Focusrite Saffire 6 has fixed audioformat quirks with multiple
endpoints assigned to a single altsetting.  Unfortunately the generic
parser couldn't detect the sync endpoint correctly as the implicit
sync due to the missing EP attribute bits.  In the former kernels, it
used to work somehow casually, but it's been broken for a while after
the large code change in 5.11.

This patch cures the regression by the following:
- Allow the static quirk table to provide the sync EP information;
  we just need to fill the fields and let the generic parser skipping
  parsing if sync_ep is already set.
- Add the sync endpoint information to the entry for Saffire 6.

Fixes: 7b0efea4 ("ALSA: usb-audio: Add missing ep_idx in fixed EP quirks")
Reported-and-tested-by: NAndré Kapelrud <a.kapelrud@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220606160910.6926-3-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

When ep_idx is already non-zero, it means usually a capture stream
that is set up explicity by a fixed-format quirk, and applying the
check for generic (non-implicit-fb) sync EPs might hit incorrectly,
resulting in a bogus sync endpoint for the capture stream.

This patch adds a check for the ep_idx and skip if it's a secondary
endpoint.  It's a part of the fixes for regressions on Saffire 6.

Fixes: 7b0efea4 ("ALSA: usb-audio: Add missing ep_idx in fixed EP quirks")
Reported-and-tested-by: NAndré Kapelrud <a.kapelrud@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220606160910.6926-2-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

Devices such as the TC-Helicon GoXLR require the sync endpoint to be
configured in advance of the data endpoint in order for sound output
to work.

This patch simply changes the ordering of EP configuration to resolve
this.

Fixes: bf6313a0 ("ALSA: usb-audio: Refactor endpoint management")
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=215079Signed-off-by: NCraig McLure <craig@mclure.net>
Reviewed-by: NJaroslav Kysela <perex@perex.cz>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220524062115.25968-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

In the previous fix, we increased the max buffer bytes from 1MB to 4MB
so that we can use bigger buffers for the modern HiFi devices with
higher rates, more channels and wider formats.  OTOH, extending this
has a concern that too big buffer is allowed for the lower rates, less
channels and narrower formats; when an application tries to allocate
as big buffer as possible, it'll lead to unexpectedly too huge size.

Also, we had a problem about the inconsistent max buffer and period
bytes for the implicit feedback mode when both streams have different
channels.  This was fixed by the (relatively complex) patch to reduce
the max buffer and period bytes accordingly.

This is an alternative fix for those, a patch to kill two birds with
one stone (*): instead of increasing the max buffer bytes blindly and
applying the reduction per channels, we simply use the hw constraints
for the buffer and period "time".  Meanwhile the max buffer and period
bytes are set unlimited instead.

Since the inconsistency of buffer (and period) bytes comes from the
difference of the channels in the tied streams, as long as we care
only about the buffer (and period) time, it doesn't matter; the buffer
time is same for different channels, although we still allow higher
buffer size.  Similarly, this will allow more buffer bytes for HiFi
devices while it also keeps the reasonable size for the legacy
devices, too.

As of this patch, the max period and buffer time are set to 1 and 2
seconds, which should be large enough for all possible use cases.

(*) No animals were harmed in the making of this patch.

Fixes: 98c27add ("ALSA: usb-audio: Cap upper limits of buffer/period bytes for implicit fb")
Fixes: fee2ec8c ("ALSA: usb-audio: Increase max buffer size")
Link: https://lore.kernel.org/r/20220412130740.18933-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The current limit of max buffer size 1MB seems too small for modern
devices with lots of channels and high sample rates.
Let's make bigger, 4MB.
Reviewed-by: NJaroslav Kysela <perex@perex.cz>
Link: https://lore.kernel.org/r/20220407212740.17920-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

In the implicit feedback mode, some parameters are tied between both
playback and capture streams.  One of the tied parameters is the
period size, and this can be a problem if the device has different
number of channels to both streams.  Assume that an application opens
a playback stream that has an implicit feedback from a capture stream,
and it allocates up to the max period and buffer size as much as
possible.  When the capture device supports only more channels than
the playback, the minimum period and buffer sizes become larger than
the sizes the playback stream took.  That is, the minimum size will be
over the max size the driver limits, and PCM core sees as if no
available configuration is found, returning -EINVAL mercilessly.

