d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
09828ba947
freebsd-dev/sys/dev/sound/pcm/dsp.c

2917 lines
72 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
update my email address.
2003-09-07 16:28:03 +00:00
* Copyright (c) 1999 Cameron Grant <cg@freebsd.org>
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
* 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.
*/
repo-copied to make way for newmidi, this commit updates include paths
1999-11-20 16:50:33 +00:00
#include <dev/sound/pcm/sound.h>
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
#include <sys/ctype.h>
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
SND_DECLARE_FILE("$FreeBSD$");
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
static int dsp_mmap_allow_prot_exec = 0;
SYSCTL_INT(_hw_snd, OID_AUTO, compat_linux_mmap, CTLFLAG_RW,
&dsp_mmap_allow_prot_exec, 0, "linux mmap compatibility");
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 dsp_cdevinfo {
struct pcm_channel *rdch, *wrch;
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 busy, simplex;
TAILQ_ENTRY(dsp_cdevinfo) link;
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
#define PCM_RDCH(x) (((struct dsp_cdevinfo *)(x)->si_drv1)->rdch)
#define PCM_WRCH(x) (((struct dsp_cdevinfo *)(x)->si_drv1)->wrch)
#define PCM_SIMPLEX(x) (((struct dsp_cdevinfo *)(x)->si_drv1)->simplex)
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
#define DSP_CDEVINFO_CACHESIZE 8
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
#define DSP_REGISTERED(x, y) (PCM_REGISTERED(x) && \
(y) != NULL && (y)->si_drv1 != 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
fix the return value of the SOUND_PCM_WRITE_CHANNELS ioctl re-enable old pcm ioctls
2000-05-27 14:40:17 +00:00
#define OLDPCM_IOCTL
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static d_open_t dsp_open;
static d_close_t dsp_close;
static d_read_t dsp_read;
static d_write_t dsp_write;
static d_ioctl_t dsp_ioctl;
static d_poll_t dsp_poll;
static d_mmap_t dsp_mmap;
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +00:00
struct cdevsw dsp_cdevsw = {
Device megapatch 4/6: Introduce d_version field in struct cdevsw, this must always be initialized to D_VERSION. Flip sense of D_NOGIANT flag to D_NEEDGIANT, this involves removing four D_NOGIANT flags and adding 145 D_NEEDGIANT flags.
2004-02-21 21:10:55 +00:00
.d_version = D_VERSION,
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_open = dsp_open,
.d_close = dsp_close,
.d_read = dsp_read,
.d_write = dsp_write,
.d_ioctl = dsp_ioctl,
.d_poll = dsp_poll,
.d_mmap = dsp_mmap,
.d_name = "dsp",
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
};
add defines and ifdefs so this code will compile on 4.x add spls so this code will work on 4.x
2001-06-23 17:36:51 +00:00
#ifdef USING_DEVFS
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
static eventhandler_tag dsp_ehtag = NULL;
static int dsp_umax = -1;
static int dsp_cmax = -1;
add defines and ifdefs so this code will compile on 4.x add spls so this code will work on 4.x
2001-06-23 17:36:51 +00:00
#endif
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
static int dsp_oss_syncgroup(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_syncgroup *group);
static int dsp_oss_syncstart(int sg_id);
static int dsp_oss_policy(struct pcm_channel *wrch, struct pcm_channel *rdch, int policy);
#ifdef OSSV4_EXPERIMENT
static int dsp_oss_cookedmode(struct pcm_channel *wrch, struct pcm_channel *rdch, int enabled);
static int dsp_oss_getchnorder(struct pcm_channel *wrch, struct pcm_channel *rdch, unsigned long long *map);
static int dsp_oss_setchnorder(struct pcm_channel *wrch, struct pcm_channel *rdch, unsigned long long *map);
static int dsp_oss_getlabel(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_label_t *label);
static int dsp_oss_setlabel(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_label_t *label);
static int dsp_oss_getsong(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *song);
static int dsp_oss_setsong(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *song);
static int dsp_oss_setname(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *name);
#endif
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static struct snddev_info *
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_get_info(struct cdev *dev)
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
{
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
return (devclass_get_softc(pcm_devclass, PCMUNIT(dev)));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
static uint32_t
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_get_flags(struct cdev *dev)
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
{
device_t bdev;
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
bdev = devclass_get_device(pcm_devclass, PCMUNIT(dev));
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
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
return ((bdev != NULL) ? pcm_getflags(bdev) : 0xffffffff);
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
}
static void
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
dsp_set_flags(struct cdev *dev, uint32_t flags)
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
{
device_t bdev;
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
bdev = devclass_get_device(pcm_devclass, PCMUNIT(dev));
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
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 (bdev != NULL)
pcm_setflags(bdev, flags);
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
/*
Fix typos and clean up some comments. Approved by: ariff
2006-01-25 21:13:46 +00:00
* return the channels associated with an open device instance.
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
* lock channels specified.
*/
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
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
getchns(struct cdev *dev, struct pcm_channel **rdch, struct pcm_channel **wrch,
uint32_t prio)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
struct snddev_info *d;
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 pcm_channel *ch;
uint32_t flags;
fix dma underrun issues mutate some panics to kasserts add more spl protection PR: kern/14990 Partially Submitted by: Vladimir N.Silyaev <vns@delta.odessa.ua> Reviewed by: dfr
1999-12-05 19:09:13 +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 (PCM_SIMPLEX(dev) != 0) {
d = dsp_get_info(dev);
if (!PCM_REGISTERED(d))
return (ENXIO);
pcm_lock(d);
PCM_WAIT(d);
PCM_ACQUIRE(d);
/*
* Note: order is important -
* pcm flags -> prio query flags -> wild guess
*/
ch = NULL;
flags = dsp_get_flags(dev);
if (flags & SD_F_PRIO_WR) {
ch = PCM_RDCH(dev);
PCM_RDCH(dev) = NULL;
} else if (flags & SD_F_PRIO_RD) {
ch = PCM_WRCH(dev);
PCM_WRCH(dev) = NULL;
} else if (prio & SD_F_PRIO_WR) {
ch = PCM_RDCH(dev);
PCM_RDCH(dev) = NULL;
flags |= SD_F_PRIO_WR;
} else if (prio & SD_F_PRIO_RD) {
ch = PCM_WRCH(dev);
PCM_WRCH(dev) = NULL;
flags |= SD_F_PRIO_RD;
} else if (PCM_WRCH(dev) != NULL) {
ch = PCM_RDCH(dev);
PCM_RDCH(dev) = NULL;
flags |= SD_F_PRIO_WR;
} else if (PCM_RDCH(dev) != NULL) {
ch = PCM_WRCH(dev);
PCM_WRCH(dev) = NULL;
flags |= SD_F_PRIO_RD;
}
PCM_SIMPLEX(dev) = 0;
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
dsp_set_flags(dev, 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
if (ch != NULL) {
CHN_LOCK(ch);
pcm_chnref(ch, -1);
pcm_chnrelease(ch);
}
PCM_RELEASE(d);
pcm_unlock(d);
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
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
*rdch = PCM_RDCH(dev);
*wrch = PCM_WRCH(dev);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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 (*rdch != NULL && (prio & SD_F_PRIO_RD))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
CHN_LOCK(*rdch);
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 (*wrch != NULL && (prio & SD_F_PRIO_WR))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
CHN_LOCK(*wrch);
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);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
/* unlock specified channels */
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
static void
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
relchns(struct cdev *dev, struct pcm_channel *rdch, struct pcm_channel *wrch,
uint32_t prio)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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 (wrch != NULL && (prio & SD_F_PRIO_WR))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
CHN_UNLOCK(wrch);
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 (rdch != NULL && (prio & SD_F_PRIO_RD))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
CHN_UNLOCK(rdch);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
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
static void
dsp_cdevinfo_alloc(struct cdev *dev,
struct pcm_channel *rdch, struct pcm_channel *wrch)
{
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 snddev_info *d;
struct dsp_cdevinfo *cdi;
int simplex;
d = dsp_get_info(dev);
KASSERT(PCM_REGISTERED(d) && dev != NULL && dev->si_drv1 == NULL &&
rdch != wrch,
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
("bogus %s(), what are you trying to accomplish here?", __func__));
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
PCM_BUSYASSERT(d);
mtx_assert(d->lock, MA_OWNED);
simplex = (dsp_get_flags(dev) & SD_F_SIMPLEX) ? 1 : 0;
/*
* Scan for free instance entry and put it into the end of list.
* Create new one if necessary.
