freebsd-dev/sys/dev/sound/pcm/channel.c

2209 lines
53 KiB
C
Raw Normal View History

/*-
2003-09-07 16:28:03 +00:00
* Copyright (c) 1999 Cameron Grant <cg@freebsd.org>
* Portions Copyright by Luigi Rizzo - 1997-99
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include "opt_isa.h"
#include <dev/sound/pcm/sound.h>
#include "feeder_if.h"
SND_DECLARE_FILE("$FreeBSD$");
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
int report_soft_formats = 1;
SYSCTL_INT(_hw_snd, OID_AUTO, report_soft_formats, CTLFLAG_RW,
&report_soft_formats, 1, "report software-emulated formats");
int chn_latency = CHN_LATENCY_DEFAULT;
TUNABLE_INT("hw.snd.latency", &chn_latency);
static int
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
sysctl_hw_snd_latency(SYSCTL_HANDLER_ARGS)
{
int err, val;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
val = chn_latency;
err = sysctl_handle_int(oidp, &val, 0, req);
if (err != 0 || req->newptr == NULL)
return err;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (val < CHN_LATENCY_MIN || val > CHN_LATENCY_MAX)
err = EINVAL;
else
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
chn_latency = val;
return err;
}
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
SYSCTL_PROC(_hw_snd, OID_AUTO, latency, CTLTYPE_INT | CTLFLAG_RW,
0, sizeof(int), sysctl_hw_snd_latency, "I",
"buffering latency (0=low ... 10=high)");
int chn_latency_profile = CHN_LATENCY_PROFILE_DEFAULT;
TUNABLE_INT("hw.snd.latency_profile", &chn_latency_profile);
static int
sysctl_hw_snd_latency_profile(SYSCTL_HANDLER_ARGS)
{
int err, val;
val = chn_latency_profile;
err = sysctl_handle_int(oidp, &val, 0, req);
if (err != 0 || req->newptr == NULL)
return err;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (val < CHN_LATENCY_PROFILE_MIN || val > CHN_LATENCY_PROFILE_MAX)
err = EINVAL;
else
chn_latency_profile = val;
return err;
}
SYSCTL_PROC(_hw_snd, OID_AUTO, latency_profile, CTLTYPE_INT | CTLFLAG_RW,
0, sizeof(int), sysctl_hw_snd_latency_profile, "I",
"buffering latency profile (0=aggresive 1=safe)");
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static int chn_timeout = CHN_TIMEOUT;
TUNABLE_INT("hw.snd.timeout", &chn_timeout);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
#ifdef SND_DEBUG
static int
sysctl_hw_snd_timeout(SYSCTL_HANDLER_ARGS)
{
int err, val;
val = chn_timeout;
err = sysctl_handle_int(oidp, &val, 0, req);
if (err != 0 || req->newptr == NULL)
return err;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (val < CHN_TIMEOUT_MIN || val > CHN_TIMEOUT_MAX)
err = EINVAL;
else
chn_timeout = val;
return err;
}
SYSCTL_PROC(_hw_snd, OID_AUTO, timeout, CTLTYPE_INT | CTLFLAG_RW,
0, sizeof(int), sysctl_hw_snd_timeout, "I",
"interrupt timeout (1 - 10) seconds");
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
#endif
static int chn_usefrags = 0;
TUNABLE_INT("hw.snd.usefrags", &chn_usefrags);
static int chn_syncdelay = -1;
TUNABLE_INT("hw.snd.syncdelay", &chn_syncdelay);
#ifdef SND_DEBUG
SYSCTL_INT(_hw_snd, OID_AUTO, usefrags, CTLFLAG_RW,
&chn_usefrags, 1, "prefer setfragments() over setblocksize()");
SYSCTL_INT(_hw_snd, OID_AUTO, syncdelay, CTLFLAG_RW,
&chn_syncdelay, 1,
"append (0-1000) millisecond trailing buffer delay on each sync");
#endif
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
/**
* @brief Channel sync group lock
*
* Clients should acquire this lock @b without holding any channel locks
* before touching syncgroups or the main syncgroup list.
*/
struct mtx snd_pcm_syncgroups_mtx;
MTX_SYSINIT(pcm_syncgroup, &snd_pcm_syncgroups_mtx, "PCM channel sync group lock", MTX_DEF);
/**
* @brief syncgroups' master list
*
* Each time a channel syncgroup is created, it's added to this list. This
* list should only be accessed with @sa snd_pcm_syncgroups_mtx held.
*
* See SNDCTL_DSP_SYNCGROUP for more information.
*/
struct pcm_synclist snd_pcm_syncgroups = SLIST_HEAD_INITIALIZER(head);
static int chn_buildfeeder(struct pcm_channel *c);
static void
chn_lockinit(struct pcm_channel *c, int dir)
{
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
switch (dir) {
case PCMDIR_PLAY:
c->lock = snd_mtxcreate(c->name, "pcm play channel");
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->intr_cv, "pcmwr");
break;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
case PCMDIR_PLAY_VIRTUAL:
c->lock = snd_mtxcreate(c->name, "pcm virtual play channel");
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->intr_cv, "pcmwrv");
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
break;
case PCMDIR_REC:
c->lock = snd_mtxcreate(c->name, "pcm record channel");
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->intr_cv, "pcmrd");
break;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
case PCMDIR_REC_VIRTUAL:
c->lock = snd_mtxcreate(c->name, "pcm virtual record channel");
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->intr_cv, "pcmrdv");
break;
case 0:
c->lock = snd_mtxcreate(c->name, "pcm fake channel");
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->intr_cv, "pcmfk");
break;
}
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_init(&c->cv, "pcmchn");
}
static void
chn_lockdestroy(struct pcm_channel *c)
{
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_LOCKASSERT(c);
CHN_BROADCAST(&c->cv);
CHN_BROADCAST(&c->intr_cv);
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
cv_destroy(&c->cv);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
cv_destroy(&c->intr_cv);
snd_mtxfree(c->lock);
}
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
/**
* @brief Determine channel is ready for I/O
*
* @retval 1 = ready for I/O
* @retval 0 = not ready for I/O
*/
static int
chn_polltrigger(struct pcm_channel *c)
{
struct snd_dbuf *bs = c->bufsoft;
unsigned amt, lim;
CHN_LOCKASSERT(c);
if (c->flags & CHN_F_MAPPED) {
if (sndbuf_getprevblocks(bs) == 0)
return 1;
2001-03-25 21:43:24 +00:00
else
return (sndbuf_getblocks(bs) > sndbuf_getprevblocks(bs))? 1 : 0;
} else {
amt = (c->direction == PCMDIR_PLAY)? sndbuf_getfree(bs) : sndbuf_getready(bs);
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
#if 0
lim = (c->flags & CHN_F_HAS_SIZE)? sndbuf_getblksz(bs) : 1;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
#endif
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
lim = c->lw;
return (amt >= lim) ? 1 : 0;
}
return 0;
}
static int
chn_pollreset(struct pcm_channel *c)
{
struct snd_dbuf *bs = c->bufsoft;
CHN_LOCKASSERT(c);
sndbuf_updateprevtotal(bs);
return 1;
}
static void
chn_wakeup(struct pcm_channel *c)
{
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
struct snd_dbuf *bs;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
struct pcm_channel *ch;
CHN_LOCKASSERT(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
bs = c->bufsoft;
if (CHN_EMPTY(c, children.busy)) {
if (SEL_WAITING(sndbuf_getsel(bs)) && chn_polltrigger(c))
selwakeuppri(sndbuf_getsel(bs), PRIBIO);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (c->flags & CHN_F_SLEEPING) {
/*
* Ok, I can just panic it right here since it is
* quite obvious that we never allow multiple waiters
* from userland. I'm too generous...
*/
CHN_BROADCAST(&c->intr_cv);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
}
} else {
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
CHN_FOREACH(ch, c, children.busy) {
CHN_LOCK(ch);
chn_wakeup(ch);
CHN_UNLOCK(ch);
}
}
}
static int
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
chn_sleep(struct pcm_channel *c, int timeout)
{
int ret;
CHN_LOCKASSERT(c);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (c->flags & CHN_F_DEAD)
return (EINVAL);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
c->flags |= CHN_F_SLEEPING;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = cv_timedwait_sig(&c->intr_cv, c->lock, timeout);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
c->flags &= ~CHN_F_SLEEPING;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return ((c->flags & CHN_F_DEAD) ? EINVAL : ret);
}
/*
* chn_dmaupdate() tracks the status of a dma transfer,
* updating pointers.
*/
static unsigned int
chn_dmaupdate(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
unsigned int delta, old, hwptr, amt;
KASSERT(sndbuf_getsize(b) > 0, ("bufsize == 0"));
CHN_LOCKASSERT(c);
old = sndbuf_gethwptr(b);
hwptr = chn_getptr(c);
delta = (sndbuf_getsize(b) + hwptr - old) % sndbuf_getsize(b);
sndbuf_sethwptr(b, hwptr);
if (c->direction == PCMDIR_PLAY) {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
amt = min(delta, sndbuf_getready(b));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
amt -= amt % sndbuf_getbps(b);
if (amt > 0)
sndbuf_dispose(b, NULL, amt);
} else {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
amt = min(delta, sndbuf_getfree(b));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
amt -= amt % sndbuf_getbps(b);
if (amt > 0)
sndbuf_acquire(b, NULL, amt);
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (snd_verbose > 3 && CHN_STARTED(c) && delta == 0) {
device_printf(c->dev, "WARNING: %s DMA completion "
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
"too fast/slow ! hwptr=%u, old=%u "
"delta=%u amt=%u ready=%u free=%u\n",
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
CHN_DIRSTR(c), hwptr, old, delta, amt,
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
sndbuf_getready(b), sndbuf_getfree(b));
}
return delta;
}
void
chn_wrupdate(struct pcm_channel *c)
{
int ret;
CHN_LOCKASSERT(c);
KASSERT(c->direction == PCMDIR_PLAY, ("chn_wrupdate on bad channel"));
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if ((c->flags & (CHN_F_MAPPED | CHN_F_VIRTUAL)) || CHN_STOPPED(c))
return;
chn_dmaupdate(c);
ret = chn_wrfeed(c);
/* tell the driver we've updated the primary buffer */
chn_trigger(c, PCMTRIG_EMLDMAWR);
DEB(if (ret)
printf("chn_wrupdate: chn_wrfeed returned %d\n", ret);)
}
int
chn_wrfeed(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
unsigned int ret, amt;
CHN_LOCKASSERT(c);
if ((c->flags & CHN_F_MAPPED) && !(c->flags & CHN_F_CLOSING))
sndbuf_acquire(bs, NULL, sndbuf_getfree(bs));
amt = sndbuf_getfree(b);
ret = (amt > 0) ? sndbuf_feed(bs, b, c, c->feeder, amt) : ENOSPC;
/*
* Possible xruns. There should be no empty space left in buffer.