For avoiding this problem, we have to look through the counter part of
audioformat list for each sync ep, and checks the channels.  If more
channels are found there, we reduce the max period and buffer sizes
accordingly.

You may wonder that the patch adds only the evaluation of channels
between streams, and what about other parameters?  Both the format and
the rate are tied in the implicit fb mode, hence they are always
identical.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=215792
Fixes: 5a6c3e11 ("ALSA: usb-audio: Add hw constraint for implicit fb sync")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220407211657.15087-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The recent change made mistakenly the stream for capture started at
prepare stage.  Add the stream direction check to avoid it.

Fixes: 9c9a3b9d ("ALSA: usb-audio: Rename early_playback_start flag with lowlatency_playback")
Link: https://lore.kernel.org/r/20211119102629.7476-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The recent regression report revealed that the judgment of the
low-latency playback mode based on the runtime->stop_threshold cannot
work reliably at the prepare stage, as sw_params call may happen at
any time, and PCM dmix actually sets it up after the prepare call.
This ended up with the stall of the stream as PCM ack won't be issued
at all.

For addressing this, check the free-wheeling mode again at the PCM
trigger right before starting the stream again, and allow switching to
the non-LL mode at a late stage.

Fixes: d5f871f8 ("ALSA: usb-audio: Improved lowlatency playback support")
Reported-and-tested-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Link: https://lore.kernel.org/r/20211117161855.m45mxcqszkfcetai@box.shutemov.name
Link: https://lore.kernel.org/r/20211119102459.7055-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The recent support for the improved low-latency playback mode applied
the SNDRV_PCM_INFO_EXPLICIT_SYNC flag for the target streams, but this
was a slight overkill.  The use of the flag above disables effectively
both PCM status and control mmaps, while basically what we want to
track is only about the appl_ptr update.

For less restriction, use a more proper flag,
SNDRV_PCM_INFO_SYNC_APPLPTR instead, which disables only the control
mmap.

Fixes: d5f871f8 ("ALSA: usb-audio: Improved lowlatency playback support")
Link: https://lore.kernel.org/r/20211011103650.10182-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

When a stream is in the implicit feedback mode, it's more or less tied
with a capture stream.  Passing SNDRV_PCM_INFO_JOINT_DUPLEX may help
for user-space to understand the situation.

Link: https://lore.kernel.org/r/20211007083528.4184-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

While draining a stream, ALSA PCM core stops the stream by issuing
snd_pcm_stop() after all data has been sent out.  And, at PCM trigger
stop, currently USB-audio driver kills the in-flight URBs explicitly,
then at sync-stop ops, sync with the finish of all remaining URBs.
This might result in a drop of the drained samples as most of
USB-audio devices / hosts allow relatively long in-flight samples (as
a sort of FIFO).

For avoiding the trimming, this patch changes the stream-stop behavior
during PCM draining state.  Under that condition, the pending URBs
won't be killed.  The leftover in-flight URBs are caught by the
sync-stop operation that shall be performed after the trigger-stop
operation.

Link: https://lore.kernel.org/r/20210929080844.11583-10-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This is another attempt to improve further the handling of playback
stream in the low latency mode.  The latest workaround in commit
4267c5a8 ("ALSA: usb-audio: Work around for XRUN with low latency
playback") revealed that submitting URBs forcibly in advance may
trigger XRUN easily.  In the classical mode, this problem was avoided
by practically delaying the submission of the actual data with the
pre-submissions of silent data before triggering the stream start.
But that is exactly what we want to avoid.

Now, in this patch, instead of the previous workaround, we take a
similar approach as used in the implicit feedback mode.  The URBs are
queued at the PCM trigger start like before, but we check whether the
buffer has been already filled enough before each submission, and
stop queuing if the data overcomes the threshold.  The remaining URBs
are kept in the ready list, and they will be retrieved in the URB
complete callback of other (already queued) URBs.  In the complete
callback, we try to fill the data and submit as much as possible
again.  When there is no more available in-flight URBs that may handle
the pending data, we'll check in PCM ack callback and submit and
process URBs there in addition.  In this way, the amount of in-flight
URBs may vary dynamically and flexibly depending on the available data
without hitting XRUN.

The following things are changed to achieve the behavior above:

* The endpoint prepare callback is changed to return an error code;
  when there is no enough data available, it may return -EAGAIN.
  Currently only prepare_playback_urb() returns the error.