*/
TAILQ_FOREACH(cdi, &d->dsp_cdevinfo_pool, link) {
if (cdi->busy != 0)
break;
cdi->rdch = rdch;
cdi->wrch = wrch;
cdi->simplex = simplex;
cdi->busy = 1;
TAILQ_REMOVE(&d->dsp_cdevinfo_pool, cdi, link);
TAILQ_INSERT_TAIL(&d->dsp_cdevinfo_pool, cdi, link);
dev->si_drv1 = cdi;
return;
}
pcm_unlock(d);
cdi = malloc(sizeof(*cdi), M_DEVBUF, M_WAITOK | M_ZERO);
pcm_lock(d);
cdi->rdch = rdch;
cdi->wrch = wrch;
cdi->simplex = simplex;
cdi->busy = 1;
TAILQ_INSERT_TAIL(&d->dsp_cdevinfo_pool, cdi, link);
dev->si_drv1 = cdi;
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
}
static void
dsp_cdevinfo_free(struct cdev *dev)
{
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 snddev_info *d;
struct dsp_cdevinfo *cdi, *tmp;
uint32_t flags;
int i;
d = dsp_get_info(dev);
KASSERT(PCM_REGISTERED(d) && dev != NULL && dev->si_drv1 != 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
PCM_RDCH(dev) == NULL && PCM_WRCH(dev) == NULL,
("bogus %s(), what are you trying to accomplish here?", __func__));
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
PCM_BUSYASSERT(d);
mtx_assert(d->lock, MA_OWNED);
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
cdi = dev->si_drv1;
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
dev->si_drv1 = 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
cdi->rdch = NULL;
cdi->wrch = NULL;
cdi->simplex = 0;
cdi->busy = 0;
/*
* Once it is free, move it back to the beginning of list for
* faster new entry allocation.
*/
TAILQ_REMOVE(&d->dsp_cdevinfo_pool, cdi, link);
TAILQ_INSERT_HEAD(&d->dsp_cdevinfo_pool, cdi, link);
/*
* Scan the list, cache free entries up to DSP_CDEVINFO_CACHESIZE.
* Reset simplex flags.
*/
flags = dsp_get_flags(dev) & ~SD_F_PRIO_SET;
i = DSP_CDEVINFO_CACHESIZE;
TAILQ_FOREACH_SAFE(cdi, &d->dsp_cdevinfo_pool, link, tmp) {
if (cdi->busy != 0) {
if (cdi->simplex == 0) {
if (cdi->rdch != NULL)
flags |= SD_F_PRIO_RD;
if (cdi->wrch != NULL)
flags |= SD_F_PRIO_WR;
}
} else {
if (i == 0) {
TAILQ_REMOVE(&d->dsp_cdevinfo_pool, cdi, link);
free(cdi, M_DEVBUF);
} else
i--;
}
}
dsp_set_flags(dev, flags);
}
void
dsp_cdevinfo_init(struct snddev_info *d)
{
struct dsp_cdevinfo *cdi;
int i;
KASSERT(d != NULL, ("NULL snddev_info"));
PCM_BUSYASSERT(d);
mtx_assert(d->lock, MA_NOTOWNED);
TAILQ_INIT(&d->dsp_cdevinfo_pool);
for (i = 0; i < DSP_CDEVINFO_CACHESIZE; i++) {
cdi = malloc(sizeof(*cdi), M_DEVBUF, M_WAITOK | M_ZERO);
TAILQ_INSERT_HEAD(&d->dsp_cdevinfo_pool, cdi, link);
}
}
void
dsp_cdevinfo_flush(struct snddev_info *d)
{
struct dsp_cdevinfo *cdi, *tmp;
KASSERT(d != NULL, ("NULL snddev_info"));
PCM_BUSYASSERT(d);
mtx_assert(d->lock, MA_NOTOWNED);
cdi = TAILQ_FIRST(&d->dsp_cdevinfo_pool);
while (cdi != NULL) {
tmp = TAILQ_NEXT(cdi, link);
free(cdi, M_DEVBUF);
cdi = tmp;
}
TAILQ_INIT(&d->dsp_cdevinfo_pool);
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
}
/* duplex / simplex cdev type */
enum {
DSP_CDEV_TYPE_RDONLY, /* simplex read-only (record) */
DSP_CDEV_TYPE_WRONLY, /* simplex write-only (play) */
DSP_CDEV_TYPE_RDWR, /* duplex read, write, or both */
};
#define DSP_F_VALID(x) ((x) & (FREAD | FWRITE))
#define DSP_F_DUPLEX(x) (((x) & (FREAD | FWRITE)) == (FREAD | FWRITE))
#define DSP_F_SIMPLEX(x) (!DSP_F_DUPLEX(x))
#define DSP_F_READ(x) ((x) & FREAD)
#define DSP_F_WRITE(x) ((x) & FWRITE)
static const struct {
int type;
char *name;
char *sep;
int use_sep;
int hw;
int max;
uint32_t fmt, spd;
int query;
} dsp_cdevs[] = {
{ SND_DEV_DSP, "dsp", ".", 0, 0, 0,
AFMT_U8, DSP_DEFAULT_SPEED, DSP_CDEV_TYPE_RDWR },
{ SND_DEV_AUDIO, "audio", ".", 0, 0, 0,
AFMT_MU_LAW, DSP_DEFAULT_SPEED, DSP_CDEV_TYPE_RDWR },
{ SND_DEV_DSP16, "dspW", ".", 0, 0, 0,
AFMT_S16_LE, DSP_DEFAULT_SPEED, DSP_CDEV_TYPE_RDWR },
{ SND_DEV_DSPHW_PLAY, "dsp", ".p", 1, 1, SND_MAXHWCHAN,
AFMT_S16_LE | AFMT_STEREO, 48000, DSP_CDEV_TYPE_WRONLY },
{ SND_DEV_DSPHW_VPLAY, "dsp", ".vp", 1, 1, SND_MAXVCHANS,
AFMT_S16_LE | AFMT_STEREO, 48000, DSP_CDEV_TYPE_WRONLY },
{ SND_DEV_DSPHW_REC, "dsp", ".r", 1, 1, SND_MAXHWCHAN,
AFMT_S16_LE | AFMT_STEREO, 48000, DSP_CDEV_TYPE_RDONLY },
{ SND_DEV_DSPHW_VREC, "dsp", ".vr", 1, 1, SND_MAXVCHANS,
AFMT_S16_LE | AFMT_STEREO, 48000, DSP_CDEV_TYPE_RDONLY },
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
{ SND_DEV_DSPHW_CD, "dspcd", ".", 0, 0, 0,
AFMT_S16_LE | AFMT_STEREO, 44100, DSP_CDEV_TYPE_RDWR },
{ SND_DEV_DSP_MMAP, "dsp_mmap", ".", 0, 0, 0,
AFMT_S16_LE | AFMT_STEREO, 48000, DSP_CDEV_TYPE_RDWR },
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
#define DSP_FIXUP_ERROR() do { \
prio = dsp_get_flags(i_dev); \
if (!DSP_F_VALID(flags)) \
error = EINVAL; \
if (!DSP_F_DUPLEX(flags) && \
((DSP_F_READ(flags) && d->reccount == 0) || \
(DSP_F_WRITE(flags) && d->playcount == 0))) \
error = ENOTSUP; \
else if (!DSP_F_DUPLEX(flags) && (prio & SD_F_SIMPLEX) && \
((DSP_F_READ(flags) && (prio & SD_F_PRIO_WR)) || \
(DSP_F_WRITE(flags) && (prio & SD_F_PRIO_RD)))) \
error = EBUSY; \
else if (DSP_REGISTERED(d, i_dev)) \
error = EBUSY; \
} while(0)
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_open(struct cdev *i_dev, int flags, int mode, struct thread *td)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct pcm_channel *rdch, *wrch;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
struct snddev_info *d;
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
uint32_t fmt, spd, prio;
int i, error, rderror, wrerror, devtype, wdevunit, rdevunit;
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
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
/* Kind of impossible.. */
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
if (i_dev == NULL || td == 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
return (ENODEV);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
d = dsp_get_info(i_dev);
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 (!PCM_REGISTERED(d))
return (EBADF);
PCM_GIANT_ENTER(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
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
/* Lock snddev so nobody else can monkey with it. */
pcm_lock(d);
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
PCM_WAIT(d);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
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
/*
* Try to acquire cloned device before someone else pick it.
* ENODEV means this is not a cloned droids.
*/
error = snd_clone_acquire(i_dev);
if (!(error == 0 || error == ENODEV)) {
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
DSP_FIXUP_ERROR();
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
pcm_unlock(d);
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
PCM_GIANT_EXIT(d);
return (error);
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
}
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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
error = 0;
DSP_FIXUP_ERROR();
add a method for recording of specific channels for devices with more than one hardware record channel. new devices, /dev/dsprX.Y where X is unit number and Y is channel index.
