatic UMP conversion and applies accordingly.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-32-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch enables the automatic conversion of UMP events from/to the
legacy ALSA sequencer MIDI events. Also, as UMP itself has two
different modes (MIDI 1.0 and MIDI 2.0), yet another converters
between them are needed, too. Namely, we have conversions between the
legacy and UMP like:
- seq legacy event -> seq UMP MIDI 1.0 event
- seq legacy event -> seq UMP MIDI 2.0 event
- seq UMP MIDI 1.0 event -> seq legacy event
- seq UMP MIDI 2.0 event -> seq legacy event
and the conversions between UMP MIDI 1.0 and 2.0 clients like:
- seq UMP MIDI 1.0 event -> seq UMP MIDI 2.0 event
- seq UMP MIDI 2.0 event -> seq UMP MIDI 1.0 event
The translation is per best-effort; some MIDI 2.0 specific events are
ignored when translated to MIDI 1.0.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-31-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add yet more new filed "ump_group" to snd_seq_port_info for specifying
the associated UMP Group number for each sequencer port. This will be
referred in the upcoming automatic UMP conversion in sequencer core.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-30-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Add a new field "direction" to snd_seq_port_info for allowing a client
to tell the expected direction of the port access. A port might still
allow subscriptions for read/write (e.g. for MIDI-CI) even if the
primary usage of the port is a single direction (either input or
output only). This new "direction" field can help to indicate such
cases.
When the direction is unspecified at creating a port and the port has
either read or write capability, the corresponding direction bits are
set automatically as default.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-29-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This is an extension to ALSA sequencer infrastructure to support the
MIDI 2.0 UMP Endpoint port. It's a "catch-all" port that is supposed
to be present for each UMP Endpoint. When this port is read via
subscription, it sends any events from all ports (UMP Groups) found in
the same client.
A UMP Endpoint port can be created with the new capability bit
SNDRV_SEQ_PORT_CAP_UMP_ENDPOINT. Although the port assignment isn't
strictly defined, it should be the port number 0.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-28-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This extends the ALSA sequencer port capability bit to indicate the
"inactive" flag. When this flag is set, the port is essentially
invisible, and doesn't appear in the port query ioctls, while the
direct access and the connection to this port are still allowed. The
active/inactive state can be flipped dynamically, so that it can be
visible at any time later.
This feature is introduced basically for UMP; some UMP Groups in a UMP
Block may be unassigned, hence those are practically invisible. On
ALSA sequencer, the corresponding sequencer ports will get this new
"inactive" flag to indicate the invisible state.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-27-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Starting from this commit, we add the basic support of UMP (Universal
MIDI Packet) events on ALSA sequencer infrastructure. The biggest
change here is that, for transferring UMP packets that are up to 128
bits, we extend the data payload of ALSA sequencer event record when
the client is declared to support for the new UMP events.
A new event flag bit, SNDRV_SEQ_EVENT_UMP, is defined and it shall be
set for the UMP packet events that have the larger payload of 128
bits, defined as struct snd_seq_ump_event.
For controlling the UMP feature enablement in kernel, a new Kconfig,
CONFIG_SND_SEQ_UMP is introduced. The extended event for UMP is
available only when this Kconfig item is set. Similarly, the size of
the internal snd_seq_event_cell also increases (in 4 bytes) when the
Kconfig item is set. (But the size increase is effective only for
32bit architectures; 64bit archs already have padding there.)
Overall, when CONFIG_SND_SEQ_UMP isn't set, there is no change in the
event and cell, keeping the old sizes.
For applications that want to access the UMP packets, first of all, a
sequencer client has to declare the user-protocol to match with the
latest one via the new SNDRV_SEQ_IOCTL_USER_PVERSION; otherwise it's
treated as if a legacy client without UMP support.
Then the client can switch to the new UMP mode (MIDI 1.0 or MIDI 2.0)
with a new field, midi_version, in snd_seq_client_info. When switched
to UMP mode (midi_version = 1 or 2), the client can write the UMP
events with SNDRV_SEQ_EVENT_UMP flag. For reads, the alignment size
is changed from snd_seq_event (28 bytes) to snd_seq_ump_event (32
bytes). When a UMP sequencer event is delivered to a legacy sequencer
client, it's ignored or handled as an error.
Conceptually, ALSA sequencer client and port correspond to the UMP
Endpoint and Group, respectively; each client may have multiple ports
and each port has the fixed number (16) of channels, total up to 256
channels.
As of this commit, ALSA sequencer core just sends and receives the UMP
events as-is from/to clients. The automatic conversions between the
legacy events and the new UMP events will be implemented in a later
patch.
Along with this commit, bump the sequencer protocol version to 1.0.3.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-26-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
For the future extension of ALSA sequencer ABI, introduce a new ioctl
SNDRV_SEQ_IOCTL_USER_PVERSION. This is similar like the ioctls used
in PCM and other interfaces, for an application to specify its
supporting ABI version.