  The evaluation of the available data is a bit messy here; we can't
  check with snd_pcm_avail() at the point of prepare callback (as
  runtime->status->hwptr hasn't been updated yet), hence we manually
  estimate the appl_ptr and compare with the internal hwptr_done to
  calculate the available frames.

* snd_usb_endpoint_start() doesn't submit full URBs if the prepare
  callback returns -EAGAIN, and puts the remaining URBs to the ready
  list for the later submission.

* snd_complete_urb() treats the URBs in the low-latency mode similarly
  like the implicit feedback mode, and submissions are done in
  (now exported) snd_usb_queue_pending_output_urbs().

* snd_usb_queue_pending_output_urbs() again checks the error value
  from the prepare callback.  If it's -EAGAIN for the normal stream
  (i.e. not implicit feedback mode), we push it back to the ready list
  again.

* PCM ack callback is introduced for the playback stream, and it calls
  snd_usb_queue_pending_output_urbs() if there is no in-flight URB
  while the stream is running.  This corresponds to the case where the
  system needs the appl_ptr update for re-submitting a new URB.

* snd_usb_queue_pending_output_urbs() and the prepare EP callback
  receive in_stream_lock argument, which is a bool flag indicating the
  call path from PCM ack.  It's needed for avoiding the deadlock of
  snd_pcm_period_elapsed() calls.

* Set the new SNDRV_PCM_INFO_EXPLICIT_SYNC flag when the new
  low-latency mode is deployed.  This assures catching each applptr
  update even in the mmap mode.

Fixes: 4267c5a8 ("ALSA: usb-audio: Work around for XRUN with low latency playback")
Link: https://lore.kernel.org/r/20210929080844.11583-9-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This is yet more preparation for the upcoming changes.

Extend snd_usb_endpoint_next_packet_size() to check the available
frames and return -EAGAIN if the next packet size is equal or exceeds
the given size.  This will be needed for avoiding XRUN during the low
latency operation.

As of this patch, avail=0 is passed, i.e. the check is skipped and no
behavior change.

Link: https://lore.kernel.org/r/20210929080844.11583-7-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

When a playback stream runs in the implicit feedback mode, its
operation is passive and won't start unless the capture packet is
received.  This behavior contradicts with the low-latency playback
mode, and we should turn off lowlatency_playback flag accordingly.

In theory, we may take the low-latency mode when the playback-first
quirk is set, but it still conflicts with the later operation with the
fixed packet numbers, so it's disabled all together for now.

Link: https://lore.kernel.org/r/20210929080844.11583-6-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The free-wheel stream operation like dmix may not update the appl_ptr
appropriately, and it doesn't fit with the low-latency playback mode.
Disable the low-latency playback operation when the stream is set up
in such a mode.

Link: https://lore.kernel.org/r/20210929080844.11583-5-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This is a preparation patch for the upcoming low-latency improvement
changes.

Rename early_playback_start flag with lowlatency_playback as it's more
intuitive.  The new flag is basically a reverse meaning.

Along with the rename, factor out the code to set the flag to a
function.  This makes the complex condition checks simpler.

Also, the same flag is introduced to snd_usb_endpoint, too, that is
carried from the snd_usb_substream flag.  Currently the endpoint flag
isn't still referred, but will be used in later patches.

Link: https://lore.kernel.org/r/20210929080844.11583-4-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

When a single clock source is shared among several endpoints, we have
to keep the same rate on all active endpoints as long as the clock is
being used.  For dealing with such a case, this patch adds one more
check in the hw params constraint for the rate to take the shared
clocks into account.  The current rate is evaluated from the endpoint
list that applies the same clock source.

BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1190418
Link: https://lore.kernel.org/r/20210929080844.11583-2-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

For making user to switch back to the old playback mode, this patch
adds a new module option 'lowlatency' to snd-usb-audio driver.
When user face a regression due to the recent low-latency playback
support, they can test easily by passing lowlatency=0 option without
rebuilding the kernel.

Fixes: 307cc9ba ("ALSA: usb-audio: Reduce latency at playback start, take#2")
Link: https://lore.kernel.org/r/20210829073830.22686-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The recent change for low latency playback works in most of test cases
but it turned out still to hit errors on some use cases, most notably
with JACK with small buffer sizes.  This is because USB-audio driver
fills up and submits full URBs at the beginning, while the URBs would
return immediately and try to fill more -- that can easily trigger
XRUN.  It was more or less expected, but in the small buffer size, the
problem became pretty obvious.

