The dsp_registers field of struct echoaduio has the volatile modifier,
but it's basically superfluous; the field is accessed only for the
base pointer of readl() and writel(), hence marking with __iomem alone
should suffice. OTOH, having the volatile prefix causes a compile
warning like:
sound/pci/echoaudio/echoaudio.c:1878:14: warning: passing argument 1 of 'iounmap' discards 'volatile' qualifier from pointer target type [-Wdiscarded-qualifiers]
So it's better to drop this superfluous modifier.
Link: https://lore.kernel.org/r/20200803143958.24324-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Distorted audio appears occasionally, affecting either playback or
capture and requiring the affected substream to be closed by all
applications and re-opened.
The best way I have found to reproduce the bug is to use dmix in
combination with Chromium, which opens the audio device multiple times
in threads. Anecdotally, the problems appear to have increased with
faster CPUs. I ruled out 32-bit counter wrapping; it often happens
much earlier.
Since applying this patch I have not had problems, where previously
they would occur several times a day.
The patch targets the following issues:
* Check for progress using the counter from the hardware, not after it
has been truncated to the buffer.
This is a clean way to address a possible bug where if a whole
ringbuffer advances between interrupts, it goes unnoticed.
* Move last_period state from chip to pipe
This more logically belongs as part of pipe, and code is reasier to
read if it is "counter position last time a period elapsed".
Now the code has no references to period count. A period is just
when the regular counter crosses a threshold. This increases
readability and reduces scope for bugs.
* Treat period notification and buffer advance independently:
This helps to clarify what is the responsibility of the interrupt
handler, and what is pcm_pointer().
Removing shared state between these operations means race conditions
are fixed without introducing locks. Synchronisation is only around
the read of pipe->dma_counter. There may be cache line contention
around "struct audiopipe" but I did not have cause to profile this.
Pay attention to be robust where dma_counter wrapping is not a
multiple of period_size or buffer_size.
This is a revised patch based on feedback from Takashi and Giuliano.
Signed-off-by: Mark Hills <mark@xwax.org>
Link: https://lore.kernel.org/r/20200708101848.3457-5-mark@xwax.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Use of atomics does not make these statements robust:
atomic_inc(&chip->opencount);
if (atomic_read(&chip->opencount) > 1 && chip->rate_set)
chip->can_set_rate=0;
and
if (atomic_read(&chip->opencount)) {
if (chip->opencount) {
changed = -EAGAIN;
} else {
changed = set_digital_mode(chip, dmode);
It would be necessary to atomically increment or decrement the value
and use the returned result. And yet we still need to prevent other
threads making use of "can_set_rate" while we set it.
However in all but one case the atomic is misleading as they are already
running with "mode_mutex" held.
Decisions are made on mode setting are often intrinsically connected
to "opencount" because some operations are not permitted unless
there is sole ownership.
So instead simplify this, and use "mode_mutex" as a lock for all reference
counting and mode setting.
Signed-off-by: Mark Hills <mark@xwax.org>
Link: https://lore.kernel.org/r/20200708101848.3457-2-mark@xwax.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Many data fields defined in echoaudio drivers are in little-endian,
hence they should be defined with __le16 or __le32. This makes it
easier to catch the forgotten conversions.
Spotted by sparse, a warning like:
sound/pci/echoaudio/echoaudio_dsp.c:990:36: warning: incorrect type in assignment (different base types)
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Drop pci_device() macro that just leads to chip->pci->dev, and pass it
directly to request_firmware(). It was introduced for allowing the
external alsa-driver kernel module builds. Since it was discontinued
years ago, we should clean it up now.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The echoaudio locally defines TRUE and FALSE. Not only is this
redundant given that C now has a boolean type it results in lots of
warnings as other headers also define these macros, causing duplicate
definitions. Fix this by removing the local defines and converting all
local users to use the standard C true and false instead, simply
removing the macros is less safe due to implicit inclusion of the other
definitons.
[fixed overlooked replacement of FALSE by tiwai]
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
removed all references of snd_printk with the standard dev_* macro.
[a few places degraded to dev_dbg(), too -- tiwai]
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
added reference of struct echoaudio to free_firmware function.
this structure will be later used to get a reference of the card
when converting snd_printk to dev_* in the next patch of the series.
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
CONFIG_HOTPLUG is going away as an option. As result the __dev*
markings will be going away.
Remove use of __devinit, __devexit_p, __devinitdata, __devinitconst,
and __devexit.
Signed-off-by: Bill Pemberton <wfp5p@virginia.edu>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch adds rearranges parts of the initialization code and adds
suspend and resume callbacks.
This patch adds suspend and resume callbacks.
It also rearranges parts of the initialization code so it can be
used in both the first initialization (when the module is loaded we
also have to load default settings) and the resume callback (where
we have to restore the previous settings).
Signed-off-by: Giuliano Pochini <pochini@shiny.it>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This patch implements a simple cache for the firmware files when CONFIG_PM is defined.
This patch changes get_firmware(), free_firmware() and adds
free_firmware_cache(). The first two functions implement a very
simple cache and the latter is used to actually release all the stored
firmwares when the module is unloaded.
When CONFIG_PM is not enabled those functions act as before, that is
free_firmware() releases the firmware immediately and
free_firmware_cache() does nothing.
Signed-off-by: Giuliano Pochini <pochini@shiny.it>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Changes the way the firmware is passed through functions.
When CONFIG_PM is enabled the firmware cannot be released because the
driver will need it again to resume the card.
With this patch the firmware is passed as an index of the struct
firmware card_fw[] in place of a pointer. That same index is then used
to locate the firmware in the firmware cache.
Signed-off-by: Giuliano Pochini <pochini@shiny.it>
Signed-off-by: Takashi Iwai <tiwai@suse.de>