2001-09-05 16:28:41 +00:00
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 (error != 0) {
(void)snd_clone_release(i_dev);
pcm_unlock(d);
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
PCM_GIANT_EXIT(d);
return (error);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
}
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
* That is just enough. Acquire and unlock pcm lock so
* the other will just have to wait until we finish doing
* everything.
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
PCM_ACQUIRE(d);
pcm_unlock(d);
Marginally simplify dsp_open error handling by adding an early test to determine liklihood of opening device in requested directions. Makes for simpler error handling and change should close kern/35004. PR: kern/35004.
2002-08-28 15:19:30 +00:00
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
devtype = PCMDEV(i_dev);
wdevunit = -1;
rdevunit = -1;
fmt = 0;
spd = 0;
Marginally simplify dsp_open error handling by adding an early test to determine liklihood of opening device in requested directions. Makes for simpler error handling and change should close kern/35004. PR: kern/35004.
2002-08-28 15:19:30 +00:00
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
for (i = 0; i < (sizeof(dsp_cdevs) / sizeof(dsp_cdevs[0])); i++) {
if (devtype != dsp_cdevs[i].type)
continue;
if (DSP_F_SIMPLEX(flags) &&
((dsp_cdevs[i].query == DSP_CDEV_TYPE_WRONLY &&
DSP_F_READ(flags)) ||
(dsp_cdevs[i].query == DSP_CDEV_TYPE_RDONLY &&
DSP_F_WRITE(flags)))) {
/*
* simplex, opposite direction? Please be gone..
*/
(void)snd_clone_release(i_dev);
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
PCM_RELEASE_QUICK(d);
PCM_GIANT_EXIT(d);
return (ENOTSUP);
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 (dsp_cdevs[i].query == DSP_CDEV_TYPE_WRONLY)
wdevunit = dev2unit(i_dev);
else if (dsp_cdevs[i].query == DSP_CDEV_TYPE_RDONLY)
rdevunit = dev2unit(i_dev);
fmt = dsp_cdevs[i].fmt;
spd = dsp_cdevs[i].spd;
break;
Marginally simplify dsp_open error handling by adding an early test to determine liklihood of opening device in requested directions. Makes for simpler error handling and change should close kern/35004. PR: kern/35004.
2002-08-28 15:19:30 +00:00
}
beginnings of virtual playback channel support instead of using two malloced arrays for storing channel lists, use an slist. convert the sndstat device to use sbufs and optionally provide more detail about channel state. vchans are software mixed playback channels. they are not enabled by this commit. they use the feeder infrastructure to emulate normal playback channels in a manner transparent to applications, whilst providing as many channels are desired, especially suitable for devices with only one hardware playback channel. in the future they will provide additional features. those wishing to test this functionality will need to add vchan.c to sys/conf/files and use 'sysctl -w hw.snd.pcm0.vchans' to enable it. blocksize and auto-rate selection are not yet supported.
2001-05-27 17:22:00 +00:00
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
/* No matching devtype? */
if (fmt == 0 || spd == 0)
panic("impossible devtype %d", devtype);
rdch = NULL;
wrch = 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
rderror = 0;
wrerror = 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
modify comments
2003-08-15 02:31:13 +00:00
/*
* if we get here, the open request is valid- either:
* * we were previously not open
* * we were open for play xor record and the opener wants
* the non-open direction
*/
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 (DSP_F_READ(flags)) {
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
/* open for read */
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
rderror = pcm_chnalloc(d, &rdch, PCMDIR_REC,
td->td_proc->p_pid, rdevunit);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +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 (rderror == 0 && (chn_reset(rdch, fmt) != 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
(chn_setspeed(rdch, spd) != 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
rderror = ENXIO;
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +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 (rderror != 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
if (rdch != NULL)
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
pcm_chnrelease(rdch);
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 (!DSP_F_DUPLEX(flags)) {
(void)snd_clone_release(i_dev);
PCM_RELEASE_QUICK(d);
PCM_GIANT_EXIT(d);
return (rderror);
}
rdch = NULL;
} else {
if (flags & O_NONBLOCK)
rdch->flags |= CHN_F_NBIO;
pcm_chnref(rdch, 1);
CHN_UNLOCK(rdch);
Marginally simplify dsp_open error handling by adding an early test to determine liklihood of opening device in requested directions. Makes for simpler error handling and change should close kern/35004. PR: kern/35004.
2002-08-28 15:19:30 +00:00
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
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 (DSP_F_WRITE(flags)) {
/* open for write */
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
wrerror = pcm_chnalloc(d, &wrch, PCMDIR_PLAY,
td->td_proc->p_pid, wdevunit);
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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 (wrerror == 0 && (chn_reset(wrch, fmt) != 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
(chn_setspeed(wrch, spd) != 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
wrerror = ENXIO;
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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 (wrerror != 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
if (wrch != NULL)
pcm_chnrelease(wrch);
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 (!DSP_F_DUPLEX(flags)) {
if (rdch != NULL) {
/*
* Lock, deref and release previously
* created record channel
*/
CHN_LOCK(rdch);
pcm_chnref(rdch, -1);
pcm_chnrelease(rdch);
}
(void)snd_clone_release(i_dev);
PCM_RELEASE_QUICK(d);
PCM_GIANT_EXIT(d);
return (wrerror);
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
wrch = NULL;
} else {
if (flags & O_NONBLOCK)
wrch->flags |= CHN_F_NBIO;
pcm_chnref(wrch, 1);
CHN_UNLOCK(wrch);
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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
}
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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 (rdch == NULL && wrch == NULL) {
(void)snd_clone_release(i_dev);
PCM_RELEASE_QUICK(d);
PCM_GIANT_EXIT(d);
return ((wrerror != 0) ? wrerror : rderror);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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
pcm_lock(d);
/*
* We're done. Allocate channels information for this cdev.
*/
dsp_cdevinfo_alloc(i_dev, rdch, wrch);
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
/*
* Increase clone refcount for its automatic garbage collector.
*/
(void)snd_clone_ref(i_dev);
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +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
PCM_RELEASE(d);
Fix lock order reversal when opening device and chn_reset fails. Submitted by: Jan-Espen Pettersen <sigsegv@leakingmemory.org> Tested by: Georg Funk <georgfunk@web.de>
2003-05-01 16:31:21 +00:00
pcm_unlock(d);
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
PCM_GIANT_LEAVE(d);
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);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_close(struct cdev *i_dev, int flags, int mode, struct thread *td)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct pcm_channel *rdch, *wrch;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
struct snddev_info *d;
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 sg_ids, refs;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
d = dsp_get_info(i_dev);
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 (!DSP_REGISTERED(d, i_dev))
return (EBADF);
PCM_GIANT_ENTER(d);
Fix LOR #174. Tested with: WITNESS, INVARIANTS and DIAGNOSTICS
2005-12-31 01:58:40 +00:00
pcm_lock(d);
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
PCM_WAIT(d);
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
rdch = PCM_RDCH(i_dev);
wrch = PCM_WRCH(i_dev);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
if (rdch || wrch) {
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
PCM_ACQUIRE(d);
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
pcm_unlock(d);
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
refs = 0;
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
if (rdch) {
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
/*
* The channel itself need not be locked because:
* a) Adding a channel to a syncgroup happens only in dsp_ioctl(),
* which cannot run concurrently to dsp_close().
* b) The syncmember pointer (sm) is protected by the global
* syncgroup list lock.
* c) A channel can't just disappear, invalidating pointers,
* unless it's closed/dereferenced first.
*/
PCM_SG_LOCK();
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
sg_ids = chn_syncdestroy(rdch);
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
PCM_SG_UNLOCK();
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 (sg_ids != 0)
free_unr(pcmsg_unrhdr, sg_ids);
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
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
CHN_LOCK(rdch);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
refs += pcm_chnref(rdch, -1);
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
chn_abort(rdch); /* won't sleep */
rdch->flags &= ~(CHN_F_RUNNING | CHN_F_MAPPED | CHN_F_DEAD);
chn_reset(rdch, 0);
pcm_chnrelease(rdch);
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
PCM_RDCH(i_dev) = NULL;
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
}
if (wrch) {
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
/*
* Please see block above.
*/
PCM_SG_LOCK();
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
sg_ids = chn_syncdestroy(wrch);
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
PCM_SG_UNLOCK();
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 (sg_ids != 0)
free_unr(pcmsg_unrhdr, sg_ids);
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
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
CHN_LOCK(wrch);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
refs += pcm_chnref(wrch, -1);
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
chn_flush(wrch); /* may sleep */
wrch->flags &= ~(CHN_F_RUNNING | CHN_F_MAPPED | CHN_F_DEAD);
chn_reset(wrch, 0);
pcm_chnrelease(wrch);
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
PCM_WRCH(i_dev) = NULL;
Fix sound LOR problems: dsp_open: rearrange to only hold one lock at a time dsp_close: ditto mixer_hwvol_init: delete locking, the only consumer seems to be the ess driver and it only call it a creation time, I think the device will be stable across the sleepable malloc. cmi interrupt routine: Release locks while caller chn_intr, either this or do what emu10k1 does which is have no locks at in the interrupt handler. Submitted by: mat@cnd.mcgill.ca
2003-11-11 05:38:28 +00:00
}
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
pcm_lock(d);
/*
* If there are no more references, release the channels.