Fixing this behavior properly would require the change of the
fundamental driver design, so it's no trivial task, unfortunately.
Instead, here we work around the problem just by switching back to the
old method when the given configuration is too fragile with the low
latency stream handling.  As a threshold, we calculate the total
buffer bytes in all plus one URBs, and check whether it's beyond the
PCM buffer bytes.  The one extra URB is needed because XRUN happens at
the next submission after the first round.

Fixes: 307cc9ba ("ALSA: usb-audio: Reduce latency at playback start, take#2")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210827203311.5987-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This is another attempt for the reduction of the latency at the start
of a USB audio playback stream.  The first attempt in the commit
9ce650a7 caused an unexpected regression (a deadlock with pipewire
usage) and was later reverted by the commit 4b820e16.  The devils
are always living in details, of course; the cause of the deadlock was
the call of snd_pcm_period_elapsed() inside prepare_playback_urb()
callback.  In the original code, this callback is never called from
the stream lock context as it's driven solely from the URB complete
callback.  Along with the movement of the URB submission into the
trigger START, this prepare call may be also executed in the stream
lock context, hence it deadlocked with the another lock in
snd_pcm_period_elapsed().  (Note that this happens only conditionally
with a small period size that matches with the URB buffer length,
which was a reason I overlooked during my tests.  Also, the problem
wasn't seen in the capture stream because the capture stream handles
the period-elapsed only at retire callback that isn't executed at the
trigger.)

If it were only about avoiding the deadlock, it'd be possible to use
snd_pcm_period_elapsed_under_stream_lock() as a solution.  However, in
general, the period elapsed notification must be sent after the actual
stream start, and replacing the call wouldn't satisfy the pattern.
A better option is to delay the notification after the stream start
procedure finished, instead.  In the case of USB framework, one of the
fitting place would be the complete callback of the first URB.

So, as a workaround of the deadlock and the order fixes above, in
addition to the re-applying the changes in the commit 9ce650a7,
this patch introduces a new flag indicating the delayed period-elapsed
handling and sets it under the possible deadlock condition
(i.e. prepare callback being called before subs->running is set).
Once when the flag is set, the period-elapsed call is handled at a
later URB complete call instead.

As a reference for the original motivation for the low-latency change,
I cite here again:

| USB-audio driver behaves a bit strangely for the playback stream --
| namely, it starts sending silent packets at PCM prepare state while
| the actual data is submitted at first when the trigger START is
| kicked off.  This is a workaround for the behavior where URBs are
| processed too quickly at the beginning.  That is, if we start
| submitting URBs at trigger START, the first few URBs will be
| immediately completed, and this would result in the immediate
| period-elapsed calls right after the start, which may confuse
| applications.
|
| OTOH, submitting the data after silent URBs would, of course, result
| in a certain delay of the actual data processing, and this is rather
| more serious problem on modern systems, in practice.
|
| This patch tries to revert the workaround and lets the URB
| submission starting at PCM trigger for the playback again.  As far
| as I've tested with various backends (native ALSA, PA, JACK, PW), I
| haven't seen any problems (famous last words :)
|
| Note that the capture stream handling needs no such workaround,
| since the capture is driven per received URB.

Link: https://lore.kernel.org/r/4e71531f-4535-fd46-040e-506a3c256bbd@marcan.st
Link: https://lore.kernel.org/r/s5hbl7li0fe.wl-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>
Link: https://lore.kernel.org/r/20210707112447.27485-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This reverts commit 9ce650a7.

This commit causes watchdog lockups on my machine, and while I have no
idea what the cause is, it bisected right to this commit, and reverting
the change promptly fixes it.

At least occasionally one of the watchdog call traces was

  Call Trace:
    _raw_spin_lock_irqsave+0x35/0x40
    snd_pcm_period_elapsed+0x1b/0xa0 [snd_pcm]
    snd_usb_endpoint_start+0x1a0/0x3c0 [snd_usb_audio]
    start_endpoints+0x23/0x90 [snd_usb_audio]
    snd_usb_substream_playback_trigger+0x7b/0x1a0 [snd_usb_audio]
    snd_pcm_common_ioctl+0x1c44/0x2360 [snd_pcm]
    snd_pcm_ioctl+0x2e/0x40 [snd_pcm]
    __se_sys_ioctl+0x72/0xc0
    do_syscall_64+0x4c/0xa0
    entry_SYSCALL_64_after_hwframe+0x44/0xae

so presumably it's a locking error on that substream spinlock that
snd_pcm_period_elapsed() takes.  But at this point I just want to have a
working system so that I can continue the merge window work tomorrow.

Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

USB-audio driver behaves a bit strangely for the playback stream --
namely, it starts sending silent packets at PCM prepare state while
the actual data is submitted at first when the trigger START is kicked
off.  This is a workaround for the behavior where URBs are processed
too quickly at the beginning.  That is, if we start submitting URBs at
trigger START, the first few URBs will be immediately completed, and
this would result in the immediate period-elapsed calls right after
the start, which may confuse applications.

OTOH, submitting the data after silent URBs would, of course, result
in a certain delay of the actual data processing, and this is rather
more serious problem on modern systems, in practice.

This patch tries to revert the workaround and lets the URB submission
starting at PCM trigger for the playback again.  As far as I've tested
with various backends (native ALSA, PA, JACK, PW), I haven't seen any
problems (famous last words :)

Note that the capture stream handling needs no such workaround, since
the capture is driven per received URB.

Link: https://lore.kernel.org/r/20210601162457.4877-6-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

Just minor code refactoring.  Like DOP DSD code, it can be better in a
separate function for code readability.

Link: https://lore.kernel.org/r/20210601162457.4877-5-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The PCM delay accounting in USB-audio driver is a bit complex to
follow, and this is an attempt to improve the readability and provide
some potential fix.

Basically, the PCM position delay is calculated from two factors: the
in-flight data on URBs and the USB frame counter.  For the playback
stream, we advance the hwptr already at submitting URBs.  Those
"in-flight" data amount is now tracked, and this is used as the base
value for the PCM delay correction.  The in-flight data is decreased
again at URB completion in return.  For the capture stream, OTOH,
there is no in-flight data, hence the delay base is zero.

The USB frame counter is used in addition for correcting the current
position.  The reference frame counter is updated at each submission
and receiving time, and the difference from the current counter value
is taken into account.

In this patch, each in-flight data bytes is recorded in the new
snd_usb_ctx.queued field, and the total in-flight amount is tracked in
snd_usb_substream.inflight_bytes field, as the replacement of
last_delay field.

Note that updating the hwptr after URB completion doesn't work for
PulseAudio who tries to scratch the buffer on the fly; USB-audio is
basically a double-buffer implementation, hence the scratching the
buffer can't work for the already submitted data.  So we always update
hwptr beforehand.  It's not ideal, but the delay account should give
enough correctness.

Link: https://lore.kernel.org/r/20210601162457.4877-4-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

There are a bunch of lines calculating the buffer size in bytes at
each time.  Keep the value in subs->buffer_bytes and use it
consistently for the code simplicity.

Link: https://lore.kernel.org/r/20210601162457.4877-3-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

It's a local function, let's make it static.

Link: https://lore.kernel.org/r/20210601162457.4877-2-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The recent fix for the hw constraints for implicit feedback streams
via commit e4ea77f8 ("ALSA: usb-audio: Always apply the hw
constraints for implicit fb sync") added the check of the matching
endpoints and whether those EPs are already opened.  This is needed
and correct, per se, even for the normal streams without the implicit
feedback, as the endpoint setup is exclusive.

However, it's reported that there seem applications that behave in
unexpected ways to update the hw_params without clearing the previous
setup via hw_free, and those hit a problem now: then hw_params is
called with still the previous EP setup kept, hence it's restricted
with the previous own setup.  Although the obvious fix is to call
snd_pcm_hw_free() API in the application side, it's a kind of
unwelcome change.

This patch tries to ease the situation: in the endpoint check, we add
a couple of more conditions and now skip the endpoint that is being
used only by the stream in question itself.  That is, in addition to
the presence check of ep (ep->cur_audiofmt is non-NULL), when the
following conditions are met, we skip such an ep:
- ep->opened == 1, and
- ep->cur_audiofmt == subs->cur_audiofmt.

subs->cur_audiofmt is non-NULL only if it's a re-setup of hw_params,
and ep->cur_audiofmt points to the currently set up parameters.  So if
those match, it must be this stream itself.