*/
if (sndbuf_getfree(b) > 0)
c->xruns++;
Whats New: 1. Support wide range sampling rate, as low as 1hz up to int32 max (which is, insane) through new feeder_rate, multiple precisions choice (32/64 bit converter). This is indeed, quite insane, but it does give us more room and flexibility. Plenty sysctl options to adjust resampling characteristics. 2. Support 24/32 bit pcm format conversion through new, much improved, simplified and optimized feeder_fmt. Changes: 1. buffer.c / dsp.c / sound.h * Support for 24/32 AFMT. 2. feeder_rate.c * New implementation of sampling rate conversion with 32/64 bit precision, 1 - int32max hz (which is, ridiculous, yet very addictive). Much improved / smarter buffer management to not cause any missing samples at the end of conversion process * Tunable sysctls for various aspect: hw.snd.feeder_rate_ratemin - minimum allowable sampling rate (default to 4000) hw.snd.feeder_rate_ratemax - maximum allowable sampling rate (default to 1102500) hw.snd.feeder_rate_buffersize - conversion buffer size (default to 8192) hw.snd.feeder_rate_scaling - scaling / conversion method (please refer to the source for explaination). Default to previous implementation type. 3. feeder_fmt.c / sound.h * New implementation, support for 24/32bit conversion, optimized, and simplified. Few routines has been removed (8 to xlaw, 16 to 8). It just doesn't make sense. 4. channel.c * Support for 24/32 AFMT * Fix wrong xruns increment, causing incorrect underruns statistic while using vchans. 5. vchan.c * Support for 24/32 AFMT * Proper speed / rate detection especially for fixed rate ac97. User can override it using kernel hint: hint.pcm.<unit>.vchanrate="xxxx". Notes / Issues: * Virtual Channels (vchans) Enabling vchans can really, really help to solve overrun issues. This is quite understandable, because it operates entirely within its own buffering system without relying on hardware interrupt / state. Even if you don't need vchan, just enable single channel can help much. Few soundcards (notably via8233x, sblive, possibly others) have their own hardware multi channel, and this is unfortunately beyond vchan reachability. * The arrival of 24/32 also come with a price. Applications that can do 24/32bit playback need to be recompiled (notably mplayer). Use (recompiled) mplayer to experiment / test / debug this various format using -af format=fmt. Note that 24bit seeking in mplayer is a little bit broken, sometimes can cause silence or loud static noise. Pausing / seeking few times can solve this problem. You don't have to rebuild world entirely for this. Simply copy /usr/src/sys/sys/soundcard.h to /usr/include/sys/soundcard.h would suffice. Few drivers also need recompilation, and this can be done via /usr/src/sys/modules/sound/. Support for 24bit hardware playback is beyond the scope of this changes. That would require spessific hardware driver changes. * Don't expect playing 9999999999hz is a wise decision. Be reasonable. The new feeder_rate implemention provide flexibility, not insanity. You can easily chew up your CPU with this kind of mind instability. Please use proper mosquito repellent device for this obvious cracked brain attempt. As for testing purposes, you can use (again) mplayer to generate / play with different sampling rate. Use something like "mplayer -af resample=192000:0:0 <files>". Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my> Tested by: multimedia@
2005-07-31 16:16:22 +00:00
if (sndbuf_getfree(b) < amt)
chn_wakeup(c);
return ret;
}
static void
chn_wrintr(struct pcm_channel *c)
{
int ret;
CHN_LOCKASSERT(c);
/* update pointers in primary buffer */
2001-03-25 21:43:24 +00:00
chn_dmaupdate(c);
/* ...and feed from secondary to primary */
ret = chn_wrfeed(c);
/* tell the driver we've updated the primary buffer */
2001-03-25 21:43:24 +00:00
chn_trigger(c, PCMTRIG_EMLDMAWR);
DEB(if (ret)
printf("chn_wrintr: chn_wrfeed returned %d\n", ret);)
}
/*
* user write routine - uiomove data into secondary buffer, trigger if necessary
* if blocking, sleep, rinse and repeat.
*
* called externally, so must handle locking
*/
int
chn_write(struct pcm_channel *c, struct uio *buf)
{
struct snd_dbuf *bs = c->bufsoft;
void *off;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
int ret, timeout, sz, t, p;
CHN_LOCKASSERT(c);
ret = 0;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
timeout = chn_timeout * hz;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (ret == 0 && buf->uio_resid > 0) {
sz = min(buf->uio_resid, sndbuf_getfree(bs));
if (sz > 0) {
/*
* The following assumes that the free space in
* the buffer can never be less around the
* unlock-uiomove-lock sequence.
*/
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (ret == 0 && sz > 0) {
p = sndbuf_getfreeptr(bs);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
t = min(sz, sndbuf_getsize(bs) - p);
off = sndbuf_getbufofs(bs, p);
CHN_UNLOCK(c);
ret = uiomove(off, t, buf);
CHN_LOCK(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sz -= t;
sndbuf_acquire(bs, NULL, t);
}
ret = 0;
if (CHN_STOPPED(c) && !(c->flags & CHN_F_NOTRIGGER)) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = chn_start(c, 0);
if (ret != 0)
c->flags |= CHN_F_DEAD;
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
} else if (c->flags & (CHN_F_NBIO | CHN_F_NOTRIGGER)) {
/**
* @todo Evaluate whether EAGAIN is truly desirable.
* 4Front drivers behave like this, but I'm
* not sure if it at all violates the "write
* should be allowed to block" model.
*
* The idea is that, while set with CHN_F_NOTRIGGER,
* a channel isn't playing, *but* without this we
* end up with "interrupt timeout / channel dead".
*/
ret = EAGAIN;
} else {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = chn_sleep(c, timeout);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (ret == EAGAIN) {
ret = EINVAL;
c->flags |= CHN_F_DEAD;
printf("%s: play interrupt timeout, "
"channel dead\n", c->name);
} else if (ret == ERESTART || ret == EINTR)
c->flags |= CHN_F_ABORTING;
}
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (ret);
}
/*
* Feed new data from the read buffer. Can be called in the bottom half.
*/
int
chn_rdfeed(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
unsigned int ret, amt;
CHN_LOCKASSERT(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (c->flags & CHN_F_MAPPED)
sndbuf_dispose(bs, NULL, sndbuf_getready(bs));
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
amt = sndbuf_getfree(bs);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
ret = (amt > 0) ? sndbuf_feed(b, bs, c, c->feeder, amt) : ENOSPC;
amt = sndbuf_getready(b);
if (amt > 0) {
c->xruns++;
sndbuf_dispose(b, NULL, amt);
}
if (sndbuf_getready(bs) > 0)
chn_wakeup(c);
return ret;
}
void
chn_rdupdate(struct pcm_channel *c)
{
int ret;
CHN_LOCKASSERT(c);
KASSERT(c->direction == PCMDIR_REC, ("chn_rdupdate on bad channel"));
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
if ((c->flags & (CHN_F_MAPPED | CHN_F_VIRTUAL)) || CHN_STOPPED(c))
return;
chn_trigger(c, PCMTRIG_EMLDMARD);
chn_dmaupdate(c);
ret = chn_rdfeed(c);
DEB(if (ret)
printf("chn_rdfeed: %d\n", ret);)
}
/* read interrupt routine. Must be called with interrupts blocked. */
static void
chn_rdintr(struct pcm_channel *c)
{
int ret;
CHN_LOCKASSERT(c);
/* tell the driver to update the primary buffer if non-dma */
2001-03-25 21:43:24 +00:00
chn_trigger(c, PCMTRIG_EMLDMARD);
/* update pointers in primary buffer */
2001-03-25 21:43:24 +00:00
chn_dmaupdate(c);
/* ...and feed from primary to secondary */
ret = chn_rdfeed(c);
}
/*
* user read routine - trigger if necessary, uiomove data from secondary buffer
* if blocking, sleep, rinse and repeat.
*
* called externally, so must handle locking
*/
int
chn_read(struct pcm_channel *c, struct uio *buf)
{
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
struct snd_dbuf *bs = c->bufsoft;
void *off;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
int ret, timeout, sz, t, p;
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STOPPED(c) && !(c->flags & CHN_F_NOTRIGGER)) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = chn_start(c, 0);
if (ret != 0) {
c->flags |= CHN_F_DEAD;
return (ret);
}
}
ret = 0;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
timeout = chn_timeout * hz;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (ret == 0 && buf->uio_resid > 0) {
sz = min(buf->uio_resid, sndbuf_getready(bs));
2001-10-24 12:33:21 +00:00
if (sz > 0) {
/*
* The following assumes that the free space in
* the buffer can never be less around the
* unlock-uiomove-lock sequence.
*/
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (ret == 0 && sz > 0) {
p = sndbuf_getreadyptr(bs);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
t = min(sz, sndbuf_getsize(bs) - p);
off = sndbuf_getbufofs(bs, p);
CHN_UNLOCK(c);
ret = uiomove(off, t, buf);
CHN_LOCK(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sz -= t;
sndbuf_dispose(bs, NULL, t);
}
ret = 0;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
} else if (c->flags & (CHN_F_NBIO | CHN_F_NOTRIGGER))
ret = EAGAIN;
else {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = chn_sleep(c, timeout);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (ret == EAGAIN) {
ret = EINVAL;
c->flags |= CHN_F_DEAD;
printf("%s: record interrupt timeout, "
"channel dead\n", c->name);
} else if (ret == ERESTART || ret == EINTR)
c->flags |= CHN_F_ABORTING;
2000-06-20 23:42:08 +00:00
}
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (ret);
}
void
chn_intr(struct pcm_channel *c)
{
Bring in USB4BSD, Hans Petter Selasky rework of the USB stack that includes significant features and SMP safety. This commit includes a more or less complete rewrite of the *BSD USB stack, including Host Controller and Device Controller drivers and updating all existing USB drivers to use the new USB API: 1) A brief feature list: - A new and mutex enabled USB API. - Many USB drivers are now running Giant free. - Linux USB kernel compatibility layer. - New UGEN backend and libusb library, finally solves the "driver unloading" problem. The new BSD licensed libusb20 library is fully compatible with libusb-0.1.12 from sourceforge. - New "usbconfig" utility, for easy configuration of USB. - Full support for Split transactions, which means you can use your full speed USB audio device on a high speed USB HUB. - Full support for HS ISOC transactions, which makes writing drivers for various HS webcams possible, for example. - Full support for USB on embedded platforms, mostly cache flushing and buffer invalidating stuff. - Safer parsing of USB descriptors. - Autodetect of annoying USB install disks. - Support for USB device side mode, also called USB gadget mode, using the same API like the USB host side. In other words the new USB stack is symmetric with regard to host and device side. - Support for USB transfers like I/O vectors, means more throughput and less interrupts. - ... see the FreeBSD quarterly status reports under "USB project" 2) To enable the driver in the default kernel build: 2.a) Remove all existing USB device options from your kernel config file. 2.b) Add the following USB device options to your kernel configuration file: # USB core support device usb2_core # USB controller support device usb2_controller device usb2_controller_ehci device usb2_controller_ohci device usb2_controller_uhci # USB mass storage support device usb2_storage device usb2_storage_mass # USB ethernet support, requires miibus device usb2_ethernet device usb2_ethernet_aue device usb2_ethernet_axe device usb2_ethernet_cdce device usb2_ethernet_cue device usb2_ethernet_kue device usb2_ethernet_rue device usb2_ethernet_dav # USB wireless LAN support device usb2_wlan device usb2_wlan_rum device usb2_wlan_ral device usb2_wlan_zyd # USB serial device support device usb2_serial device usb2_serial_ark device usb2_serial_bsa device usb2_serial_bser device usb2_serial_chcom device usb2_serial_cycom device usb2_serial_foma device usb2_serial_ftdi device usb2_serial_gensa device usb2_serial_ipaq device usb2_serial_lpt device usb2_serial_mct device usb2_serial_modem device usb2_serial_moscom device usb2_serial_plcom device usb2_serial_visor device usb2_serial_vscom # USB bluetooth support device usb2_bluetooth device usb2_bluetooth_ng # USB input device support device usb2_input device usb2_input_hid device usb2_input_kbd device usb2_input_ms # USB sound and MIDI device support device usb2_sound 2) To enable the driver at runtime: 2.a) Unload all existing USB modules. If USB is compiled into the kernel then you might have to build a new kernel. 2.b) Load the "usb2_xxx.ko" modules under /boot/kernel having the same base name like the kernel device option. Submitted by: Hans Petter Selasky hselasky at c2i dot net Reviewed by: imp, alfred
2008-11-04 02:31:03 +00:00
uint8_t do_unlock;
if (CHN_LOCK_OWNED(c)) {
/*
* Allow sound drivers to call this function with
* "CHN_LOCK()" locked:
*/
do_unlock = 0;
} else {
do_unlock = 1;
CHN_LOCK(c);
}
c->interrupts++;
2000-04-17 17:06:47 +00:00
if (c->direction == PCMDIR_PLAY)
chn_wrintr(c);
else
chn_rdintr(c);
