Compensate for the cache lag of 64 frames, and actually populate the
cache. Without these, the playback would start with garbage (which
would be (mostly?) masqueraded by the note's attack phase).
Note that we set the starting address only 61 frames ahead, to
compensate for the interpolator's epsilon. Unlike for PCM playback, we
don't even need to manually silence-fill the first frames in the cache,
because we insert some silence in front of each sample anyway.
A challenge are extremely short samples with a loop end below the cache
size, because a) we'd have to wrap the current address to be within the
loop and b) automatic pre-filling of the cache with the right data does
not work in this case.
We could pre-fill the cache manually, but that's slow, requires
additional code for each sample width, and is made even more complex by
the driver's virtual address space having no contiguous mapping for the
CPU.
We could have the engine fill the cache piece-wise (which is really what
happens when playback is running), but that would also be complex, and
we'd need to wait for the engine to handle each piece, so it wouldn't be
that much faster than the manual fill.
For the case of requiring only one loop iteration prior to reaching the
cache size, we could leverage the engine's looping mechanism around
CCR_CACHELOOPFLAG, but this special case doesn't seem worth the
complexity.
So we just unroll the loop as far as necessary to be able to play back
the sample without any fiddling.
Pedantically, this would be incorrect for loop-until-release samples
with a low loop end which are released very quickly, but that would be
relatively harmless, is not a plausible use case in the first place, and
SoundFont sample mode 3 isn't actually implemented anyway (it's
conflated with mode 1, infinite looping).
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Message-ID: <20240406064830.1029573-16-oswald.buddenhagen@gmx.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Samples are byte-sized in this mode, and thus the offset calculation
needs no shifting.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Message-ID: <20240406064830.1029573-11-oswald.buddenhagen@gmx.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
As already anticipated in the original commit, playback was broken for
very short samples. I just didn't expect it to be an actual problem,
because we're talking about less than 1.5 milliseconds here. But clearly
such wavetable samples do actually exist.
The problem was that for such short samples we'd set the current
position beyond the end of the loop, so we'd run off the end of the
sample and play garbage.
This is a bigger (more audible) problem than the original one, which was
that we'd start playback with garbage (whatever was still in the cache),
which would be mostly masked by the note's attack phase.
So revert to the old behavior for now. We'll subsequently fix it
properly with a bigger patch series.
Note that this isn't a full revert - the dead code is not re-introduced,
because that would be silly.
Fixes: df335e9a8b ("ALSA: emu10k1: fix synthesizer sample playback position and caching")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=218625
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Message-ID: <20240401145805.528794-1-oswald.buddenhagen@gmx.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This is only a very partial fix - the frequency-dependent envelope & LFO
register values aren't adjusted.
But I'm not sure they were even correct at 48 kHz to start with, as most
of them are precalculated by common code which assumes an EMU8K-specific
44.1 kHz word clock, and it seems somewhat unlikely that the hardware's
register interpretation was adjusted to compensate for the different
word clock.
In any case I'm not going to spend time on fixing that, as this code is
unlikely to be actually used by anyone today.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230612191325.1315854-6-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Instead of separate voices, we now allocate non-interleaved channels,
which may in turn contain two interleaved voices each. The higher-level
code keeps only one pointer per channel. The channels are not allocated
in one block any more, as there is no reason to do that. As a
consequence of that, and because it is cleaner regardless, we now let
the allocator store these pointers at a specified location, rather than
returning only the first one and having the calling code deduce the
remaining ones.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518140947.3725394-8-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The voice allocator clearly knows about the field (it resets it), so
it's more consistent (and leads to less duplicated code) to have the
constructor take it as a parameter.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518140947.3725394-7-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
This allows us to drop the code that tries to preserve already allocated
voices upon repeated hw_param callback invocations. Getting it right for
multi-channel voices would otherwise get a bit hairy.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518140947.3725394-5-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Make terminate_voice() actually do at all what it's supposed to do:
instantly and completely shut down the note.