*/
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 (refs == 0 && PCM_RDCH(i_dev) == NULL &&
PCM_WRCH(i_dev) == NULL) {
dsp_cdevinfo_free(i_dev);
/*
* Release clone busy state and unref it
* so the automatic garbage collector will
* get the hint and do the remaining cleanup
* process.
*/
(void)snd_clone_release(i_dev);
(void)snd_clone_unref(i_dev);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +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
PCM_RELEASE(d);
General spl* cleanup. It doesn't serve any purpose anymore. Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-10-02 15:39:07 +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
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
pcm_unlock(d);
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
PCM_GIANT_LEAVE(d);
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
return (0);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
static __inline int
dsp_io_ops(struct cdev *i_dev, struct uio *buf)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
struct snddev_info *d;
struct pcm_channel **ch, *rdch, *wrch;
int (*chn_io)(struct pcm_channel *, struct uio *);
int prio, ret;
pid_t runpid;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
KASSERT(i_dev != NULL && buf != NULL &&
(buf->uio_rw == UIO_READ || buf->uio_rw == UIO_WRITE),
("%s(): io train wreck!", __func__));
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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
d = dsp_get_info(i_dev);
if (!DSP_REGISTERED(d, i_dev))
return (EBADF);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
PCM_GIANT_ENTER(d);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
switch (buf->uio_rw) {
case UIO_READ:
prio = SD_F_PRIO_RD;
ch = &rdch;
chn_io = chn_read;
break;
case UIO_WRITE:
prio = SD_F_PRIO_WR;
ch = &wrch;
chn_io = chn_write;
break;
default:
panic("invalid/corrupted uio direction: %d", buf->uio_rw);
break;
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
rdch = NULL;
wrch = NULL;
runpid = buf->uio_td->td_proc->p_pid;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
getchns(i_dev, &rdch, &wrch, prio);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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 (*ch == NULL || !((*ch)->flags & CHN_F_BUSY)) {
PCM_GIANT_EXIT(d);
return (EBADF);
}
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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 (((*ch)->flags & (CHN_F_MAPPED | CHN_F_DEAD)) ||
(((*ch)->flags & CHN_F_RUNNING) && (*ch)->pid != runpid)) {
relchns(i_dev, rdch, wrch, prio);
PCM_GIANT_EXIT(d);
return (EINVAL);
} else if (!((*ch)->flags & CHN_F_RUNNING)) {
(*ch)->flags |= CHN_F_RUNNING;
(*ch)->pid = runpid;
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
/*
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_read/write must give up channel lock in order to copy bytes
* from/to userland, so up the "in progress" counter to make sure
* someone else doesn't come along and muss up the buffer.
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
++(*ch)->inprog;
ret = chn_io(*ch, buf);
--(*ch)->inprog;
CHN_BROADCAST(&(*ch)->cv);
relchns(i_dev, rdch, wrch, prio);
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
PCM_GIANT_LEAVE(d);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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);
}
static int
dsp_read(struct cdev *i_dev, struct uio *buf, int flag)
{
return (dsp_io_ops(i_dev, buf));
}
static int
dsp_write(struct cdev *i_dev, struct uio *buf, int flag)
{
return (dsp_io_ops(i_dev, buf));
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_ioctl(struct cdev *i_dev, u_long cmd, caddr_t arg, int mode, struct thread *td)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
struct pcm_channel *chn, *rdch, *wrch;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
struct snddev_info *d;
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 *arg_i, ret, kill, tmp, xcmd;
d = dsp_get_info(i_dev);
if (!DSP_REGISTERED(d, i_dev))
return (EBADF);
PCM_GIANT_ENTER(d);
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
arg_i = (int *)arg;
ret = 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
xcmd = 0;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
/*
* this is an evil hack to allow broken apps to perform mixer ioctls
* on dsp devices.
*/
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
if (IOCGROUP(cmd) == 'M') {
/*
* This is at least, a bug to bug compatible with OSS.
*/
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 (d->mixer_dev != NULL) {
PCM_ACQUIRE_QUICK(d);
ret = mixer_ioctl_cmd(d->mixer_dev, cmd, arg, -1, td,
MIXER_CMD_DIRECT);
PCM_RELEASE_QUICK(d);
} else
ret = EBADF;
PCM_GIANT_EXIT(d);
return (ret);
Apply more thorough fixes while dealing with device opening and closing: - Determine open direction using 'flags', not 'mode'. This bug exist since past 4 years. - Don't allow opening the same device twice, be it in a same or different direction. - O_RDWR is allowed, provided that it is done by a single open (for example by mixer(8)) and the underlying hardware support true full-duplex operation. - Do various paranoid checking in case other process/thread trying to hijack the same device twice (or more). MFC after: 5 days
2006-03-21 06:35:48 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
/*
* Certain ioctls may be made on any type of device (audio, mixer,
* and MIDI). Handle those special cases here.
*/
if (IOCGROUP(cmd) == 'X') {
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
PCM_ACQUIRE_QUICK(d);
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
switch(cmd) {
case SNDCTL_SYSINFO:
sound_oss_sysinfo((oss_sysinfo *)arg);
break;
case SNDCTL_AUDIOINFO:
ret = dsp_oss_audioinfo(i_dev, (oss_audioinfo *)arg);
break;
case SNDCTL_MIXERINFO:
ret = mixer_oss_mixerinfo(i_dev, (oss_mixerinfo *)arg);
break;
default:
ret = EINVAL;
}
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
PCM_RELEASE_QUICK(d);
PCM_GIANT_EXIT(d);
return (ret);
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
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
getchns(i_dev, &rdch, &wrch, 0);
kill = 0;
Remove paranoid thread hijacking check. It cause wierd behaviour (dangling channel - process association) especially in threaded or fork()ed apps.
2006-03-30 06:17:03 +00:00
if (wrch && (wrch->flags & CHN_F_DEAD))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
kill |= 1;
Remove paranoid thread hijacking check. It cause wierd behaviour (dangling channel - process association) especially in threaded or fork()ed apps.
2006-03-30 06:17:03 +00:00
if (rdch && (rdch->flags & CHN_F_DEAD))
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
kill |= 2;
if (kill == 3) {
relchns(i_dev, rdch, wrch, 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
PCM_GIANT_EXIT(d);
return (EINVAL);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (kill & 1)
wrch = NULL;
if (kill & 2)
rdch = 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
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 (wrch == NULL && rdch == NULL) {
relchns(i_dev, rdch, wrch, 0);
PCM_GIANT_EXIT(d);
return (EINVAL);
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
switch(cmd) {
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
#ifdef OLDPCM_IOCTL
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
/*
* we start with the new ioctl interface.