Bring in USB4BSD, Hans Petter Selasky rework of the USB stack that includes significant features and SMP safety. This commit includes a more or less complete rewrite of the *BSD USB stack, including Host Controller and Device Controller drivers and updating all existing USB drivers to use the new USB API: 1) A brief feature list: - A new and mutex enabled USB API. - Many USB drivers are now running Giant free. - Linux USB kernel compatibility layer. - New UGEN backend and libusb library, finally solves the "driver unloading" problem. The new BSD licensed libusb20 library is fully compatible with libusb-0.1.12 from sourceforge. - New "usbconfig" utility, for easy configuration of USB. - Full support for Split transactions, which means you can use your full speed USB audio device on a high speed USB HUB. - Full support for HS ISOC transactions, which makes writing drivers for various HS webcams possible, for example. - Full support for USB on embedded platforms, mostly cache flushing and buffer invalidating stuff. - Safer parsing of USB descriptors. - Autodetect of annoying USB install disks. - Support for USB device side mode, also called USB gadget mode, using the same API like the USB host side. In other words the new USB stack is symmetric with regard to host and device side. - Support for USB transfers like I/O vectors, means more throughput and less interrupts. - ... see the FreeBSD quarterly status reports under "USB project" 2) To enable the driver in the default kernel build: 2.a) Remove all existing USB device options from your kernel config file. 2.b) Add the following USB device options to your kernel configuration file: # USB core support device usb2_core # USB controller support device usb2_controller device usb2_controller_ehci device usb2_controller_ohci device usb2_controller_uhci # USB mass storage support device usb2_storage device usb2_storage_mass # USB ethernet support, requires miibus device usb2_ethernet device usb2_ethernet_aue device usb2_ethernet_axe device usb2_ethernet_cdce device usb2_ethernet_cue device usb2_ethernet_kue device usb2_ethernet_rue device usb2_ethernet_dav # USB wireless LAN support device usb2_wlan device usb2_wlan_rum device usb2_wlan_ral device usb2_wlan_zyd # USB serial device support device usb2_serial device usb2_serial_ark device usb2_serial_bsa device usb2_serial_bser device usb2_serial_chcom device usb2_serial_cycom device usb2_serial_foma device usb2_serial_ftdi device usb2_serial_gensa device usb2_serial_ipaq device usb2_serial_lpt device usb2_serial_mct device usb2_serial_modem device usb2_serial_moscom device usb2_serial_plcom device usb2_serial_visor device usb2_serial_vscom # USB bluetooth support device usb2_bluetooth device usb2_bluetooth_ng # USB input device support device usb2_input device usb2_input_hid device usb2_input_kbd device usb2_input_ms # USB sound and MIDI device support device usb2_sound 2) To enable the driver at runtime: 2.a) Unload all existing USB modules. If USB is compiled into the kernel then you might have to build a new kernel. 2.b) Load the "usb2_xxx.ko" modules under /boot/kernel having the same base name like the kernel device option. Submitted by: Hans Petter Selasky hselasky at c2i dot net Reviewed by: imp, alfred
2008-11-04 02:31:03 +00:00
if (do_unlock) {
CHN_UNLOCK(c);
}
return;
}
u_int32_t
chn_start(struct pcm_channel *c, int force)
{
u_int32_t i, j;
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
int err;
CHN_LOCKASSERT(c);
/* if we're running, or if we're prevented from triggering, bail */
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STARTED(c) || ((c->flags & CHN_F_NOTRIGGER) && !force))
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (EINVAL);
err = 0;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (force) {
i = 1;
j = 0;
} else {
if (c->direction == PCMDIR_REC) {
i = sndbuf_getfree(bs);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
j = (i > 0) ? 1 : sndbuf_getready(b);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
} else {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (sndbuf_getfree(bs) == 0) {
i = 1;
j = 0;
} else {
struct snd_dbuf *pb;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
pb = CHN_BUF_PARENT(c, b);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
i = sndbuf_xbytes(sndbuf_getready(bs), bs, pb);
j = sndbuf_getbps(pb);
}
}
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
if (snd_verbose > 3 && CHN_EMPTY(c, children))
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
printf("%s: %s (%s) threshold i=%d j=%d\n",
__func__, CHN_DIRSTR(c),
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
(c->flags & CHN_F_VIRTUAL) ? "virtual" : "hardware",
i, j);
}
if (i >= j) {
c->flags |= CHN_F_TRIGGERED;
sndbuf_setrun(b, 1);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
c->feedcount = (c->flags & CHN_F_CLOSING) ? 2 : 0;
c->interrupts = 0;
c->xruns = 0;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (c->direction == PCMDIR_PLAY && c->parentchannel == NULL) {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sndbuf_fillsilence(b);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (snd_verbose > 3)
printf("%s: %s starting! (%s) (ready=%d "
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
"force=%d i=%d j=%d intrtimeout=%u "
"latency=%dms)\n",
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
__func__,
(c->flags & CHN_F_HAS_VCHAN) ?
"VCHAN" : "HW",
(c->flags & CHN_F_CLOSING) ? "closing" :
"running",
sndbuf_getready(b),
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
force, i, j, c->timeout,
(sndbuf_getsize(b) * 1000) /
(sndbuf_getbps(b) * sndbuf_getspd(b)));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
err = chn_trigger(c, PCMTRIG_START);
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (err);
}
void
chn_resetbuf(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
c->blocks = 0;
sndbuf_reset(b);
sndbuf_reset(bs);
}
/*
* chn_sync waits until the space in the given channel goes above
* a threshold. The threshold is checked against fl or rl respectively.
* Assume that the condition can become true, do not check here...
*/
int
chn_sync(struct pcm_channel *c, int threshold)
{
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
struct snd_dbuf *b, *bs;
int ret, count, hcount, minflush, resid, residp, syncdelay, blksz;
u_int32_t cflag;
CHN_LOCKASSERT(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (c->direction != PCMDIR_PLAY)
return (EINVAL);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
bs = c->bufsoft;
if ((c->flags & (CHN_F_DEAD | CHN_F_ABORTING)) ||
(threshold < 1 && sndbuf_getready(bs) < 1))
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (0);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
/* if we haven't yet started and nothing is buffered, else start*/
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STOPPED(c)) {
if (threshold > 0 || sndbuf_getready(bs) > 0) {
ret = chn_start(c, 1);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (ret != 0)
return (ret);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
} else
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (0);
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
b = CHN_BUF_PARENT(c, c->bufhard);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
minflush = threshold + sndbuf_xbytes(sndbuf_getready(b), b, bs);
syncdelay = chn_syncdelay;
if (syncdelay < 0 && (threshold > 0 || sndbuf_getready(bs) > 0))
minflush += sndbuf_xbytes(sndbuf_getsize(b), b, bs);
/*
* Append (0-1000) millisecond trailing buffer (if needed)
* for slower / high latency hardwares (notably USB audio)
* to avoid audible truncation.
*/
if (syncdelay > 0)
minflush += (sndbuf_getbps(bs) * sndbuf_getspd(bs) *
((syncdelay > 1000) ? 1000 : syncdelay)) / 1000;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
minflush -= minflush % sndbuf_getbps(bs);
if (minflush > 0) {
threshold = min(minflush, sndbuf_getfree(bs));
sndbuf_clear(bs, threshold);
sndbuf_acquire(bs, NULL, threshold);
minflush -= threshold;
}
resid = sndbuf_getready(bs);
residp = resid;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
blksz = sndbuf_getblksz(b);
if (blksz < 1) {
printf("%s: WARNING: blksz < 1 ! maxsize=%d [%d/%d/%d]\n",
__func__, sndbuf_getmaxsize(b), sndbuf_getsize(b),
sndbuf_getblksz(b), sndbuf_getblkcnt(b));
if (sndbuf_getblkcnt(b) > 0)
blksz = sndbuf_getsize(b) / sndbuf_getblkcnt(b);
if (blksz < 1)
blksz = 1;
}
count = sndbuf_xbytes(minflush + resid, bs, b) / blksz;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
hcount = count;
ret = 0;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (snd_verbose > 3)
printf("%s: [begin] timeout=%d count=%d "
"minflush=%d resid=%d\n", __func__, c->timeout, count,
minflush, resid);
cflag = c->flags & CHN_F_CLOSING;
c->flags |= CHN_F_CLOSING;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (count > 0 && (resid > 0 || minflush > 0)) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
ret = chn_sleep(c, c->timeout);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (ret == ERESTART || ret == EINTR) {
c->flags |= CHN_F_ABORTING;
break;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
} else if (ret == 0 || ret == EAGAIN) {
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
resid = sndbuf_getready(bs);
if (resid == residp) {
--count;
if (snd_verbose > 3)
printf("%s: [stalled] timeout=%d "
"count=%d hcount=%d "
"resid=%d minflush=%d\n",
__func__, c->timeout, count,
hcount, resid, minflush);
} else if (resid < residp && count < hcount) {
++count;
if (snd_verbose > 3)
printf("%s: [resume] timeout=%d "
"count=%d hcount=%d "
"resid=%d minflush=%d\n",
__func__, c->timeout, count,
hcount, resid, minflush);
}
if (minflush > 0 && sndbuf_getfree(bs) > 0) {
threshold = min(minflush,
sndbuf_getfree(bs));
sndbuf_clear(bs, threshold);
sndbuf_acquire(bs, NULL, threshold);
resid = sndbuf_getready(bs);
minflush -= threshold;
}
residp = resid;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
} else
break;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
}
c->flags &= ~CHN_F_CLOSING;
c->flags |= cflag;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (snd_verbose > 3)
printf("%s: timeout=%d count=%d hcount=%d resid=%d residp=%d "
"minflush=%d ret=%d\n",
__func__, c->timeout, count, hcount, resid, residp,
minflush, ret);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (0);
}
/* called externally, handle locking */
int
chn_poll(struct pcm_channel *c, int ev, struct thread *td)
{
struct snd_dbuf *bs = c->bufsoft;
int ret;
CHN_LOCKASSERT(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (!(c->flags & (CHN_F_MAPPED | CHN_F_TRIGGERED))) {
ret = chn_start(c, 1);
if (ret != 0)
return (0);
}
ret = 0;
if (chn_polltrigger(c) && chn_pollreset(c))
ret = ev;
else
selrecord(td, sndbuf_getsel(bs));
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (ret);
}
/*
* chn_abort terminates a running dma transfer. it may sleep up to 200ms.
* it returns the number of bytes that have not been transferred.
*
* called from: dsp_close, dsp_ioctl, with channel locked
*/
int
chn_abort(struct pcm_channel *c)
{
int missing = 0;
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STOPPED(c))
return 0;
c->flags |= CHN_F_ABORTING;
c->flags &= ~CHN_F_TRIGGERED;
/* kill the channel */
chn_trigger(c, PCMTRIG_ABORT);
sndbuf_setrun(b, 0);
if (!(c->flags & CHN_F_VIRTUAL))
chn_dmaupdate(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
missing = sndbuf_getready(bs);
c->flags &= ~CHN_F_ABORTING;
2001-03-25 21:43:24 +00:00
return missing;
}
/*
* this routine tries to flush the dma transfer. It is called
* on a close of a playback channel.
* first, if there is data in the buffer, but the dma has not yet
* begun, we need to start it.
* next, we wait for the play buffer to drain
* finally, we stop the dma.
*
* called from: dsp_close, not valid for record channels.
*/
int
chn_flush(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
CHN_LOCKASSERT(c);
KASSERT(c->direction == PCMDIR_PLAY, ("chn_flush on bad channel"));
DEB(printf("chn_flush: c->flags 0x%08x\n", c->flags));
2001-03-25 21:43:24 +00:00
c->flags |= CHN_F_CLOSING;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
chn_sync(c, 0);
c->flags &= ~CHN_F_TRIGGERED;
/* kill the channel */
chn_trigger(c, PCMTRIG_ABORT);
sndbuf_setrun(b, 0);
c->flags &= ~CHN_F_CLOSING;
return 0;
}
int
fmtvalid(u_int32_t fmt, u_int32_t *fmtlist)
{
int i;
for (i = 0; fmtlist[i]; i++)
if (fmt == fmtlist[i])
return 1;
return 0;
}
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
static struct afmtstr_table default_afmtstr_table[] = {
{ "alaw", AFMT_A_LAW }, { "mulaw", AFMT_MU_LAW },
{ "u8", AFMT_U8 }, { "s8", AFMT_S8 },
{ "s16le", AFMT_S16_LE }, { "s16be", AFMT_S16_BE },
{ "u16le", AFMT_U16_LE }, { "u16be", AFMT_U16_BE },
{ "s24le", AFMT_S24_LE }, { "s24be", AFMT_S24_BE },
{ "u24le", AFMT_U24_LE }, { "u24be", AFMT_U24_BE },
{ "s32le", AFMT_S32_LE }, { "s32be", AFMT_S32_BE },
{ "u32le", AFMT_U32_LE }, { "u32be", AFMT_U32_BE },
{ NULL, 0 },
};
int
afmtstr_swap_sign(char *s)
{
if (s == NULL || strlen(s) < 2) /* full length of "s8" */
return 0;
if (*s == 's')
*s = 'u';
else if (*s == 'u')
*s = 's';
else
return 0;
return 1;
}
int
afmtstr_swap_endian(char *s)
{
if (s == NULL || strlen(s) < 5) /* full length of "s16le" */
return 0;
if (s[3] == 'l')
s[3] = 'b';
else if (s[3] == 'b')
s[3] = 'l';
else
return 0;
return 1;
}
u_int32_t
afmtstr2afmt(struct afmtstr_table *tbl, const char *s, int stereo)
{
size_t fsz, sz;
sz = (s == NULL) ? 0 : strlen(s);
if (sz > 1) {
if (tbl == NULL)
tbl = default_afmtstr_table;
for (; tbl->fmtstr != NULL; tbl++) {
fsz = strlen(tbl->fmtstr);
if (sz < fsz)
continue;
if (strncmp(s, tbl->fmtstr, fsz) != 0)
continue;
if (fsz == sz)
return tbl->format |
((stereo) ? AFMT_STEREO : 0);
if ((sz - fsz) < 2 || s[fsz] != ':')
break;
/*
* For now, just handle mono/stereo.