Fixes: e4ea77f8 ("ALSA: usb-audio: Always apply the hw constraints for implicit fb sync")
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=211941
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210228080138.9936-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

In the later patch, we're going to issue the PCM sync_stop calls at
disconnection.  But currently the USB-audio driver can't handle it
because it has a check of shutdown flag for stopping the URBs.  This
is basically superfluous (the stopping URBs are safe at disconnection
state), so let's drop the check.

Fixes: dc5eafe7 ("ALSA: usb-audio: Support PCM sync_stop")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210206203052.15606-4-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The commit f274baa4 ("ALSA: usb-audio: Allow non-vmalloc buffer
for PCM buffers") introduced the mode to allocate coherent pages for
PCM buffers, and it used bus->controller device as its DMA device.
It turned out, however, that bus->sysdev is a more appropriate device
to be used for DMA mapping in HCD code.

This patch corrects the device reference accordingly.

Note that, on most platforms, both point to the very same device,
hence this patch doesn't change anything practically.  But on
platforms like xhcd-plat hcd, the change becomes effective.

Fixes: f274baa4 ("ALSA: usb-audio: Allow non-vmalloc buffer for PCM buffers")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210205144559.29555-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

Since the recent refactoring, it's been reported that some USB-audio
devices (typically webcams) are no longer detected properly by
PulseAudio.  The debug session revealed that it's failing at probing
by PA to try the sample rate 44.1kHz while the device has discrete
sample rates other than 44.1kHz.  But the puzzle was that arecord
works as is, and some other devices with the discrete rates work,
either.

After all, this turned out to be the lack of the dependencies in a few
hw constraint rules: snd_pcm_hw_rule_add() has the (variable)
arguments specifying the dependent parameters, and some functions
didn't set the target parameter itself as the dependencies.  This
resulted in an invalid parameter that could be generated only in a
certain call pattern.  This bug itself has been present in the code,
but it didn't trigger errors just because the rules were casually
avoiding such a corner case.  After the recent refactoring and
cleanup, however, the hw constraints work "as expected", and the
problem surfaced now.

For fixing the problem above, this patch adds the missing dependent
parameters to each snd_pcm_hw_rule() call.

Fixes: bc4e94aa ("ALSA: usb-audio: Handle discrete rates properly in hw constraints")
BugLink: http://bugzilla.opensuse.org/show_bug.cgi?id=1181014
Link: https://lore.kernel.org/r/20210120204554.30177-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

Since the commit 5a6c3e11 ("ALSA: usb-audio: Add hw constraint for
implicit fb sync"), we apply the hw constraints for the implicit
feedback sync to make the secondary open aligned with the already
opened stream setup.  This change assumed that the secondary open is
performed after the first stream has been already set up, and adds the
hw constraints to sync with the first stream's parameters only when
the EP setup for the first stream was confirmed at the open time.
However, most of applications handling the full-duplex operations do
open both playback and capture streams at first, then set up both
streams.  This results in skipping the additional hw constraints since
the counter-part stream hasn't been set up yet at the open of the
second stream, and it eventually leads to "incompatible EP" error in
the end.

This patch corrects the behavior by always applying the hw constraints
for the implicit fb sync.  The hw constraint rules are defined so that
they check the sync EP dynamically at each invocation, instead.  This
covers the concurrent stream setups better and lets the hw refine
calls resolving to the right configuration.

Also this patch corrects a minor error that has existed in the debug
print that isn't built as default.

Fixes: 5a6c3e11 ("ALSA: usb-audio: Add hw constraint for implicit fb sync")
Link: https://lore.kernel.org/r/20210111081611.12790-1-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The code dealing with the implicit feedback mode grew recently, and
it's becoming messy.  As we receive more and more devices that need
the similar handling, it's better to be processed through a table
instead of the open code.

This patch moves the code that is relevant with parsing the implicit
feedback mode and some helpers into another file, implicit.c.  The
detection and the setup of the implicit feedback sync EPs are
rewritten to use the ID/class matching table instead.

There should be no functional changes.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-38-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The capture stream of BOSS GT-001 seems always requiring to be tied
with the playback stream.  OTOH, the playback stream of this device
doesn't seem working in the implicit fb mode, per se, since the
playback must be running before the capture stream.