*/
case AIONWRITE: /* how many bytes can write ? */
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (wrch) {
CHN_LOCK(wrch);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
/*
if (wrch && wrch->bufhard.dl)
while (chn_wrfeed(wrch) == 0);
*/
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
*arg_i = sndbuf_getfree(wrch->bufsoft);
CHN_UNLOCK(wrch);
} else {
*arg_i = 0;
ret = EINVAL;
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case AIOSSIZE: /* set the current blocksize */
{
struct snd_size *p = (struct snd_size *)arg;
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
p->play_size = 0;
p->rec_size = 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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
fix the return value of the SOUND_PCM_WRITE_CHANNELS ioctl re-enable old pcm ioctls
2000-05-27 14:40:17 +00:00
chn_setblocksize(wrch, 2, p->play_size);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
p->play_size = sndbuf_getblksz(wrch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (rdch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
fix the return value of the SOUND_PCM_WRITE_CHANNELS ioctl re-enable old pcm ioctls
2000-05-27 14:40:17 +00:00
chn_setblocksize(rdch, 2, p->rec_size);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
p->rec_size = sndbuf_getblksz(rdch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
break;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case AIOGSIZE: /* get the current blocksize */
{
struct snd_size *p = (struct snd_size *)arg;
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (wrch) {
CHN_LOCK(wrch);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
p->play_size = sndbuf_getblksz(wrch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
}
if (rdch) {
CHN_LOCK(rdch);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
p->rec_size = sndbuf_getblksz(rdch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case AIOSFMT:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
case AIOGFMT:
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
snd_chan_param *p = (snd_chan_param *)arg;
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
Add some safeguards to AIOSFMT: - Return EINVAL if play_format or rec_format is set but the corresponding sample rate is 0. - Don't try to set the playback or recording format to 0. Previously, issuing an AIOSFMT ioctl with an all-zeroes snd_chan_param would trigger a KASSERT in chn_fmtchain(); I'm unsure about the effects on a kernel without INVARIANTS. After this commit, issuing AIOSFMT with an all-zeroes snd_chan_param is equivalent to issuing AIOGFMT. MFC after: 2 weeks
2005-10-30 10:03:11 +00:00
if (cmd == AIOSFMT &&
((p->play_format != 0 && p->play_rate == 0) ||
(p->rec_format != 0 && p->rec_rate == 0))) {
ret = EINVAL;
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
PCM_ACQUIRE_QUICK(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
Add some safeguards to AIOSFMT: - Return EINVAL if play_format or rec_format is set but the corresponding sample rate is 0. - Don't try to set the playback or recording format to 0. Previously, issuing an AIOSFMT ioctl with an all-zeroes snd_chan_param would trigger a KASSERT in chn_fmtchain(); I'm unsure about the effects on a kernel without INVARIANTS. After this commit, issuing AIOSFMT with an all-zeroes snd_chan_param is equivalent to issuing AIOGFMT. MFC after: 2 weeks
2005-10-30 10:03:11 +00:00
if (cmd == AIOSFMT && p->play_format != 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn_setformat(wrch, p->play_format);
chn_setspeed(wrch, p->play_rate);
}
p->play_rate = wrch->speed;
p->play_format = wrch->format;
CHN_UNLOCK(wrch);
} else {
p->play_rate = 0;
p->play_format = 0;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
if (rdch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
Add some safeguards to AIOSFMT: - Return EINVAL if play_format or rec_format is set but the corresponding sample rate is 0. - Don't try to set the playback or recording format to 0. Previously, issuing an AIOSFMT ioctl with an all-zeroes snd_chan_param would trigger a KASSERT in chn_fmtchain(); I'm unsure about the effects on a kernel without INVARIANTS. After this commit, issuing AIOSFMT with an all-zeroes snd_chan_param is equivalent to issuing AIOGFMT. MFC after: 2 weeks
2005-10-30 10:03:11 +00:00
if (cmd == AIOSFMT && p->rec_format != 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn_setformat(rdch, p->rec_format);
chn_setspeed(rdch, p->rec_rate);
}
p->rec_rate = rdch->speed;
p->rec_format = rdch->format;
CHN_UNLOCK(rdch);
} else {
p->rec_rate = 0;
p->rec_format = 0;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
PCM_RELEASE_QUICK(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case AIOGCAP: /* get capabilities */
{
snd_capabilities *p = (snd_capabilities *)arg;
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct pcmchan_caps *pcaps = NULL, *rcaps = NULL;
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
struct cdev *pdev;
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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
pcm_lock(d);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (rdch) {
CHN_LOCK(rdch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
rcaps = chn_getcaps(rdch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
}
if (wrch) {
CHN_LOCK(wrch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
pcaps = chn_getcaps(wrch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
p->rate_min = max(rcaps? rcaps->minspeed : 0,
pcaps? pcaps->minspeed : 0);
p->rate_max = min(rcaps? rcaps->maxspeed : 1000000,
pcaps? pcaps->maxspeed : 1000000);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
p->bufsize = min(rdch? sndbuf_getsize(rdch->bufsoft) : 1000000,
wrch? sndbuf_getsize(wrch->bufsoft) : 1000000);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
/* XXX bad on sb16 */
rework feeder sytem to allow feeders in klds modify driver capability reporting format to list every audio format seperately- required for above and because we could not previously indicate that mono was unsupported. there should be no functional impact.
2000-08-20 22:18:56 +00:00
p->formats = (rdch? chn_getformats(rdch) : 0xffffffff) &
(wrch? chn_getformats(wrch) : 0xffffffff);
handle duplex properly in the AIOGCAP ioctl - this may (partially?) fix rat
2000-06-07 11:51:30 +00:00
if (rdch && wrch)
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
p->formats |= (dsp_get_flags(i_dev) & SD_F_SIMPLEX)? 0 : AFMT_FULLDUPLEX;
As previously announced: discontinue use of makedev() call in soundcode. This takes us a lot closer to refcounting dev_t. This patch originally by cg@ with a few minor changes by me. It is largely untested, but has been HEADSUP'ed twice, so presumably people have not found any issues with it. Submitted by: cg@
2004-01-17 10:37:11 +00:00
pdev = d->mixer_dev;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
p->mixers = 1; /* default: one mixer */
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
p->inputs = pdev->si_drv1? mix_getdevs(pdev->si_drv1) : 0;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
p->left = p->right = 100;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (wrch)
CHN_UNLOCK(wrch);
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 (rdch)
CHN_UNLOCK(rdch);
pcm_unlock(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case AIOSTOP:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (*arg_i == AIOSYNC_PLAY && wrch) {
CHN_LOCK(wrch);
update code dealing with snd_dbuf objects to do so using a functional interface modify chn_setblocksize() to pick a default soft-blocksize appropriate to the sample rate and format in use. it will aim for a power of two size small enough to generate block sizes of at most 20ms. it will also set the hard-blocksize taking into account rate/format conversions in use. update drivers to implement setblocksize correctly: updated, tested: sb16, emu10k1, maestro, solo updated, untested: ad1816, ess, mss, sb8, csa not updated: ds1, es137x, fm801, neomagic, t4dwave, via82c686 i lack hardware to test: ad1816, csa, fm801, neomagic others will be updated/tested in the next few days.
2000-12-23 03:16:13 +00:00
*arg_i = chn_abort(wrch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
} else if (*arg_i == AIOSYNC_CAPTURE && rdch) {
CHN_LOCK(rdch);
update code dealing with snd_dbuf objects to do so using a functional interface modify chn_setblocksize() to pick a default soft-blocksize appropriate to the sample rate and format in use. it will aim for a power of two size small enough to generate block sizes of at most 20ms. it will also set the hard-blocksize taking into account rate/format conversions in use. update drivers to implement setblocksize correctly: updated, tested: sb16, emu10k1, maestro, solo updated, untested: ad1816, ess, mss, sb8, csa not updated: ds1, es137x, fm801, neomagic, t4dwave, via82c686 i lack hardware to test: ad1816, csa, fm801, neomagic others will be updated/tested in the next few days.
2000-12-23 03:16:13 +00:00
*arg_i = chn_abort(rdch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
} else {
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
printf("AIOSTOP: bad channel 0x%x\n", *arg_i);
*arg_i = 0;
}
break;
case AIOSYNC:
printf("AIOSYNC chan 0x%03lx pos %lu unimplemented\n",
((snd_sync_parm *)arg)->chan, ((snd_sync_parm *)arg)->pos);
break;
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
#endif
Fix the "dsp sync" debugging printf that was unconditional. Fix a few formatting nits/glitches/whatever.
1999-09-19 09:53:57 +00:00
/*
* here follow the standard ioctls (filio.h etc.)
*/
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case FIONREAD: /* get # bytes to read */
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (rdch) {
CHN_LOCK(rdch);
/* if (rdch && rdch->bufhard.dl)
while (chn_rdfeed(rdch) == 0);
*/
*arg_i = sndbuf_getready(rdch->bufsoft);
CHN_UNLOCK(rdch);
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
} else {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
*arg_i = 0;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
ret = EINVAL;
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case FIOASYNC: /*set/clear async i/o */
DEB( printf("FIOASYNC\n") ; )
break;
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
The SNDCTL_DSP_NONBLOCK ioctl doesn't take an argument, from what I can tell by browsing the Internet (Linux, OSS, etc.).
2006-09-29 09:10:54 +00:00
case SNDCTL_DSP_NONBLOCK: /* set non-blocking i/o */
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case FIONBIO: /* set/clear non-blocking i/o */
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (rdch) {
CHN_LOCK(rdch);
The SNDCTL_DSP_NONBLOCK ioctl doesn't take an argument, from what I can tell by browsing the Internet (Linux, OSS, etc.).
2006-09-29 09:10:54 +00:00
if (cmd == SNDCTL_DSP_NONBLOCK || *arg_i)
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
rdch->flags |= CHN_F_NBIO;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
else
rdch->flags &= ~CHN_F_NBIO;
CHN_UNLOCK(rdch);
}
if (wrch) {
CHN_LOCK(wrch);
The SNDCTL_DSP_NONBLOCK ioctl doesn't take an argument, from what I can tell by browsing the Internet (Linux, OSS, etc.).