*/
if ((s[fsz + 2] == '\0' && (s[fsz + 1] == 'm' ||
s[fsz + 1] == '1')) ||
strcmp(s + fsz + 1, "mono") == 0)
return tbl->format;
if ((s[fsz + 2] == '\0' && (s[fsz + 1] == 's' ||
s[fsz + 1] == '2')) ||
strcmp(s + fsz + 1, "stereo") == 0)
return tbl->format | AFMT_STEREO;
break;
}
}
return 0;
}
u_int32_t
afmt2afmtstr(struct afmtstr_table *tbl, u_int32_t afmt, char *dst,
size_t len, int type, int stereo)
{
u_int32_t fmt = 0;
char *fmtstr = NULL, *tag = "";
if (tbl == NULL)
tbl = default_afmtstr_table;
for (; tbl->format != 0; tbl++) {
if (tbl->format == 0)
break;
if ((afmt & ~AFMT_STEREO) != tbl->format)
continue;
fmt = afmt;
fmtstr = tbl->fmtstr;
break;
}
if (fmt != 0 && fmtstr != NULL && dst != NULL && len > 0) {
strlcpy(dst, fmtstr, len);
switch (type) {
case AFMTSTR_SIMPLE:
tag = (fmt & AFMT_STEREO) ? ":s" : ":m";
break;
case AFMTSTR_NUM:
tag = (fmt & AFMT_STEREO) ? ":2" : ":1";
break;
case AFMTSTR_FULL:
tag = (fmt & AFMT_STEREO) ? ":stereo" : ":mono";
break;
case AFMTSTR_NONE:
default:
break;
}
if (strlen(tag) > 0 && ((stereo && !(fmt & AFMT_STEREO)) || \
(!stereo && (fmt & AFMT_STEREO))))
strlcat(dst, tag, len);
}
return fmt;
}
int
chn_reset(struct pcm_channel *c, u_int32_t fmt)
{
int hwspd, r;
CHN_LOCKASSERT(c);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
c->feedcount = 0;
c->flags &= CHN_F_RESET;
c->interrupts = 0;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
c->timeout = 1;
c->xruns = 0;
r = CHANNEL_RESET(c->methods, c->devinfo);
if (fmt != 0) {
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
#if 0
hwspd = DSP_DEFAULT_SPEED;
/* only do this on a record channel until feederbuilder works */
if (c->direction == PCMDIR_REC)
RANGE(hwspd, chn_getcaps(c)->minspeed, chn_getcaps(c)->maxspeed);
c->speed = hwspd;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
#endif
hwspd = chn_getcaps(c)->minspeed;
c->speed = hwspd;
if (r == 0)
r = chn_setformat(c, fmt);
if (r == 0)
r = chn_setspeed(c, hwspd);
#if 0
if (r == 0)
r = chn_setvolume(c, 100, 100);
#endif
2000-04-17 17:06:47 +00:00
}
if (r == 0)
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
r = chn_setlatency(c, chn_latency);
if (r == 0) {
2001-03-25 21:43:24 +00:00
chn_resetbuf(c);
r = CHANNEL_RESETDONE(c->methods, c->devinfo);
}
return r;
}
int
chn_init(struct pcm_channel *c, void *devinfo, int dir, int direction)
{
struct feeder_class *fc;
struct snd_dbuf *b, *bs;
int ret;
if (chn_timeout < CHN_TIMEOUT_MIN || chn_timeout > CHN_TIMEOUT_MAX)
chn_timeout = CHN_TIMEOUT;
chn_lockinit(c, dir);
b = NULL;
bs = NULL;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
CHN_INIT(c, children);
CHN_INIT(c, children.busy);
c->devinfo = NULL;
c->feeder = NULL;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
c->latency = -1;
c->timeout = 1;
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
ret = ENOMEM;
b = sndbuf_create(c->dev, c->name, "primary", c);
if (b == NULL)
goto out;
bs = sndbuf_create(c->dev, c->name, "secondary", c);
if (bs == NULL)
goto out;
CHN_LOCK(c);
ret = EINVAL;
fc = feeder_getclass(NULL);
if (fc == NULL)
goto out;
if (chn_addfeeder(c, fc, NULL))
goto out;
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
/*
* XXX - sndbuf_setup() & sndbuf_resize() expect to be called
* with the channel unlocked because they are also called
* from driver methods that don't know about locking
*/
CHN_UNLOCK(c);
sndbuf_setup(bs, NULL, 0);
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
CHN_LOCK(c);
c->bufhard = b;
c->bufsoft = bs;
c->flags = 0;
c->feederflags = 0;
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
c->sm = NULL;
ret = ENODEV;
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
CHN_UNLOCK(c); /* XXX - Unlock for CHANNEL_INIT() malloc() call */
c->devinfo = CHANNEL_INIT(c->methods, devinfo, b, c, direction);
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
CHN_LOCK(c);
if (c->devinfo == NULL)
goto out;
ret = ENOMEM;
if ((sndbuf_getsize(b) == 0) && ((c->flags & CHN_F_VIRTUAL) == 0))
goto out;
ret = chn_setdir(c, direction);
if (ret)
goto out;
ret = sndbuf_setfmt(b, AFMT_U8);
if (ret)
goto out;
ret = sndbuf_setfmt(bs, AFMT_U8);
if (ret)
goto out;
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
/**
* @todo Should this be moved somewhere else? The primary buffer
* is allocated by the driver or via DMA map setup, and tmpbuf
* seems to only come into existence in sndbuf_resize().
*/
if (c->direction == PCMDIR_PLAY) {
bs->sl = sndbuf_getmaxsize(bs);
bs->shadbuf = malloc(bs->sl, M_DEVBUF, M_NOWAIT);
if (bs->shadbuf == NULL) {
ret = ENOMEM;
goto out;
}
}
out:
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
CHN_UNLOCK(c);
if (ret) {
if (c->devinfo) {
if (CHANNEL_FREE(c->methods, c->devinfo))
sndbuf_free(b);
}
if (bs)
sndbuf_destroy(bs);
if (b)
sndbuf_destroy(b);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_LOCK(c);
c->flags |= CHN_F_DEAD;
chn_lockdestroy(c);
return ret;
}
return 0;
}
int
chn_kill(struct pcm_channel *c)
{
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
if (CHN_STARTED(c)) {
CHN_LOCK(c);
chn_trigger(c, PCMTRIG_ABORT);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
CHN_UNLOCK(c);
}
while (chn_removefeeder(c) == 0)
;
if (CHANNEL_FREE(c->methods, c->devinfo))
sndbuf_free(b);
sndbuf_destroy(bs);
sndbuf_destroy(b);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_LOCK(c);
c->flags |= CHN_F_DEAD;
chn_lockdestroy(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (0);
}
int
chn_setdir(struct pcm_channel *c, int dir)
{
#ifdef DEV_ISA
struct snd_dbuf *b = c->bufhard;
#endif
int r;
CHN_LOCKASSERT(c);
c->direction = dir;
r = CHANNEL_SETDIR(c->methods, c->devinfo, c->direction);
#ifdef DEV_ISA
if (!r && SND_DMA(b))
sndbuf_dmasetdir(b, c->direction);
#endif
return r;
}
int
chn_setvolume(struct pcm_channel *c, int left, int right)
{
CHN_LOCKASSERT(c);
/* should add a feeder for volume changing if channel returns -1 */
if (left > 100)
left = 100;
if (left < 0)
left = 0;
if (right > 100)
right = 100;
if (right < 0)
right = 0;
c->volume = left | (right << 8);
2001-03-25 21:43:24 +00:00
return 0;
}
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
static u_int32_t
round_pow2(u_int32_t v)
{
u_int32_t ret;
if (v < 2)
v = 2;
ret = 0;
while (v >> ret)
ret++;
ret = 1 << (ret - 1);
while (ret < v)
ret <<= 1;
return ret;
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static u_int32_t
round_blksz(u_int32_t v, int round)
{
u_int32_t ret, tmp;
if (round < 1)
round = 1;
ret = min(round_pow2(v), CHN_2NDBUFMAXSIZE >> 1);
if (ret > v && (ret >> 1) > 0 && (ret >> 1) >= ((v * 3) >> 2))
ret >>= 1;
tmp = ret - (ret % round);
while (tmp < 16 || tmp < round) {
ret <<= 1;
tmp = ret - (ret % round);
}
return ret;
}
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
/*
* 4Front call it DSP Policy, while we call it "Latency Profile". The idea
* is to keep 2nd buffer short so that it doesn't cause long queue during
* buffer transfer.
*
* Latency reference table for 48khz stereo 16bit: (PLAY)
*
* +---------+------------+-----------+------------+
* | Latency | Blockcount | Blocksize | Buffersize |
* +---------+------------+-----------+------------+
* | 0 | 2 | 64 | 128 |
* +---------+------------+-----------+------------+
* | 1 | 4 | 128 | 512 |
* +---------+------------+-----------+------------+
* | 2 | 8 | 512 | 4096 |
* +---------+------------+-----------+------------+
* | 3 | 16 | 512 | 8192 |
* +---------+------------+-----------+------------+
* | 4 | 32 | 512 | 16384 |
* +---------+------------+-----------+------------+
* | 5 | 32 | 1024 | 32768 |
* +---------+------------+-----------+------------+
* | 6 | 16 | 2048 | 32768 |
* +---------+------------+-----------+------------+
* | 7 | 8 | 4096 | 32768 |
* +---------+------------+-----------+------------+
* | 8 | 4 | 8192 | 32768 |
* +---------+------------+-----------+------------+
* | 9 | 2 | 16384 | 32768 |
* +---------+------------+-----------+------------+
* | 10 | 2 | 32768 | 65536 |
* +---------+------------+-----------+------------+
*
* Recording need a different reference table. All we care is
* gobbling up everything within reasonable buffering threshold.
*
* Latency reference table for 48khz stereo 16bit: (REC)
*
* +---------+------------+-----------+------------+
* | Latency | Blockcount | Blocksize | Buffersize |
* +---------+------------+-----------+------------+
* | 0 | 512 | 32 | 16384 |
* +---------+------------+-----------+------------+
* | 1 | 256 | 64 | 16384 |
* +---------+------------+-----------+------------+
* | 2 | 128 | 128 | 16384 |
* +---------+------------+-----------+------------+
* | 3 | 64 | 256 | 16384 |
* +---------+------------+-----------+------------+
* | 4 | 32 | 512 | 16384 |
* +---------+------------+-----------+------------+
* | 5 | 32 | 1024 | 32768 |
* +---------+------------+-----------+------------+
* | 6 | 16 | 2048 | 32768 |
* +---------+------------+-----------+------------+
* | 7 | 8 | 4096 | 32768 |
* +---------+------------+-----------+------------+
* | 8 | 4 | 8192 | 32768 |
* +---------+------------+-----------+------------+
* | 9 | 2 | 16384 | 32768 |
* +---------+------------+-----------+------------+
* | 10 | 2 | 32768 | 65536 |
* +---------+------------+-----------+------------+
*
* Calculations for other data rate are entirely based on these reference
* tables. For normal operation, Latency 5 seems give the best, well
* balanced performance for typical workload. Anything below 5 will
* eat up CPU to keep up with increasing context switches because of
* shorter buffer space and usually require the application to handle it
* aggresively through possibly real time programming technique.