The bogus behavior was mostly harmless, as usually the voice is freed
right afterwards, which implicitly terminates it anyway.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518140339.3722308-2-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Compensate for the cache delay, and actually populate the cache.
Without these, the playback would start with garbage (which would be
(mostly?) masqueraded by the attack phase).
Unlike for the PCM voices, this doesn't try to compensate for the
interpolator read-ahead, because it's pointless to be super-exact here.
Note that this code is probably still broken for particularly short
samples, because we ignore the loop-related parts of CCR. But I'm not
going to reverse-engineer that now ...
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518140339.3722308-1-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
While this nicely denoises the code, the real intent is being able to
write many registers pseudo-atomically, which will come in handy later.
Idea stolen from kX-project.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230518093134.3697955-1-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Using the *_MASK defines for "maximal value" is debatable. I got the
idea from FreeBSD, and it sorta makes sense to me.
Some hunks look a bit incomplete, because code that is going to be
subsequently removed is not touched here.
Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Link: https://lore.kernel.org/r/20230428080732.1697695-1-oswald.buddenhagen@gmx.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
PCI EMU10k1 driver code contains a few assignments in if condition,
which is a bad coding style that may confuse readers and occasionally
lead to bugs.
This patch is merely for coding-style fixes, no functional changes.
Link: https://lore.kernel.org/r/20210608140540.17885-41-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Based on 1 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version this program is distributed in the
hope that it will be useful but without any warranty without even
the implied warranty of merchantability or fitness for a particular
purpose see the gnu general public license for more details you
should have received a copy of the gnu general public license along
with this program if not write to the free software foundation inc
59 temple place suite 330 boston ma 02111 1307 usa
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 1334 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Allison Randal <allison@lohutok.net>
Reviewed-by: Richard Fontana <rfontana@redhat.com>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070033.113240726@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Declare snd_emux_operators structure as const as it is only copied into
another structure. So, snd_emux_operators structures having this property
can be made const.
Signed-off-by: Bhumika Goyal <bhumirks@gmail.com>
Reviewed-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Looks like audigy emu10k2 (probably emu10k1 - sb live too) support two
modes for DMA. Second mode is useful for 64 bit os with more then 2 GB
of ram (fixes problems with big soundfont loading)
1) 32MB from 2 GB address space using 8192 pages (used now as default)
2) 16MB from 4 GB address space using 4096 pages
Mode is set using HCFG_EXPANDED_MEM flag in HCFG register.
Also format of emu10k2 page table is then different.
Signed-off-by: Peter Zubaj <pzubaj@marticonet.sk>
Tested-by: Takashi Iwai <tiwai@suse.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The emu10k1 voice allocator takes voice_lock spinlock. When there is
no empty stream available, it tries to release a voice used by synth,
and calls get_synth_voice. The callback function,
snd_emu10k1_synth_get_voice(), however, also takes the voice_lock,
thus it deadlocks.
The fix is simply removing the voice_lock holds in
snd_emu10k1_synth_get_voice(), as this is always called in the
spinlock context.
Reported-and-tested-by: Arthur Marsh <arthur.marsh@internode.on.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
A simplified version of the semantic patch that finds this problem is as
follows: (http://coccinelle.lip6.fr/)
// <smpl>
@r1@
statement S;
position p,p1;
@@
S@p1;@p
@script:python r2@
p << r1.p;
p1 << r1.p1;
@@
if p[0].line != p1[0].line_end:
cocci.include_match(False)
@@
position r1.p;
@@
-;@p
// </smpl>
Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
These aren't modules, but they do make use of these macros, so
they will need export.h to get that definition. Previously,
they got it via the implicit module.h inclusion.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Kill snd_assert() in sound/pci/*, either removed or replaced with
if () with snd_BUG_ON().
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Modules: EMU10K1/EMU10K2 driver,Common EMU synth
This patch fixes problems with voices cutting off or not
sounding at all.
Signed-off-by: Tim <tedon@rogers.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