2006-09-29 09:10:54 +00:00
if (cmd == SNDCTL_DSP_NONBLOCK || *arg_i)
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
wrch->flags |= CHN_F_NBIO;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
else
wrch->flags &= ~CHN_F_NBIO;
CHN_UNLOCK(wrch);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
/*
Fix the "dsp sync" debugging printf that was unconditional. Fix a few formatting nits/glitches/whatever.
1999-09-19 09:53:57 +00:00
* Finally, here is the linux-compatible ioctl interface
*/
#define THE_REAL_SNDCTL_DSP_GETBLKSIZE _IOWR('P', 4, int)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case THE_REAL_SNDCTL_DSP_GETBLKSIZE:
case SNDCTL_DSP_GETBLKSIZE:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn = wrch ? wrch : rdch;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (chn) {
CHN_LOCK(chn);
*arg_i = sndbuf_getblksz(chn->bufsoft);
CHN_UNLOCK(chn);
} else {
*arg_i = 0;
ret = EINVAL;
}
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;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SNDCTL_DSP_SETBLKSIZE:
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
RANGE(*arg_i, 16, 65536);
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
PCM_ACQUIRE_QUICK(d);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (wrch) {
CHN_LOCK(wrch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
chn_setblocksize(wrch, 2, *arg_i);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
}
if (rdch) {
CHN_LOCK(rdch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
chn_setblocksize(rdch, 2, *arg_i);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
}
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
PCM_RELEASE_QUICK(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_RESET:
DEB(printf("dsp reset\n"));
Reduce latency when using the SNDCTL_DSP_RESET ioctl by calling chn_resetbuf(). Submited by: Pyun YongHyeon <yongari@kt-is.co.kr> Approved by: tanimura (mentor)
2004-01-20 05:30:09 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
chn_abort(wrch);
Reduce latency when using the SNDCTL_DSP_RESET ioctl by calling chn_resetbuf(). Submited by: Pyun YongHyeon <yongari@kt-is.co.kr> Approved by: tanimura (mentor)
2004-01-20 05:30:09 +00:00
chn_resetbuf(wrch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
Reduce latency when using the SNDCTL_DSP_RESET ioctl by calling chn_resetbuf(). Submited by: Pyun YongHyeon <yongari@kt-is.co.kr> Approved by: tanimura (mentor)
2004-01-20 05:30:09 +00:00
}
if (rdch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
chn_abort(rdch);
Reduce latency when using the SNDCTL_DSP_RESET ioctl by calling chn_resetbuf(). Submited by: Pyun YongHyeon <yongari@kt-is.co.kr> Approved by: tanimura (mentor)
2004-01-20 05:30:09 +00:00
chn_resetbuf(rdch);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
Reduce latency when using the SNDCTL_DSP_RESET ioctl by calling chn_resetbuf(). Submited by: Pyun YongHyeon <yongari@kt-is.co.kr> Approved by: tanimura (mentor)
2004-01-20 05:30:09 +00:00
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_SYNC:
Fix the "dsp sync" debugging printf that was unconditional. Fix a few formatting nits/glitches/whatever.
1999-09-19 09:53:57 +00:00
DEB(printf("dsp sync\n"));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
/* chn_sync may sleep */
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (wrch) {
CHN_LOCK(wrch);
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(wrch, 0);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
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
ret = EINVAL;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_SPEED:
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
/* chn_setspeed may sleep */
tmp = 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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
move various fields from pcm_channel to snd_dbuf improve chn_read code- don't stop/restart on overrun, just dump data more error checking on ioctls
2000-05-26 21:55:13 +00:00
ret = chn_setspeed(wrch, *arg_i);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
tmp = wrch->speed;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (rdch && ret == 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
move various fields from pcm_channel to snd_dbuf improve chn_read code- don't stop/restart on overrun, just dump data more error checking on ioctls
2000-05-26 21:55:13 +00:00
ret = chn_setspeed(rdch, *arg_i);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (tmp == 0)
tmp = rdch->speed;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
*arg_i = tmp;
break;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SOUND_PCM_READ_RATE:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn = wrch ? wrch : rdch;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (chn) {
CHN_LOCK(chn);
*arg_i = chn->speed;
CHN_UNLOCK(chn);
} else {
*arg_i = 0;
ret = EINVAL;
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_STEREO:
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
tmp = -1;
*arg_i = (*arg_i)? AFMT_STEREO : 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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
ret = chn_setformat(wrch, (wrch->format & ~AFMT_STEREO) | *arg_i);
tmp = (wrch->format & AFMT_STEREO)? 1 : 0;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (rdch && ret == 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
ret = chn_setformat(rdch, (rdch->format & ~AFMT_STEREO) | *arg_i);
if (tmp == -1)
tmp = (rdch->format & AFMT_STEREO)? 1 : 0;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
*arg_i = tmp;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SOUND_PCM_WRITE_CHANNELS:
move various fields from pcm_channel to snd_dbuf improve chn_read code- don't stop/restart on overrun, just dump data more error checking on ioctls
2000-05-26 21:55:13 +00:00
/* case SNDCTL_DSP_CHANNELS: ( == SOUND_PCM_WRITE_CHANNELS) */
handle bogus values passed to SOUND_PCM_WRITE_CHANNELS/SNDCTL_DSP_CHANNELS ioctls better- if 0 is passed, return the current setting. if 1 or 2 are passed, behave accordingly. treat values >2 as 2. PR: kern/29964
2001-08-23 12:21:12 +00:00
if (*arg_i != 0) {
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
tmp = 0;
handle bogus values passed to SOUND_PCM_WRITE_CHANNELS/SNDCTL_DSP_CHANNELS ioctls better- if 0 is passed, return the current setting. if 1 or 2 are passed, behave accordingly. treat values >2 as 2. PR: kern/29964
2001-08-23 12:21:12 +00:00
*arg_i = (*arg_i != 1)? AFMT_STEREO : 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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
ret = chn_setformat(wrch, (wrch->format & ~AFMT_STEREO) | *arg_i);
tmp = (wrch->format & AFMT_STEREO)? 2 : 1;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (rdch && ret == 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
ret = chn_setformat(rdch, (rdch->format & ~AFMT_STEREO) | *arg_i);
if (tmp == 0)
tmp = (rdch->format & AFMT_STEREO)? 2 : 1;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
*arg_i = tmp;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
} else {
chn = wrch ? wrch : rdch;
CHN_LOCK(chn);
*arg_i = (chn->format & AFMT_STEREO) ? 2 : 1;
CHN_UNLOCK(chn);
}
move various fields from pcm_channel to snd_dbuf improve chn_read code- don't stop/restart on overrun, just dump data more error checking on ioctls
2000-05-26 21:55:13 +00:00
break;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SOUND_PCM_READ_CHANNELS:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn = wrch ? wrch : rdch;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (chn) {
CHN_LOCK(chn);
*arg_i = (chn->format & AFMT_STEREO) ? 2 : 1;
CHN_UNLOCK(chn);
} else {
*arg_i = 0;
ret = EINVAL;
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_GETFMTS: /* returns a mask of supported fmts */
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn = wrch ? wrch : rdch;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (chn) {
CHN_LOCK(chn);
*arg_i = chn_getformats(chn);
CHN_UNLOCK(chn);
} else {
*arg_i = 0;
ret = EINVAL;
}
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;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SNDCTL_DSP_SETFMT: /* sets _one_ format */
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
if ((*arg_i != AFMT_QUERY)) {
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
tmp = 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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
ret = chn_setformat(wrch, (*arg_i) | (wrch->format & AFMT_STEREO));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
tmp = wrch->format & ~AFMT_STEREO;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (rdch && ret == 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
ret = chn_setformat(rdch, (*arg_i) | (rdch->format & AFMT_STEREO));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (tmp == 0)
tmp = rdch->format & ~AFMT_STEREO;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
*arg_i = tmp;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
} else {
chn = wrch ? wrch : rdch;
CHN_LOCK(chn);
*arg_i = chn->format & ~AFMT_STEREO;
CHN_UNLOCK(chn);
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_SETFRAGMENT:
DEB(printf("SNDCTL_DSP_SETFRAGMENT 0x%08x\n", *(int *)arg));
{
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
uint32_t fragln = (*arg_i) & 0x0000ffff;
uint32_t maxfrags = ((*arg_i) & 0xffff0000) >> 16;
uint32_t fragsz;
uint32_t r_maxfrags, r_fragsz;
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
RANGE(fragln, 4, 16);
fragsz = 1 << fragln;
if (maxfrags == 0)
maxfrags = CHN_2NDBUFMAXSIZE / fragsz;
Set a lower bound on fragment size rather than returning a failure when the user specifies a maximum fragment size < 2. This is the behavior that Linux provides and fixes the problem I've observed in Tribes2 where sounds effects are delayed by 1/2 a second.