*
*/
#define CHN_LATENCY_PBLKCNT_REF \
{{1, 2, 3, 4, 5, 5, 4, 3, 2, 1, 1}, \
{1, 2, 3, 4, 5, 5, 4, 3, 2, 1, 1}}
#define CHN_LATENCY_PBUFSZ_REF \
{{7, 9, 12, 13, 14, 15, 15, 15, 15, 15, 16}, \
{11, 12, 13, 14, 15, 16, 16, 16, 16, 16, 17}}
#define CHN_LATENCY_RBLKCNT_REF \
{{9, 8, 7, 6, 5, 5, 4, 3, 2, 1, 1}, \
{9, 8, 7, 6, 5, 5, 4, 3, 2, 1, 1}}
#define CHN_LATENCY_RBUFSZ_REF \
{{14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 16}, \
{15, 15, 15, 15, 15, 16, 16, 16, 16, 16, 17}}
#define CHN_LATENCY_DATA_REF 192000 /* 48khz stereo 16bit ~ 48000 x 2 x 2 */
static int
chn_calclatency(int dir, int latency, int bps, u_int32_t datarate,
u_int32_t max, int *rblksz, int *rblkcnt)
{
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static int pblkcnts[CHN_LATENCY_PROFILE_MAX + 1][CHN_LATENCY_MAX + 1] =
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LATENCY_PBLKCNT_REF;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static int pbufszs[CHN_LATENCY_PROFILE_MAX + 1][CHN_LATENCY_MAX + 1] =
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LATENCY_PBUFSZ_REF;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static int rblkcnts[CHN_LATENCY_PROFILE_MAX + 1][CHN_LATENCY_MAX + 1] =
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LATENCY_RBLKCNT_REF;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
static int rbufszs[CHN_LATENCY_PROFILE_MAX + 1][CHN_LATENCY_MAX + 1] =
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LATENCY_RBUFSZ_REF;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
u_int32_t bufsz;
int lprofile, blksz, blkcnt;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (latency < CHN_LATENCY_MIN || latency > CHN_LATENCY_MAX ||
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
bps < 1 || datarate < 1 ||
!(dir == PCMDIR_PLAY || dir == PCMDIR_REC)) {
if (rblksz != NULL)
*rblksz = CHN_2NDBUFMAXSIZE >> 1;
if (rblkcnt != NULL)
*rblkcnt = 2;
printf("%s: FAILED dir=%d latency=%d bps=%d "
"datarate=%u max=%u\n",
__func__, dir, latency, bps, datarate, max);
return CHN_2NDBUFMAXSIZE;
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
lprofile = chn_latency_profile;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (dir == PCMDIR_PLAY) {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
blkcnt = pblkcnts[lprofile][latency];
bufsz = pbufszs[lprofile][latency];
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
} else {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
blkcnt = rblkcnts[lprofile][latency];
bufsz = rbufszs[lprofile][latency];
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
bufsz = round_pow2(snd_xbytes(1 << bufsz, CHN_LATENCY_DATA_REF,
datarate));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (bufsz > max)
bufsz = max;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
blksz = round_blksz(bufsz >> blkcnt, bps);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (rblksz != NULL)
*rblksz = blksz;
if (rblkcnt != NULL)
*rblkcnt = 1 << blkcnt;
return blksz << blkcnt;
}
static int
chn_resizebuf(struct pcm_channel *c, int latency,
int blkcnt, int blksz)
{
struct snd_dbuf *b, *bs, *pb;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
int sblksz, sblkcnt, hblksz, hblkcnt, limit = 1;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
int ret;
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if ((c->flags & (CHN_F_MAPPED | CHN_F_TRIGGERED)) ||
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
!(c->direction == PCMDIR_PLAY || c->direction == PCMDIR_REC))
return EINVAL;
if (latency == -1) {
c->latency = -1;
latency = chn_latency;
} else if (latency == -2) {
latency = c->latency;
if (latency < CHN_LATENCY_MIN || latency > CHN_LATENCY_MAX)
latency = chn_latency;
} else if (latency < CHN_LATENCY_MIN || latency > CHN_LATENCY_MAX)
return EINVAL;
else {
c->latency = latency;
limit = 0;
}
bs = c->bufsoft;
b = c->bufhard;
if (!(blksz == 0 || blkcnt == -1) &&
(blksz < 16 || blksz < sndbuf_getbps(bs) || blkcnt < 2 ||
(blksz * blkcnt) > CHN_2NDBUFMAXSIZE))
return EINVAL;
chn_calclatency(c->direction, latency, sndbuf_getbps(bs),
sndbuf_getbps(bs) * sndbuf_getspd(bs), CHN_2NDBUFMAXSIZE,
&sblksz, &sblkcnt);
if (blksz == 0 || blkcnt == -1) {
if (blkcnt == -1)
c->flags &= ~CHN_F_HAS_SIZE;
if (c->flags & CHN_F_HAS_SIZE) {
blksz = sndbuf_getblksz(bs);
blkcnt = sndbuf_getblkcnt(bs);
}
} else
c->flags |= CHN_F_HAS_SIZE;
if (c->flags & CHN_F_HAS_SIZE) {
/*
* The application has requested their own blksz/blkcnt.
* Just obey with it, and let them toast alone. We can
* clamp it to the nearest latency profile, but that would
* defeat the purpose of having custom control. The least
* we can do is round it to the nearest ^2 and align it.
*/
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sblksz = round_blksz(blksz, sndbuf_getbps(bs));
sblkcnt = round_pow2(blkcnt);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
limit = 0;
}
if (c->parentchannel != NULL) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
pb = CHN_BUF_PARENT(c, NULL);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_UNLOCK(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_LOCK(c->parentchannel);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
chn_notify(c->parentchannel, CHN_N_BLOCKSIZE);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_UNLOCK(c->parentchannel);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LOCK(c);
limit = (limit != 0 && pb != NULL) ?
sndbuf_xbytes(sndbuf_getsize(pb), pb, bs) : 0;
c->timeout = c->parentchannel->timeout;
} else {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
hblkcnt = 2;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (c->flags & CHN_F_HAS_SIZE) {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
hblksz = round_blksz(sndbuf_xbytes(sblksz, bs, b),
sndbuf_getbps(b));
hblkcnt = round_pow2(sndbuf_getblkcnt(bs));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
} else
chn_calclatency(c->direction, latency,
sndbuf_getbps(b),
sndbuf_getbps(b) * sndbuf_getspd(b),
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
CHN_2NDBUFMAXSIZE, &hblksz, &hblkcnt);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if ((hblksz << 1) > sndbuf_getmaxsize(b))
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
hblksz = round_blksz(sndbuf_getmaxsize(b) >> 1,
sndbuf_getbps(b));
while ((hblksz * hblkcnt) > sndbuf_getmaxsize(b)) {
if (hblkcnt < 4)
hblksz >>= 1;
else
hblkcnt >>= 1;
}
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
hblksz -= hblksz % sndbuf_getbps(b);
#if 0
hblksz = sndbuf_getmaxsize(b) >> 1;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
hblksz -= hblksz % sndbuf_getbps(b);
hblkcnt = 2;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
#endif
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_UNLOCK(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (chn_usefrags == 0 ||
CHANNEL_SETFRAGMENTS(c->methods, c->devinfo,
hblksz, hblkcnt) < 1)
sndbuf_setblksz(b, CHANNEL_SETBLOCKSIZE(c->methods,
c->devinfo, hblksz));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
CHN_LOCK(c);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
if (!CHN_EMPTY(c, children)) {
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sblksz = round_blksz(
sndbuf_xbytes(sndbuf_getsize(b) >> 1, b, bs),
sndbuf_getbps(bs));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
sblkcnt = 2;
limit = 0;
} else if (limit != 0)
limit = sndbuf_xbytes(sndbuf_getsize(b), b, bs);
/*
* Interrupt timeout
*/
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
c->timeout = ((u_int64_t)hz * sndbuf_getsize(b)) /
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
((u_int64_t)sndbuf_getspd(b) * sndbuf_getbps(b));
if (c->timeout < 1)
c->timeout = 1;
}
if (limit > CHN_2NDBUFMAXSIZE)
limit = CHN_2NDBUFMAXSIZE;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
#if 0
while (limit > 0 && (sblksz * sblkcnt) > limit) {
if (sblkcnt < 4)
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
break;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
sblkcnt >>= 1;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
#endif
while ((sblksz * sblkcnt) < limit)
sblkcnt <<= 1;
while ((sblksz * sblkcnt) > CHN_2NDBUFMAXSIZE) {
if (sblkcnt < 4)
sblksz >>= 1;
else
sblkcnt >>= 1;
}
sblksz -= sblksz % sndbuf_getbps(bs);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (sndbuf_getblkcnt(bs) != sblkcnt || sndbuf_getblksz(bs) != sblksz ||
sndbuf_getsize(bs) != (sblkcnt * sblksz)) {
ret = sndbuf_remalloc(bs, sblkcnt, sblksz);
if (ret != 0) {
printf("%s: Failed: %d %d\n", __func__,
sblkcnt, sblksz);
return ret;
}
}
/*
* OSSv4 docs: "By default OSS will set the low water level equal
* to the fragment size which is optimal in most cases."
*/
c->lw = sndbuf_getblksz(bs);
chn_resetbuf(c);
if (snd_verbose > 3)
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
printf("%s: %s (%s) timeout=%u "
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
"b[%d/%d/%d] bs[%d/%d/%d] limit=%d\n",
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
__func__, CHN_DIRSTR(c),
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
(c->flags & CHN_F_VIRTUAL) ? "virtual" : "hardware",
c->timeout,
sndbuf_getsize(b), sndbuf_getblksz(b),
sndbuf_getblkcnt(b),
sndbuf_getsize(bs), sndbuf_getblksz(bs),
sndbuf_getblkcnt(bs), limit);
return 0;
}
int
chn_setlatency(struct pcm_channel *c, int latency)
{
CHN_LOCKASSERT(c);
/* Destroy blksz/blkcnt, enforce latency profile. */
return chn_resizebuf(c, latency, -1, 0);
}
int
chn_setblocksize(struct pcm_channel *c, int blkcnt, int blksz)
{
CHN_LOCKASSERT(c);
/* Destroy latency profile, enforce blksz/blkcnt */
return chn_resizebuf(c, -1, blkcnt, blksz);
}
static int
chn_tryspeed(struct pcm_channel *c, int speed)
{
struct pcm_feeder *f;
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
struct snd_dbuf *x;
int r, delta;
CHN_LOCKASSERT(c);
DEB(printf("setspeed, channel %s\n", c->name));
DEB(printf("want speed %d, ", speed));
if (speed <= 0)
return EINVAL;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STOPPED(c)) {
r = 0;
c->speed = speed;
sndbuf_setspd(bs, speed);
RANGE(speed, chn_getcaps(c)->minspeed, chn_getcaps(c)->maxspeed);
DEB(printf("try speed %d, ", speed));
sndbuf_setspd(b, CHANNEL_SETSPEED(c->methods, c->devinfo, speed));
DEB(printf("got speed %d\n", sndbuf_getspd(b)));
delta = sndbuf_getspd(b) - sndbuf_getspd(bs);
if (delta < 0)
delta = -delta;
c->feederflags &= ~(1 << FEEDER_RATE);
Whats New: 1. Support wide range sampling rate, as low as 1hz up to int32 max (which is, insane) through new feeder_rate, multiple precisions choice (32/64 bit converter). This is indeed, quite insane, but it does give us more room and flexibility. Plenty sysctl options to adjust resampling characteristics. 2. Support 24/32 bit pcm format conversion through new, much improved, simplified and optimized feeder_fmt. Changes: 1. buffer.c / dsp.c / sound.h * Support for 24/32 AFMT. 2. feeder_rate.c * New implementation of sampling rate conversion with 32/64 bit precision, 1 - int32max hz (which is, ridiculous, yet very addictive). Much improved / smarter buffer management to not cause any missing samples at the end of conversion process * Tunable sysctls for various aspect: hw.snd.feeder_rate_ratemin - minimum allowable sampling rate (default to 4000) hw.snd.feeder_rate_ratemax - maximum allowable sampling rate (default to 1102500) hw.snd.feeder_rate_buffersize - conversion buffer size (default to 8192) hw.snd.feeder_rate_scaling - scaling / conversion method (please refer to the source for explaination). Default to previous implementation type. 3. feeder_fmt.c / sound.h * New implementation, support for 24/32bit conversion, optimized, and simplified. Few routines has been removed (8 to xlaw, 16 to 8). It just doesn't make sense. 4. channel.c * Support for 24/32 AFMT * Fix wrong xruns increment, causing incorrect underruns statistic while using vchans. 5. vchan.c * Support for 24/32 AFMT * Proper speed / rate detection especially for fixed rate ac97. User can override it using kernel hint: hint.pcm.<unit>.vchanrate="xxxx". Notes / Issues: * Virtual Channels (vchans) Enabling vchans can really, really help to solve overrun issues. This is quite understandable, because it operates entirely within its own buffering system without relying on hardware interrupt / state. Even if you don't need vchan, just enable single channel can help much. Few soundcards (notably via8233x, sblive, possibly others) have their own hardware multi channel, and this is unfortunately beyond vchan reachability. * The arrival of 24/32 also come with a price. Applications that can do 24/32bit playback need to be recompiled (notably mplayer). Use (recompiled) mplayer to experiment / test / debug this various format using -af format=fmt. Note that 24bit seeking in mplayer is a little bit broken, sometimes can cause silence or loud static noise. Pausing / seeking few times can solve this problem. You don't have to rebuild world entirely for this. Simply copy /usr/src/sys/sys/soundcard.h to /usr/include/sys/soundcard.h would suffice. Few drivers also need recompilation, and this can be done via /usr/src/sys/modules/sound/. Support for 24bit hardware playback is beyond the scope of this changes. That would require spessific hardware driver changes. * Don't expect playing 9999999999hz is a wise decision. Be reasonable. The new feeder_rate implemention provide flexibility, not insanity. You can easily chew up your CPU with this kind of mind instability. Please use proper mosquito repellent device for this obvious cracked brain attempt. As for testing purposes, you can use (again) mplayer to generate / play with different sampling rate. Use something like "mplayer -af resample=192000:0:0 <files>". Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my> Tested by: multimedia@
2005-07-31 16:16:22 +00:00
/*
* Used to be 500. It was too big!