The use of this ioctl will be mandatory for the upcoming UMP support.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-25-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Some port numbers are special, such as 254 for subscribers and 255 for
broadcast. Return error if application tries to create such a port.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-24-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The client type and the port info validity check should be done before
actually creating a port, instead of unnecessary create-and-scratch.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-23-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
We didn't check if a port with the given port number has been already
present at creating a new port. Check it and return -EBUSY properly
if the port number conflicts.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-22-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Introduce the new helpers, snd_seq_kernel_client_get() and _put() for
kernel client drivers to treat the snd_seq_client more directly.
This allows us to reduce the exported symbols and APIs at each time we
need to access some field in future.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-20-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Create a new variant of snd_seq_expand_var_event() for expanding the
data starting from the given byte offset. It'll be used by the new
UMP sequencer code later.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-19-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
There can be a small memory hole that may not be cleared at expanding
an event with the variable length type. Make sure to clear it.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-18-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
snd_rawmidi_kernel_open() is used only internally from ALSA sequencer,
so far, and parsing the card / device matching table at each open is
redundant, as each sequencer client already gets the rawmidi object
beforehand.
This patch optimizes the path by passing the rawmidi object directly
at snd_rawmidi_kernel_open(). This is also a preparation for the
upcoming UMP rawmidi I/O support.
Reviewed-by: Jaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20230523075358.9672-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
With clang's kernel control flow integrity (kCFI, CONFIG_CFI_CLANG),
indirect call targets are validated against the expected function
pointer prototype to make sure the call target is valid to help mitigate
ROP attacks. If they are not identical, there is a failure at run time,
which manifests as either a kernel panic or thread getting killed.
seq_copy_in_user() and seq_copy_in_kernel() did not have prototypes
matching snd_seq_dump_func_t. Adjust this and remove the casts. There
are not resulting binary output differences.
This was found as a result of Clang's new -Wcast-function-type-strict
flag, which is more sensitive than the simpler -Wcast-function-type,
which only checks for type width mismatches.
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/lkml/202211041527.HD8TLSE1-lkp@intel.com
Cc: Jaroslav Kysela <perex@perex.cz>
Cc: Takashi Iwai <tiwai@suse.com>
Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Cc: alsa-devel@alsa-project.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20221118232346.never.380-kees@kernel.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
It's been reported that there is a possible data-race accessing to the
global card_requested[] array at ALSA sequencer core, which is used
for determining whether to call request_module() for the card or not.
This data race itself is almost harmless, as it might end up with one
extra request_module() call for the already loaded module at most.
But it's still better to fix.
This patch addresses the possible data race of card_requested[] and
client_requested[] arrays by replacing them with bitmask.
It's an atomic operation and can work without locks.
Reported-by: Abhishek Shah <abhishek.shah@columbia.edu>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/CAEHB24_ay6YzARpA1zgCsE7=H9CSJJzux618E=Ka4h0YdKn=qA@mail.gmail.com
Link: https://lore.kernel.org/r/20220823072717.1706-2-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
ALSA OSS sequencer refers to a global variable max_midi_devs at
creating a new port, storing it to its own field. Meanwhile this
variable may be changed by other sequencer events at
snd_seq_oss_midi_check_exit_port() in parallel, which may cause a data
race.
OTOH, this data race itself is almost harmless, as the access to the
MIDI device is done via get_mdev() and it's protected with a refcount,
hence its presence is guaranteed.
Though, it's sill better to address the data-race from the code sanity
POV, and this patch adds the proper spinlock for the protection.
Reported-by: Abhishek Shah <abhishek.shah@columbia.edu>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/CAEHB2493pZRXs863w58QWnUTtv3HHfg85aYhLn5HJHCwxqtHQg@mail.gmail.com
Link: https://lore.kernel.org/r/20220823072717.1706-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
If a driver does not supply a drain operation for outputs, a default code
path will execute msleep(50). Especially for a virtual midi device
this severely limmits the throughput.
This implementation for the virtual midi driver simply flushes the output
workqueue.
Signed-off-by: Stefan Sauer <st_kost@gmx.de>
Link: https://lore.kernel.org/r/20220106124145.17254-1-st_kost@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Currently ALSA sequencer core tries to process the queued events as
much as possible when they become dispatchable. If applications try
to queue too massive events to be processed at the very same timing,
the sequencer core would still try to process such all events, either
in the interrupt context or via some notifier; in either away, it
might be a cause of RCU stall or such problems.
As a potential workaround for those problems, this patch adds the
upper limit of the amount of events to be processed. The remaining
events are processed in the next batch, so they won't be lost.
For the time being, it's limited up to 1000 events per queue, which
should be high enough for any normal usages.
Reported-by: Zqiang <qiang.zhang1211@gmail.com>
Reported-by: syzbot+bb950e68b400ab4f65f8@syzkaller.appspotmail.com
Link: https://lore.kernel.org/r/20211102033222.3849-1-qiang.zhang1211@gmail.com
Link: https://lore.kernel.org/r/20211207165146.2888-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