2003-06-26 13:13:18 +00:00
if (maxfrags < 2)
maxfrags = 2;
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
if (maxfrags * fragsz > CHN_2NDBUFMAXSIZE)
maxfrags = CHN_2NDBUFMAXSIZE / fragsz;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
DEB(printf("SNDCTL_DSP_SETFRAGMENT %d frags, %d sz\n", maxfrags, fragsz));
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
PCM_ACQUIRE_QUICK(d);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
if (rdch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
ret = chn_setblocksize(rdch, maxfrags, fragsz);
From luigi: In SNDCTL_DSP_SETFRAGMENT, if you specify both read and write channels, the existing code first acts on the read channel, but as a side effect it updates the arguments (maxfrags, fragsz) passed by the caller according to acceptable values for the read channel, and then uses the modified values to act on the write channel. The problem with this approach is that, given a (maxfrags, fragsz) user-specified value, the actual values computed by the read and write channels may differ: e.g. the read channel might want to allocate more fragments than what the user specified because it has no side-effects on the delay and it helps in case of slow readers, whereas the write channel needs to use as few fragments as possible to keep the audio latency low (very important with telephony apps). This patch stores the values computed by the read channel into temproary variables so the write channel will use the actual arguments of the ioctl. This patch is very helpful with telephony apps such as asterisk. Submitted by: luigi Approved by: netchild (mentor)
2005-11-14 18:20:47 +00:00
r_maxfrags = sndbuf_getblkcnt(rdch->bufsoft);
r_fragsz = sndbuf_getblksz(rdch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
From luigi: In SNDCTL_DSP_SETFRAGMENT, if you specify both read and write channels, the existing code first acts on the read channel, but as a side effect it updates the arguments (maxfrags, fragsz) passed by the caller according to acceptable values for the read channel, and then uses the modified values to act on the write channel. The problem with this approach is that, given a (maxfrags, fragsz) user-specified value, the actual values computed by the read and write channels may differ: e.g. the read channel might want to allocate more fragments than what the user specified because it has no side-effects on the delay and it helps in case of slow readers, whereas the write channel needs to use as few fragments as possible to keep the audio latency low (very important with telephony apps). This patch stores the values computed by the read channel into temproary variables so the write channel will use the actual arguments of the ioctl. This patch is very helpful with telephony apps such as asterisk. Submitted by: luigi Approved by: netchild (mentor)
2005-11-14 18:20:47 +00:00
} else {
r_maxfrags = maxfrags;
r_fragsz = fragsz;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
if (wrch && ret == 0) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
ret = chn_setblocksize(wrch, maxfrags, fragsz);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
maxfrags = sndbuf_getblkcnt(wrch->bufsoft);
fragsz = sndbuf_getblksz(wrch->bufsoft);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
From luigi: In SNDCTL_DSP_SETFRAGMENT, if you specify both read and write channels, the existing code first acts on the read channel, but as a side effect it updates the arguments (maxfrags, fragsz) passed by the caller according to acceptable values for the read channel, and then uses the modified values to act on the write channel. The problem with this approach is that, given a (maxfrags, fragsz) user-specified value, the actual values computed by the read and write channels may differ: e.g. the read channel might want to allocate more fragments than what the user specified because it has no side-effects on the delay and it helps in case of slow readers, whereas the write channel needs to use as few fragments as possible to keep the audio latency low (very important with telephony apps). This patch stores the values computed by the read channel into temproary variables so the write channel will use the actual arguments of the ioctl. This patch is very helpful with telephony apps such as asterisk. Submitted by: luigi Approved by: netchild (mentor)
2005-11-14 18:20:47 +00:00
} else { /* use whatever came from the read channel */
maxfrags = r_maxfrags;
fragsz = r_fragsz;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
PCM_RELEASE_QUICK(d);
make mmap sort-of work. there seem to be interactions with certain hw drivers, so still work in progress. do various mmap-related ioctls right. improve blocksize control. bits of cleanup.
2000-04-15 05:04:12 +00:00
move various fields from pcm_channel to snd_dbuf improve chn_read code- don't stop/restart on overrun, just dump data more error checking on ioctls
2000-05-26 21:55:13 +00:00
fragln = 0;
while (fragsz > 1) {
fragln++;
fragsz >>= 1;
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
*arg_i = (maxfrags << 16) | fragln;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
fix SNDCTL_DSP_GETISPACE behaviour
2001-10-24 12:40:54 +00:00
case SNDCTL_DSP_GETISPACE:
- latest 2ndbuffer patch - make chn_setdir work for rec on isa cards - note: es1371 does not irq in smp Submitted by: tanimura
1999-12-29 03:46:54 +00:00
/* return the size of data available in the input queue */
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
{
audio_buf_info *a = (audio_buf_info *)arg;
if (rdch) {
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct snd_dbuf *bs = rdch->bufsoft;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
fix SNDCTL_DSP_GETISPACE behaviour
2001-10-24 12:40:54 +00:00
a->bytes = sndbuf_getready(bs);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
a->fragments = a->bytes / sndbuf_getblksz(bs);
a->fragstotal = sndbuf_getblkcnt(bs);
a->fragsize = sndbuf_getblksz(bs);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
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
ret = EINVAL;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case SNDCTL_DSP_GETOSPACE:
/* return space available in the output queue */
{
audio_buf_info *a = (audio_buf_info *)arg;
if (wrch) {
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct snd_dbuf *bs = wrch->bufsoft;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
Disable frantic DMA update within few SNDCTL_DSP_* ioctl. This should reduce huge playback / recording latency for applications that try to act smarter and manage their own buffering (XMMS, Skype, etc.). Note to Skype + via8xxx users: Remove previous hackish "hint.pcm.<unit>.via_dxs_disabled" from kernel hint and see whether this changes cure all those annoying sound issues.
2005-12-30 07:33:28 +00:00
/* XXX abusive DMA update: chn_wrupdate(wrch); */
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
a->bytes = sndbuf_getfree(bs);
a->fragments = a->bytes / sndbuf_getblksz(bs);
a->fragstotal = sndbuf_getblkcnt(bs);
a->fragsize = sndbuf_getblksz(bs);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
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
ret = EINVAL;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case SNDCTL_DSP_GETIPTR:
{
count_info *a = (count_info *)arg;
if (rdch) {
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct snd_dbuf *bs = rdch->bufsoft;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
Disable frantic DMA update within few SNDCTL_DSP_* ioctl. This should reduce huge playback / recording latency for applications that try to act smarter and manage their own buffering (XMMS, Skype, etc.). Note to Skype + via8xxx users: Remove previous hackish "hint.pcm.<unit>.via_dxs_disabled" from kernel hint and see whether this changes cure all those annoying sound issues.
2005-12-30 07:33:28 +00:00
/* XXX abusive DMA update: chn_rdupdate(rdch); */
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
a->bytes = sndbuf_gettotal(bs);
a->blocks = sndbuf_getblocks(bs) - rdch->blocks;
a->ptr = sndbuf_getreadyptr(bs);
rdch->blocks = sndbuf_getblocks(bs);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
} else
ret = EINVAL;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case SNDCTL_DSP_GETOPTR:
{
count_info *a = (count_info *)arg;
if (wrch) {
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct snd_dbuf *bs = wrch->bufsoft;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
Disable frantic DMA update within few SNDCTL_DSP_* ioctl. This should reduce huge playback / recording latency for applications that try to act smarter and manage their own buffering (XMMS, Skype, etc.). Note to Skype + via8xxx users: Remove previous hackish "hint.pcm.<unit>.via_dxs_disabled" from kernel hint and see whether this changes cure all those annoying sound issues.
2005-12-30 07:33:28 +00:00
/* XXX abusive DMA update: chn_wrupdate(wrch); */
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
a->bytes = sndbuf_gettotal(bs);
a->blocks = sndbuf_getblocks(bs) - wrch->blocks;
a->ptr = sndbuf_getreadyptr(bs);
wrch->blocks = sndbuf_getblocks(bs);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
} else
ret = EINVAL;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case SNDCTL_DSP_GETCAPS:
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
pcm_lock(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
*arg_i = DSP_CAP_REALTIME | DSP_CAP_MMAP | DSP_CAP_TRIGGER;
many changes: * add new channels to the end of the list so channels used in order of addition * de-globalise definition of struct snddev_info and provide accessor functions where necessary. * move the $FreeBSD$ tag in each .c file into a macro and allow the /dev/sndstat handler to display these when set to maximum verbosity to aid debugging. * allow each device to register its own sndstat handler to reduce the amount of groping sndstat must do in foreign structs.