*/
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (delta > feeder_rate_round)
c->feederflags |= 1 << FEEDER_RATE;
else
sndbuf_setspd(bs, sndbuf_getspd(b));
r = chn_buildfeeder(c);
DEB(printf("r = %d\n", r));
if (r)
goto out;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
if (!(c->feederflags & (1 << FEEDER_RATE)))
goto out;
r = EINVAL;
f = chn_findfeeder(c, FEEDER_RATE);
DEB(printf("feedrate = %p\n", f));
if (f == NULL)
goto out;
x = (c->direction == PCMDIR_REC)? b : bs;
r = FEEDER_SET(f, FEEDRATE_SRC, sndbuf_getspd(x));
DEB(printf("feeder_set(FEEDRATE_SRC, %d) = %d\n", sndbuf_getspd(x), r));
if (r)
goto out;
x = (c->direction == PCMDIR_REC)? bs : b;
r = FEEDER_SET(f, FEEDRATE_DST, sndbuf_getspd(x));
DEB(printf("feeder_set(FEEDRATE_DST, %d) = %d\n", sndbuf_getspd(x), r));
out:
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
if (!r)
r = CHANNEL_SETFORMAT(c->methods, c->devinfo,
sndbuf_getfmt(b));
if (!r)
sndbuf_setfmt(bs, c->format);
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (!r)
r = chn_resizebuf(c, -2, 0, 0);
DEB(printf("setspeed done, r = %d\n", r));
2001-03-25 21:43:24 +00:00
return r;
} else
return EINVAL;
}
int
chn_setspeed(struct pcm_channel *c, int speed)
{
int r, oldspeed = c->speed;
r = chn_tryspeed(c, speed);
if (r) {
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (snd_verbose > 3)
printf("Failed to set speed %d falling back to %d\n",
speed, oldspeed);
r = chn_tryspeed(c, oldspeed);
}
return r;
}
static int
chn_tryformat(struct pcm_channel *c, u_int32_t fmt)
{
struct snd_dbuf *b = c->bufhard;
struct snd_dbuf *bs = c->bufsoft;
int r;
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (CHN_STOPPED(c)) {
DEB(printf("want format %d\n", fmt));
c->format = fmt;
r = chn_buildfeeder(c);
if (r == 0) {
sndbuf_setfmt(bs, c->format);
2001-03-25 21:43:24 +00:00
chn_resetbuf(c);
r = CHANNEL_SETFORMAT(c->methods, c->devinfo, sndbuf_getfmt(b));
if (r == 0)
r = chn_tryspeed(c, c->speed);
2001-03-25 21:43:24 +00:00
}
return r;
} else
return EINVAL;
}
int
chn_setformat(struct pcm_channel *c, u_int32_t fmt)
{
u_int32_t oldfmt = c->format;
int r;
r = chn_tryformat(c, fmt);
if (r) {
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (snd_verbose > 3)
printf("Format change 0x%08x failed, reverting to 0x%08x\n",
fmt, oldfmt);
chn_tryformat(c, oldfmt);
}
return r;
}
int
chn_trigger(struct pcm_channel *c, int go)
{
#ifdef DEV_ISA
struct snd_dbuf *b = c->bufhard;
#endif
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
struct snddev_info *d = c->parentsnddev;
int ret;
CHN_LOCKASSERT(c);
#ifdef DEV_ISA
if (SND_DMA(b) && (go == PCMTRIG_EMLDMAWR || go == PCMTRIG_EMLDMARD))
sndbuf_dmabounce(b);
#endif
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (!PCMTRIG_COMMON(go))
return (CHANNEL_TRIGGER(c->methods, c->devinfo, go));
if (go == c->trigger)
return (0);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
ret = CHANNEL_TRIGGER(c->methods, c->devinfo, go);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (ret != 0)
return (ret);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
switch (go) {
case PCMTRIG_START:
if (snd_verbose > 3)
device_printf(c->dev,
"%s() %s: calling go=0x%08x , "
"prev=0x%08x\n", __func__, c->name, go,
c->trigger);
if (c->trigger != PCMTRIG_START) {
c->trigger = go;
CHN_UNLOCK(c);
pcm_lock(d);
CHN_INSERT_HEAD(d, c, channels.pcm.busy);
pcm_unlock(d);
CHN_LOCK(c);
}
break;
case PCMTRIG_STOP:
case PCMTRIG_ABORT:
if (snd_verbose > 3)
device_printf(c->dev,
"%s() %s: calling go=0x%08x , "
"prev=0x%08x\n", __func__, c->name, go,
c->trigger);
if (c->trigger == PCMTRIG_START) {
c->trigger = go;
CHN_UNLOCK(c);
pcm_lock(d);
CHN_REMOVE(d, c, channels.pcm.busy);
pcm_unlock(d);
CHN_LOCK(c);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
break;
default:
break;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (0);
}
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
/**
* @brief Queries sound driver for sample-aligned hardware buffer pointer index
*
* This function obtains the hardware pointer location, then aligns it to
* the current bytes-per-sample value before returning. (E.g., a channel
* running in 16 bit stereo mode would require 4 bytes per sample, so a
* hwptr value ranging from 32-35 would be returned as 32.)
*
* @param c PCM channel context
* @returns sample-aligned hardware buffer pointer index
*/
int
chn_getptr(struct pcm_channel *c)
{
int hwptr;
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
hwptr = (CHN_STARTED(c)) ? CHANNEL_GETPTR(c->methods, c->devinfo) : 0;
return (hwptr - (hwptr % sndbuf_getbps(c->bufhard)));
}
struct pcmchan_caps *
chn_getcaps(struct pcm_channel *c)
{
CHN_LOCKASSERT(c);
return CHANNEL_GETCAPS(c->methods, c->devinfo);
}
u_int32_t
chn_getformats(struct pcm_channel *c)
{
u_int32_t *fmtlist, fmts;
int i;
fmtlist = chn_getcaps(c)->fmtlist;
fmts = 0;
for (i = 0; fmtlist[i]; i++)
fmts |= fmtlist[i];
/* report software-supported formats */
if (report_soft_formats)
Whats New: 1. Support wide range sampling rate, as low as 1hz up to int32 max (which is, insane) through new feeder_rate, multiple precisions choice (32/64 bit converter). This is indeed, quite insane, but it does give us more room and flexibility. Plenty sysctl options to adjust resampling characteristics. 2. Support 24/32 bit pcm format conversion through new, much improved, simplified and optimized feeder_fmt. Changes: 1. buffer.c / dsp.c / sound.h * Support for 24/32 AFMT. 2. feeder_rate.c * New implementation of sampling rate conversion with 32/64 bit precision, 1 - int32max hz (which is, ridiculous, yet very addictive). Much improved / smarter buffer management to not cause any missing samples at the end of conversion process * Tunable sysctls for various aspect: hw.snd.feeder_rate_ratemin - minimum allowable sampling rate (default to 4000) hw.snd.feeder_rate_ratemax - maximum allowable sampling rate (default to 1102500) hw.snd.feeder_rate_buffersize - conversion buffer size (default to 8192) hw.snd.feeder_rate_scaling - scaling / conversion method (please refer to the source for explaination). Default to previous implementation type. 3. feeder_fmt.c / sound.h * New implementation, support for 24/32bit conversion, optimized, and simplified. Few routines has been removed (8 to xlaw, 16 to 8). It just doesn't make sense. 4. channel.c * Support for 24/32 AFMT * Fix wrong xruns increment, causing incorrect underruns statistic while using vchans. 5. vchan.c * Support for 24/32 AFMT * Proper speed / rate detection especially for fixed rate ac97. User can override it using kernel hint: hint.pcm.<unit>.vchanrate="xxxx". Notes / Issues: * Virtual Channels (vchans) Enabling vchans can really, really help to solve overrun issues. This is quite understandable, because it operates entirely within its own buffering system without relying on hardware interrupt / state. Even if you don't need vchan, just enable single channel can help much. Few soundcards (notably via8233x, sblive, possibly others) have their own hardware multi channel, and this is unfortunately beyond vchan reachability. * The arrival of 24/32 also come with a price. Applications that can do 24/32bit playback need to be recompiled (notably mplayer). Use (recompiled) mplayer to experiment / test / debug this various format using -af format=fmt. Note that 24bit seeking in mplayer is a little bit broken, sometimes can cause silence or loud static noise. Pausing / seeking few times can solve this problem. You don't have to rebuild world entirely for this. Simply copy /usr/src/sys/sys/soundcard.h to /usr/include/sys/soundcard.h would suffice. Few drivers also need recompilation, and this can be done via /usr/src/sys/modules/sound/. Support for 24bit hardware playback is beyond the scope of this changes. That would require spessific hardware driver changes. * Don't expect playing 9999999999hz is a wise decision. Be reasonable. The new feeder_rate implemention provide flexibility, not insanity. You can easily chew up your CPU with this kind of mind instability. Please use proper mosquito repellent device for this obvious cracked brain attempt. As for testing purposes, you can use (again) mplayer to generate / play with different sampling rate. Use something like "mplayer -af resample=192000:0:0 <files>". Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my> Tested by: multimedia@
2005-07-31 16:16:22 +00:00
fmts |= AFMT_MU_LAW|AFMT_A_LAW|AFMT_U32_LE|AFMT_U32_BE|
AFMT_S32_LE|AFMT_S32_BE|AFMT_U24_LE|AFMT_U24_BE|
AFMT_S24_LE|AFMT_S24_BE|AFMT_U16_LE|AFMT_U16_BE|
AFMT_S16_LE|AFMT_S16_BE|AFMT_U8|AFMT_S8;
return fmts;
}
static int
chn_buildfeeder(struct pcm_channel *c)
{
struct feeder_class *fc;
struct pcm_feederdesc desc;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
struct snd_mixer *m;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
u_int32_t tmp[2], type, flags, hwfmt, *fmtlist;
int err;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
char fmtstr[AFMTSTR_MAXSZ];
CHN_LOCKASSERT(c);
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
while (chn_removefeeder(c) == 0)
;
KASSERT((c->feeder == NULL), ("feeder chain not empty"));
c->align = sndbuf_getalign(c->bufsoft);
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
if (CHN_EMPTY(c, children) || c->direction == PCMDIR_REC) {
/*
* Virtual rec need this.