This patch tries to address the points above:
- Avoid the implicit fb mode for the playback
- Set up a fake sync EP for the capture stream with the hard-coded
  playback stream using the implicit fb mode
Reported-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-37-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

There are two places calculating the next packet size for the playback
stream in the exactly same way.  Provide the single helper for this
purpose and use it from both places gracefully.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-32-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

Some fields like interface and alt_idx in snd_usb_substream are mostly
useless now as they can be referred via either cur_audiofmt or
data_endpoint assigned to the substream.  Drop those, and also assure
the concurrency about the access of cur_audiofmt field.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-31-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The audioformat is referred in many places but most of usages are
read-only.  Let's add const prefix in the possible places.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-28-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

This is an intensive surgery for the endpoint and stream management
for achieving more robust and clean code.

The goals of this patch are:
- More clear endpoint resource changes
- The interface altsetting control in a single place
Below are brief description of the whole changes.

First off, most of the endpoint operations are moved into endpoint.c,
so that the snd_usb_endpoint object is only referred in other places.
The endpoint object is acquired and released via the new functions
snd_usb_endpoint_open() and snd_usb_endpoint_close() that are called
at PCM hw_params and hw_free callbacks, respectively.  Those are
ref-counted and EPs can manage the multiple opens.

The open callback receives the audioformat and hw_params arguments,
and those are used for initializing the EP parameters; especially the
endpoint, interface and altset numbers are read from there, as well as
the PCM parameters like the format, rate and channels.  Those are
stored in snd_usb_endpoint object.  If it's the secondary open, the
function checks whether the given parameters are compatible with the
already opened EP setup, too.

The coupling with a sync EP (including an implicit feedback sync) is
done by the sole snd_usb_endpoint_set_sync() call.

The configuration of each endpoint is done in a single shot via
snd_usb_endpoint_configure() call.  This is the place where most of
PCM configurations are done.  A few flags and special handling in the
snd_usb_substream are dropped along with this change.

A significant difference wrt the configuration from the previous code
is the order of USB host interface setups.  Now the interface is
always disabled at beginning and (re-)enabled at the last step of
snd_usb_endpoint_configure(), in order to be compliant with the
standard UAC2/3.  For UAC1, the interface is set before the parameter
setups since there seem devices that require it (e.g. Yamaha THR10),
just like how it was done in the previous driver code.

The start/stop are almost same as before, also single-shots.  The URB
callbacks need to be set via snd_usb_endpoint_set_callback() like the
previous code at the trigger phase, too.

Finally, the flag for the re-setup is set at the device suspend
through the full EP list, instead of PCM trigger.  This catches the
overlooked cases where the PCM hasn't been running yet but the device
needs the full setup after resume.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-26-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The function to evaluate the match of the parameters with an EP
assumes only the discrete rate tables and doesn't handle the
continuous rates properly.

This patch fixes match_endpoint_audioformats() to handle the
continuous rates.  Also the almost useless debug prints there are
dropped.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-25-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The commit 92adc96f ("ALSA: usb-audio: set the interface format
after resume on Dell WD19") introduced the workaround for the broken
setup after the resume specifically on a Dell dock model.  However,
the full setup should have been performed after the resume on all
devices, as we can't guarantee the same state.  So this patch removes
the conditional check and applies the workaround always.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-24-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>

The prepare_data_urb and retire_data_urb fields of the endpoint object
are set dynamically at PCM trigger start/stop.  Those are evaluated in
the endpoint handler, but there can be a race, especially if two
different PCM substreams are handling the same endpoint for the
implicit feedback case.  Also, the data_subs field of the endpoint is
set and accessed dynamically, too, which has the same risk.

As a slight improvement for the concurrency, this patch introduces the
function to set the callbacks and the data in a shot with the memory
barrier.  In the reader side, it's also fetched with the memory
barrier.

There is still a room of race if prepare and retire callbacks are set
during executing the URB completion.  But such an inconsistency may
happen only for the implicit fb source, i.e. it's only about the
capture stream.  And luckily, the capture stream never sets the
prepare callback, hence the problem doesn't happen practically.
Tested-by: NKeith Milner <kamilner@superlative.org>
Tested-by: NDylan Robinson <dylan_robinson@motu.com>
Link: https://lore.kernel.org/r/20201123085347.19667-23-tiwai@suse.deSigned-off-by: NTakashi Iwai <tiwai@suse.de>