*/
fc = feeder_getclass(NULL);
KASSERT(fc != NULL, ("can't find root feeder"));
err = chn_addfeeder(c, fc, NULL);
if (err) {
DEB(printf("can't add root feeder, err %d\n", err));
return err;
}
c->feeder->desc->out = c->format;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
} else if (c->direction == PCMDIR_PLAY) {
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
if (c->flags & CHN_F_HAS_VCHAN) {
desc.type = FEEDER_MIXER;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
desc.in = c->format;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
} else {
DEB(printf("can't decide which feeder type to use!\n"));
return EOPNOTSUPP;
}
desc.out = c->format;
desc.flags = 0;
fc = feeder_getclass(&desc);
if (fc == NULL) {
DEB(printf("can't find vchan feeder\n"));
return EOPNOTSUPP;
}
err = chn_addfeeder(c, fc, &desc);
if (err) {
DEB(printf("can't add vchan feeder, err %d\n", err));
return err;
}
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
} else
return EOPNOTSUPP;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
/* XXX These are too much.. */
if (c->parentsnddev != NULL && c->parentsnddev->mixer_dev != NULL &&
c->parentsnddev->mixer_dev->si_drv1 != NULL)
m = c->parentsnddev->mixer_dev->si_drv1;
else
m = NULL;
c->feederflags &= ~(1 << FEEDER_VOLUME);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (c->direction == PCMDIR_PLAY && !(c->flags & CHN_F_VIRTUAL) && m &&
(c->parentsnddev->flags & SD_F_SOFTPCMVOL))
c->feederflags |= 1 << FEEDER_VOLUME;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (!(c->flags & CHN_F_VIRTUAL) && c->parentsnddev &&
((c->direction == PCMDIR_PLAY &&
(c->parentsnddev->flags & SD_F_PSWAPLR)) ||
(c->direction == PCMDIR_REC &&
(c->parentsnddev->flags & SD_F_RSWAPLR))))
c->feederflags |= 1 << FEEDER_SWAPLR;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
flags = c->feederflags;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
fmtlist = chn_getcaps(c)->fmtlist;
DEB(printf("feederflags %x\n", flags));
for (type = FEEDER_RATE; type < FEEDER_LAST; type++) {
if (flags & (1 << type)) {
desc.type = type;
desc.in = 0;
desc.out = 0;
desc.flags = 0;
DEB(printf("find feeder type %d, ", type));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
if (type == FEEDER_VOLUME || type == FEEDER_RATE) {
if (c->feeder->desc->out & AFMT_32BIT)
strlcpy(fmtstr,"s32le", sizeof(fmtstr));
else if (c->feeder->desc->out & AFMT_24BIT)
strlcpy(fmtstr, "s24le", sizeof(fmtstr));
else {
/*
* 8bit doesn't provide enough headroom
* for proper processing without
* creating too much noises. Force to
* 16bit instead.
*/
strlcpy(fmtstr, "s16le", sizeof(fmtstr));
}
if (!(c->feeder->desc->out & AFMT_8BIT) &&
c->feeder->desc->out & AFMT_BIGENDIAN)
afmtstr_swap_endian(fmtstr);
if (!(c->feeder->desc->out & (AFMT_A_LAW | AFMT_MU_LAW)) &&
!(c->feeder->desc->out & AFMT_SIGNED))
afmtstr_swap_sign(fmtstr);
desc.in = afmtstr2afmt(NULL, fmtstr, AFMTSTR_MONO_RETURN);
if (desc.in == 0)
desc.in = AFMT_S16_LE;
/* feeder_volume need stereo processing */
if (type == FEEDER_VOLUME ||
c->feeder->desc->out & AFMT_STEREO)
desc.in |= AFMT_STEREO;
desc.out = desc.in;
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
} else if (type == FEEDER_SWAPLR) {
desc.in = c->feeder->desc->out;
desc.in |= AFMT_STEREO;
desc.out = desc.in;
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
fc = feeder_getclass(&desc);
DEB(printf("got %p\n", fc));
if (fc == NULL) {
DEB(printf("can't find required feeder type %d\n", type));
return EOPNOTSUPP;
}
[stage: 5/9] channel.c/channel_if.m: - Macros cleanups, prefer inlined min() over MIN(). - Rework chn_read()/chn_write() for better dead interrupt detection policy. Reduce scheduling overhead by doing pure 5 seconds sleep before giving up, instead of several cycle of brute micro sleeping. - Avoid calling wakeup_one() for non-sleeping channel (for example, vchan parent channel). - EWOULDBLOCK -> EAGAIN. - Fix possible divide-by-zero panic on chn_sync(). - Re-enforce ^2 blocksize policy, since there are too many broken userland apps that blindly assume it without even trying to do serious calculations. - New channel method - CHANNEL_SETFRAGMENTS(), a refined version of CHANNEL_SETBLOCKSIZE(). It accept _both_ blocksize and blockcount arguments, so the driver internals will have better hints for buffering and timing calculations. - Hook FEEDER_SWAPLR into feederchain building process. feeder_fmt.c: - Unified version of various filters, avoiding duplications. - malloc()less feeder_fmt. Informations can be retrieved dynamically by doing table lookup on static data. For cases such as converting from stereo to mono or reducing bit depth where input data is larger than output, cycle remaining available free space until it has been exhausted and start kicking 8 bytes reservoir space from there to complete the remaining requested count. - Introduce FEEDER_SWAPLR. Few super broken hardwares (found on several extremely cheap uaudio stick, possibly others) mistakenly wired left and right channels wrongly, screwing output or input.
2007-03-16 17:16:24 +00:00
if (desc.in == 0 || desc.out == 0)
desc = *fc->desc;
DEB(printf("build fmtchain from 0x%08x to 0x%08x: ", c->feeder->desc->out, fc->desc->in));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
tmp[0] = desc.in;
tmp[1] = 0;
if (chn_fmtchain(c, tmp) == 0) {
DEB(printf("failed\n"));
return ENODEV;
}
DEB(printf("ok\n"));
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
err = chn_addfeeder(c, fc, &desc);
if (err) {
DEB(printf("can't add feeder %p, output 0x%x, err %d\n", fc, fc->desc->out, err));
return err;
}
DEB(printf("added feeder %p, output 0x%x\n", fc, c->feeder->desc->out));
}
}
if (c->direction == PCMDIR_REC) {
tmp[0] = c->format;
tmp[1] = 0;
hwfmt = chn_fmtchain(c, tmp);
} else
hwfmt = chn_fmtchain(c, fmtlist);
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
if (hwfmt == 0 || !fmtvalid(hwfmt, fmtlist)) {
DEB(printf("Invalid hardware format: 0x%08x\n", hwfmt));
return ENODEV;
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
} else if (c->direction == PCMDIR_REC && !CHN_EMPTY(c, children)) {
/*
* Kind of awkward. This whole "MIXER" concept need a
* rethinking, I guess :) . Recording is the inverse
* of Playback, which is why we push mixer vchan down here.
*/
if (c->flags & CHN_F_HAS_VCHAN) {
desc.type = FEEDER_MIXER;
desc.in = c->format;
} else
return EOPNOTSUPP;
desc.out = c->format;
desc.flags = 0;
fc = feeder_getclass(&desc);
if (fc == NULL)
return EOPNOTSUPP;
err = chn_addfeeder(c, fc, &desc);
if (err != 0)
return err;
- channel.h * New definition CHN_F_HAS_VCHAN. - channel.c * Use CHN_F_HAS_VCHAN to mark channel with vchan capability instead of relying on SLIST_EMPTY(&channel->children) == true for better clarification and future possible usages of children (like 'slave' channel). * Various fixes, including blocksize / format bps allignment, better 24bit seeking (mplayer, others). * Improve format chain building, it's now possible to record something to a format non-native to the soundcard through various feeder format converters or to higher sampling rate. This also gains another feature, like doing vchan mixing on non s16le soundcard such as sb8. - sound.c * Increase robustness within various function that handle vchan creation / termination (these function need a total rewrite, but that would cause other major rewrite within various places too!). As far as its robustness can be guaranteed, leave it as is. * Optimize channel ordering, prefer *real* hardware playback channels over virtual channels. cat /dev/sndstat should look better. * Increase sndstat verbosity to include bufsoft/bufhard allocation. - vchan.c * Fix LOR 119. - http://sources.zabbadoz.net/freebsd/lor.html#119 * Reorder / increase robustness of vchan_create() / destroy(). Enforce destroy_dev() during destroy operation, fix possible panic / dangling character device. - http://lists.freebsd.org/pipermail/freebsd-current/2005-May/050308.html * Tolerate a little bit more during mixing process, this should help non s16le soundcards. Note: Recoring in a non-native rate/format may result in overruns. A friendly application is wavrec from audio/wavplay. The problem is under investigation. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 18:10:31 +00:00
}
sndbuf_setfmt(c->bufhard, hwfmt);
if ((flags & (1 << FEEDER_VOLUME))) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
u_int32_t parent;
int vol, left, right;
CHN_UNLOCK(c);
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
vol = mix_get(m, SOUND_MIXER_PCM);
if (vol == -1) {
device_printf(c->dev,
"Soft PCM Volume: Failed to read default value\n");
vol = 100 | (100 << 8);
}
left = vol & 0x7f;
right = (vol >> 8) & 0x7f;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
parent = mix_getparent(m, SOUND_MIXER_PCM);
if (parent != SOUND_MIXER_NONE) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
vol = mix_get(m, parent);
if (vol == -1) {
device_printf(c->dev,
"Soft Volume: Failed to read parent "
"default value\n");
vol = 100 | (100 << 8);
}
left = (left * (vol & 0x7f)) / 100;
right = (right * ((vol >> 8) & 0x7f)) / 100;
}
CHN_LOCK(c);
chn_setvolume(c, left, right);
}
return 0;
}
int
chn_notify(struct pcm_channel *c, u_int32_t flags)
{
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
int err, run, nrun;
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
CHN_LOCKASSERT(c);
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
if (CHN_EMPTY(c, children))
return (ENODEV);
err = 0;
/*
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
* If the hwchan is running, we can't change its rate, format or
* blocksize
*/
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
run = (CHN_STARTED(c)) ? 1 : 0;
if (run)
flags &= CHN_N_VOLUME | CHN_N_TRIGGER;
if (flags & CHN_N_RATE) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
/* XXX I'll make good use of this someday. */
}
if (flags & CHN_N_FORMAT) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
/* XXX I'll make good use of this someday. */
}
if (flags & CHN_N_VOLUME) {
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
/* XXX I'll make good use of this someday. */
}
if (flags & CHN_N_BLOCKSIZE) {
/*
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
* Set to default latency profile
*/
Welcome to Once-a-year Sound Mega-Commit. Enjoy numerous updates and fixes in every sense. General ------- - Multichannel safe, endian safe, format safe * Large part of critical pcm filters such as vchan.c, feeder_rate.c, feeder_volume.c, feeder_fmt.c and feeder.c has been rewritten so that using them does not cause the pcm data to be converted to 16bit little endian. * Macrosses for accessing pcm data safely are defined within sound.h in the form of PCM_READ_* / PCM_WRITE_* * Currently, most of them are probably limited for mono/stereo handling, but the future addition of true multichannel will be much easier. - Low latency operation * Well, this require lot more works to do not just within sound driver, but we're heading towards right direction. Buffer/block sizing within channel.c is rewritten to calculate precise allocation for various combination of sample/data/rate size. As a result, applying correct SNDCTL_DSP_POLICY value will achive expected latency behaviour simmilar to what commercial 4front driver do. * Signal handling fix. ctrl+c of "cat /dev/zero > /dev/dsp" does not result long delay. * Eliminate sound truncation if the sound data is too small. DIY: 1) Download / extract http://people.freebsd.org/~ariff/lowlatency/shortfiles.tar.gz 2) Do a comparison between "cat state*.au > /dev/dsp" and "for x in state*.au ; do cat $x > /dev/dsp ; done" - there should be no "perceivable" differences. Double close for PR kern/31445. CAVEAT: Low latency come with (unbearable) price especially for poorly written applications. Applications that trying to act smarter by requesting (wrong) blocksize/blockcount will suffer the most. Fixup samples/patches can be found at: http://people.freebsd.org/~ariff/ports/ - Switch minimum/maximum sampling rate limit to "1" and "2016000" (48k * 42) due to closer compatibility with 4front driver. Discussed with: marcus@ (long time ago?) - All driver specific sysctls in the form of "hw.snd.pcm%d.*" have been moved to their own dev sysctl nodes, notably: hw.snd.pcm%d.vchans -> dev.pcm.%d.vchans Bump __FreeBSD_version. Driver specific --------------- - Ditto for sysctls. - snd_atiixp, snd_es137x, snd_via8233, snd_hda * Numerous cleanups and fixes. * _EXPERIMENTAL_ polling mode support using simple callout_* mechanisme. This was intended for pure debugging and latency measurement, but proven good enough in few unexpected and rare cases (such as problematic shared IRQ with GIANT devices - USB). Polling can be enabled/disabled through dev.pcm.0.polling. Disabled by default. - snd_ich * Fix possible overflow during speed calibration. Delay final initialization (pcm_setstatus) after calibration finished. PR: kern/100169 Tested by: Kevin Overman <oberman@es.net> * Inverted EAPD for few Nec VersaPro. PR: kern/104715 Submitted by: KAWATA Masahiko <kawata@mta.biglobe.ne.jp> Thanks to various people, notably Joel Dahl, Yuriy Tsibizov, Kevin Oberman, those at #freebsd-azalia @ freenode and others for testing. Joel Dahl will do the manpage update.
2006-11-26 12:24:06 +00:00
chn_setlatency(c, chn_latency);
}
if (flags & CHN_N_TRIGGER) {
Last major commit and updates for RELENG_7: - Rework the entire pcm_channel structure: * Remove rarely used link placeholder, instead, make each pcm_channel as head/link of each own/each other. Unlock - Lock sequence due to sleep malloc has been reduced. * Implement "busy" queue which will contain list of busy/active channels. This greatly reduce locking contention for example while servicing interrupt for hardware with many channels or when virtual channels reach its 256 peak channels. - So I heard you like v chan ... O RLY? Welcome to Virtual **Record** Channels (vrec, rec vchans, vchans for recording, Rec-Chan, you decide), the ultimate solutions for your nagging O_RDWR full-duplex wannabe (note: flash plugins) monopolizing single record channel causing EBUSY. Vrec works exactly like Vchans (or, should I rename it to "Vplay" :) , except that it operates on the opposite direction (recording). Up to 256 vrecs (like vchans) are possible. Notes: * Relocate dev.pcm.%d.{vchans,vchanformat,vchanrate} to each of its respective node/direction: dev.pcm.%d.play.* for "play" (cdev = dsp%d.vp%d) dev.pcm.%d.rec.* for "record" (cdev = dsp%d.vr%d) * Don't expect that it will magically give you ability to split "recording source" (eg: 1 channel for cdrom, 1 channel for mic, etc). Just admit that you only have a *single* recording source / channel. Please bug your hardware vendor instead :) - Bump maxautovchans from 4 to 16. For a full-fledged multimedia desktop/workstation with too many soundservers installed (esound, artsd, jackd, pulse/polypaudio, ding-dong pling plong mudkip fuh fuh, etc), 4 seems inadequate. There will be no memory penalty here, since virtual channels are allocate only by demand. - Nuke/Rework the entire statically created cdev entries. Everything is clonable through snd own clone manager which designed to withstand many kind of abusive devfs droids such as: * while : ; do /bin/test -e /dev/dsp ; done * jot 16777216 0 | while read x ; do ls /dev/dsp0.$x ; done * hundreds (could be thousands) concurrent threads/process opening "/dev/dsp" (previously, this might result EBUSY even with just 3 contesting threads/procs). o Reusable clone objects (instead of creating new one like there's no tomorrow) after certain expiration deadline. The clone allocator will decide whether to reuse, share, or creating new clone. o Automatic garbage collector. - Dynamic unit magic allocator. Maximum attached soundcards can be tuned using tunable "hw.snd.maxunit" (Default to 512). Minimum is 16, and maximum is 2048. - ..other fixes, mostly related to concurrency issues. joel@ will do the manpage updates on sound(4). Have fun.
2007-05-31 18:43:33 +00:00
nrun = CHN_EMPTY(c, children.busy) ? 0 : 1;
if (nrun && !run)
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
err = chn_start(c, 1);
if (!nrun && run)
chn_abort(c);
}
Last (again ?!?) major commit for RELENG_7, featuring total Giant eradication in/from userland path, countless locking fixes, etc. - General sleep call through msleep(9) has been converted to condvar(9) with better consistencies. - Heavily guard every possible "slow path" entries (open(), close(), few ioctl()s, sysctls), but once it entering "fast path" (io, interrupt started), they are free to fly on their own. - Rearrange locking sequences, resulting better concurrency and serialization. Large part doesn't even need locking at all, and will be removed in future. Less clutter, except in few places due to lock ordering. - Anonymous mixer object creation/deletion to simplify mixer handling beyond typical mixer ioctls. Submitted by: chibis (with modifications) - Add few mix_[get|set|..] functions to avoid calling mixer_ioctl() directly using cryptic arguments. - Locking fixes to avoid possible deadlock with (still under Giant) USB. - Better simplex/duplex device handling. - Recover mmap() functionality for recording, which has been lost since 2.2.x - 3.x (the introduction of newpcm). Full-duplex mmap still doesn't work (due to VM/page design), but people still can mmap both by opening each direction separately. mmaped playback is guarantee to work either way. - New sysctl: "hw.snd.compat_linux_mmap" to allow PROT_EXEC page mapping, due to recent changes in linux compatibility layer which require it. All linux applications that using sound + mmap() (mostly games) require this to be enabled. Disabled by default. - Other goodies.. too many, that will increase releng7 shareholder value and make users of releng6 (and below) cry ;) * This commit should be atomic. If anything goes wrong (not counting problem originated from elsewhere), I will not hesitate to revert everything back within 12 hours. This substantial changes itself not a rocket science and the process has begun for almost 2 years, and lots of incremental changes are already in place during that period of time. * Some issues does occur in snd_emu10kx (note the 'x') due to various internal locking issues and it is currently being worked on by chibis. Tested by: chibis (Yuriy Tsibizov), joel, Alexandre Vieira, many innocent souls...
2007-06-16 03:37:28 +00:00
return (err);
}
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
/**
* @brief Fetch array of supported discrete sample rates
*
* Wrapper for CHANNEL_GETRATES. Please see channel_if.m:getrates() for
* detailed information.
*
* @note If the operation isn't supported, this function will just return 0
* (no rates in the array), and *rates will be set to NULL. Callers
* should examine rates @b only if this function returns non-zero.
*
* @param c pcm channel to examine
* @param rates pointer to array of integers; rate table will be recorded here
*
* @return number of rates in the array pointed to be @c rates
*/
int
chn_getrates(struct pcm_channel *c, int **rates)
{
KASSERT(rates != NULL, ("rates is null"));
CHN_LOCKASSERT(c);
return CHANNEL_GETRATES(c->methods, c->devinfo, rates);
}
/**
* @brief Remove channel from a sync group, if there is one.
*
* This function is initially intended for the following conditions:
* - Starting a syncgroup (@c SNDCTL_DSP_SYNCSTART ioctl)
* - Closing a device. (A channel can't be destroyed if it's still in use.)
*
* @note Before calling this function, the syncgroup list mutex must be
* held. (Consider pcm_channel::sm protected by the SG list mutex
* whether @c c is locked or not.)
*
* @param c channel device to be started or closed
* @returns If this channel was the only member of a group, the group ID
* is returned to the caller so that the caller can release it
* via free_unr() after giving up the syncgroup lock. Else it
* returns 0.
*/
int
chn_syncdestroy(struct pcm_channel *c)
{
struct pcmchan_syncmember *sm;
struct pcmchan_syncgroup *sg;
int sg_id;
sg_id = 0;
PCM_SG_LOCKASSERT(MA_OWNED);
if (c->sm != NULL) {
sm = c->sm;
sg = sm->parent;
c->sm = NULL;
KASSERT(sg != NULL, ("syncmember has null parent"));
SLIST_REMOVE(&sg->members, sm, pcmchan_syncmember, link);
free(sm, M_DEVBUF);
if (SLIST_EMPTY(&sg->members)) {
SLIST_REMOVE(&snd_pcm_syncgroups, sg, pcmchan_syncgroup, link);
sg_id = sg->id;
free(sg, M_DEVBUF);
}
}
return sg_id;
}
Change KASSERT() in feed_vchan16() into an explicit test and call to panic() so that the buffer overflow just beyond this point is always caught, even when the code is not compiled with INVARIANTS. Change chn_setblocksize() buffer reallocation code to attempt to avoid the feed_vchan16() buffer overflow by attempting to always keep the bufsoft buffer at least as large as the bufhard buffer. Print a diagnositic message Danger! %s bufsoft size increasing from %d to %d after CHANNEL_SETBLOCKSIZE() if our best attempts fail. If feed_vchan16() were to be called by the interrupt handler while locks are dropped in chn_setblocksize() to increase the size bufsoft to match the size of bufhard, the panic() code in feed_vchan16() will be triggered. If the diagnostic message is printed, it is a warning that a panic is possible if the system were to see events in an "unlucky" order. Change the locking code to avoid the need for MTX_RECURSIVE mutexes. Add the MTX_DUPOK option to the channel mutexes and change the locking sequence to always lock the parent channel before its children to avoid the possibility of deadlock. Actually implement locking assertions for the channel mutexes and fix the problems found by the resulting assertion violations. Clean up the locking code in dsp_ioctl(). Allocate the channel buffers using the malloc() M_WAITOK option instead of M_NOWAIT so that buffer allocation won't fail. Drop locks across the malloc() calls. Add/modify KASSERTS() in attempt to detect problems early. Abuse layering by adding a pointer to the snd_dbuf structure that points back to the pcm_channel that owns it. This allows sndbuf_resize() to do proper locking without having to change the its API, which is used by the hardware drivers. Don't dereference a NULL pointer when setting hw.snd.maxautovchans if a hardware driver is not loaded. Noticed by Ryan Sommers <ryans at gamersimpact.com>. Tested by: Stefan Ehmann <shoesoft AT gmx.net> Tested by: matk (Mathew Kanner) Tested by: Gordon Bergling <gbergling AT 0xfce3.net>
2004-01-28 08:02:15 +00:00
void
chn_lock(struct pcm_channel *c)
{
CHN_LOCK(c);
}
void
chn_unlock(struct pcm_channel *c)
{
CHN_UNLOCK(c);
}
MFp4 the sound Google Summer of Code project: The goal was to sync with the OSSv4 API 4Front Technologies uses in their proprietary OSS driver. This was successful as far as possible. The part of the API which is stable is implemented, for the rest there are some stubs already. New system ioctls: - SNDCTL_SYSINFO - obtain audio system info (version, # of audio/midi/ mixer devices, etc.) - SNDCTL_AUDIOINFO - fetch details about a specific audio device - SNDCTL_MIXERINFO - fetch details about a specific mixer device New audio ioctls: - Sync groups (SNDCTL_DSP_SYNCGROUP/SNDCTL_DSP_SYNCSTART) which allow triggered playback/recording on multiple devices (even across processes simultaneously). - Peak meters (SNDCTL_DSP_GETIPEAKS/SNDCTL_DSP_GETOPEAKS) - can query audio drivers for peak levels (needs driver support, disabled for now). - Per channel playback/recording levels - SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL. Note that these are still in name only, just wrapping around the AC97-style mixer at the moment. The next step is to push them down to the drivers. Audio ioctls still under development by 4Front (for which stubs may exist in this commit): - SNDCTL_GETNAME, SNDCTL_{GET,SET}{SONG,LABEL} - SNDCTL_DSP_{GET,SET}_CHNORDER - SNDCTL_MIX_ENUMINFO, SNDCTL_MIX_EXTINFO - (might be documented enough in the OSS releases to work on this. These ioctls cover the cool "twiddle any knob on your card" features.) Missing: - SNDCTL_DSP_COOKEDMODE -- this ioctl is used to give applications direct access to a card's buffers, bypassing the feeder architecture. It's a toughy -- "someone" needs to decide : (a) if this is desireable, and (b) if it's reasonably feasible. Updates for driver writers: So far, only two routines to the channel class (in channel_if.m) are added. One is for fetching a list of discrete supported playback/recording rates of a channel, and the other is for fetching peak level info (useful for drawing peak meters). Interested parties may want to help pushing down SNDCTL_DSP_{GET,SET}{PLAY,REC}VOL into the drivers. To use the new stuff you need to rebuild the sound drivers or your kernel (depending on if you use modules or not) and to install soundcard.h (a buildworld/installworld handles this). Sponsored by: Google SoC 2006 Submitted by: ryanb Many thanks to: 4Front Technologies for their cooperation, explanations and the nice license of their soundcard.h.
2006-09-23 20:45:47 +00:00
#ifdef OSSV4_EXPERIMENT
int
chn_getpeaks(struct pcm_channel *c, int *lpeak, int *rpeak)
{
CHN_LOCKASSERT(c);
return CHANNEL_GETPEAKS(c->methods, c->devinfo, lpeak, rpeak);
}
#endif