2001-08-23 11:30:52 +00:00
if (rdch && wrch && !(dsp_get_flags(i_dev) & SD_F_SIMPLEX))
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
*arg_i |= DSP_CAP_DUPLEX;
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
pcm_unlock(d);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SOUND_PCM_READ_BITS:
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
chn = wrch ? wrch : rdch;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
if (chn) {
CHN_LOCK(chn);
if (chn->format & AFMT_8BIT)
*arg_i = 8;
else if (chn->format & AFMT_16BIT)
*arg_i = 16;
else if (chn->format & AFMT_24BIT)
*arg_i = 24;
else if (chn->format & AFMT_32BIT)
*arg_i = 32;
else
ret = EINVAL;
CHN_UNLOCK(chn);
} else {
*arg_i = 0;
Release lock for a while during chn_reset() / pcm_chnalloc() operation while malloc()ing, this fixes LOR 129. See - http://lists.freebsd.org/pipermail/freebsd-current/2005-June/051157.html - http://lists.freebsd.org/pipermail/freebsd-current/2005-August/054620.html - http://sources.zabbadoz.net/freebsd/lor.html#129 Submitted by: Ariff Abdullah <skywizard@MyBSD.org.my>
2005-09-10 17:51:38 +00:00
ret = EINVAL;
Fix some potential NULL pointer dereferences. This is supposed to fix some Coverity Prevent errors (Ariff didn't looked at the CID's (ENOTIME), I just told him that there are some problems in function dsp_ioctl()). CID: 215-218 Found with: Coverity Prevent(tm) Submitted by: ariff MFC after: 5 days
2006-01-29 16:48:41 +00:00
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
case SNDCTL_DSP_SETTRIGGER:
if (rdch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(rdch);
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
rdch->flags &= ~(CHN_F_TRIGGERED | CHN_F_NOTRIGGER);
MFS: make SNDCTL_DSP_SETTRIGGER work
2001-08-23 11:58:38 +00:00
if (*arg_i & PCM_ENABLE_INPUT)
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
chn_start(rdch, 1);
MFS: make SNDCTL_DSP_SETTRIGGER work
2001-08-23 11:58:38 +00:00
else
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
rdch->flags |= CHN_F_NOTRIGGER;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(rdch);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
wrch->flags &= ~(CHN_F_TRIGGERED | CHN_F_NOTRIGGER);
MFS: make SNDCTL_DSP_SETTRIGGER work
2001-08-23 11:58:38 +00:00
if (*arg_i & PCM_ENABLE_OUTPUT)
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
chn_start(wrch, 1);
MFS: make SNDCTL_DSP_SETTRIGGER work
2001-08-23 11:58:38 +00:00
else
implement settrigger according to spec fixes for non-blocking mode
2000-04-23 18:09:18 +00:00
wrch->flags |= CHN_F_NOTRIGGER;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
break;
case SNDCTL_DSP_GETTRIGGER:
*arg_i = 0;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
if (wrch) {
CHN_LOCK(wrch);
if (wrch->flags & CHN_F_TRIGGERED)
*arg_i |= PCM_ENABLE_OUTPUT;
CHN_UNLOCK(wrch);
}
if (rdch) {
CHN_LOCK(rdch);
if (rdch->flags & CHN_F_TRIGGERED)
*arg_i |= PCM_ENABLE_INPUT;
CHN_UNLOCK(rdch);
}
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
break;
- latest 2ndbuffer patch - make chn_setdir work for rec on isa cards - note: es1371 does not irq in smp Submitted by: tanimura
1999-12-29 03:46:54 +00:00
case SNDCTL_DSP_GETODELAY:
if (wrch) {
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
struct snd_dbuf *bs = wrch->bufsoft;
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
Disable frantic DMA update within few SNDCTL_DSP_* ioctl. This should reduce huge playback / recording latency for applications that try to act smarter and manage their own buffering (XMMS, Skype, etc.). Note to Skype + via8xxx users: Remove previous hackish "hint.pcm.<unit>.via_dxs_disabled" from kernel hint and see whether this changes cure all those annoying sound issues.
2005-12-30 07:33:28 +00:00
/* XXX abusive DMA update: chn_wrupdate(wrch); */
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
*arg_i = sndbuf_getready(bs);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
- latest 2ndbuffer patch - make chn_setdir work for rec on isa cards - note: es1371 does not irq in smp Submitted by: tanimura
1999-12-29 03:46:54 +00:00
} else
ret = EINVAL;
break;
implement SNDCTL_DSP_POST
2000-10-27 22:20:45 +00:00
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
case SNDCTL_DSP_POST:
if (wrch) {
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_LOCK(wrch);
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
wrch->flags &= ~CHN_F_NOTRIGGER;
chn_start(wrch, 1);
Lock channels only as necessary in dsp_ioctl(), and only lock one channel at a time unless it is actually necessary to lock both. This avoids problems with lock order reversal and malloc() calls with a mutex held when lower level code unlocks a channel, calls malloc(), and relocks the channel. This also avoids the cost of some unnecessary locking and unlocking. Tested by: matk
2004-02-28 19:42:48 +00:00
CHN_UNLOCK(wrch);
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
ret = EINVAL;
handle AFMT_QUERY right
2000-10-26 01:34:54 +00:00
break;
implement SNDCTL_DSP_POST
2000-10-27 22:20:45 +00:00
Implement SNDCTL_DSP_SETDUPLEX. This may fix sound apps which want to use full duplex mode. Approved by: matk
2004-07-02 15:31:44 +00:00
case SNDCTL_DSP_SETDUPLEX:
/*
* switch to full-duplex mode if card is in half-duplex
* mode and is able to work in full-duplex mode
*/
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
pcm_lock(d);
Implement SNDCTL_DSP_SETDUPLEX. This may fix sound apps which want to use full duplex mode. Approved by: matk
2004-07-02 15:31:44 +00:00
if (rdch && wrch && (dsp_get_flags(i_dev) & SD_F_SIMPLEX))
dsp_set_flags(i_dev, dsp_get_flags(i_dev)^SD_F_SIMPLEX);
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
pcm_unlock(d);
Implement SNDCTL_DSP_SETDUPLEX. This may fix sound apps which want to use full duplex mode. Approved by: matk
2004-07-02 15:31:44 +00:00
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
/*
* The following four ioctls are simple wrappers around mixer_ioctl
* with no further processing. xcmd is short for "translated
* command".
*/
case SNDCTL_DSP_GETRECVOL:
if (xcmd == 0)
xcmd = SOUND_MIXER_READ_RECLEV;
/* FALLTHROUGH */
case SNDCTL_DSP_SETRECVOL:
if (xcmd == 0)
xcmd = SOUND_MIXER_WRITE_RECLEV;
/* FALLTHROUGH */
case SNDCTL_DSP_GETPLAYVOL:
if (xcmd == 0)
xcmd = SOUND_MIXER_READ_PCM;
/* FALLTHROUGH */
case SNDCTL_DSP_SETPLAYVOL:
if (xcmd == 0)
xcmd = SOUND_MIXER_WRITE_PCM;
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 (d->mixer_dev != NULL) {
PCM_ACQUIRE_QUICK(d);
ret = mixer_ioctl_cmd(d->mixer_dev, xcmd, arg, -1, td,
MIXER_CMD_DIRECT);
PCM_RELEASE_QUICK(d);
} else
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
ret = ENOTSUP;
break;
case SNDCTL_DSP_GET_RECSRC_NAMES:
case SNDCTL_DSP_GET_RECSRC:
case SNDCTL_DSP_SET_RECSRC:
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 (d->mixer_dev != NULL) {
PCM_ACQUIRE_QUICK(d);
ret = mixer_ioctl_cmd(d->mixer_dev, cmd, arg, -1, td,
MIXER_CMD_DIRECT);
PCM_RELEASE_QUICK(d);
} else
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
ret = ENOTSUP;
break;
/*
* The following 3 ioctls aren't very useful at the moment. For
* now, only a single channel is associated with a cdev (/dev/dspN
* instance), so there's only a single output routing to use (i.e.,
* the wrch bound to this cdev).
*/
case SNDCTL_DSP_GET_PLAYTGT_NAMES:
{
oss_mixer_enuminfo *ei;
ei = (oss_mixer_enuminfo *)arg;
ei->dev = 0;
ei->ctrl = 0;
ei->version = 0; /* static for now */
ei->strindex[0] = 0;
if (wrch != NULL) {
ei->nvalues = 1;
strlcpy(ei->strings, wrch->name,
sizeof(ei->strings));
} else {
ei->nvalues = 0;
ei->strings[0] = '\0';
}
}
break;
case SNDCTL_DSP_GET_PLAYTGT:
case SNDCTL_DSP_SET_PLAYTGT: /* yes, they are the same for now */
/*
* Re: SET_PLAYTGT
* OSSv4: "The value that was accepted by the device will
* be returned back in the variable pointed by the
* argument."
*/
if (wrch != NULL)
*arg_i = 0;
else
ret = EINVAL;
break;
case SNDCTL_DSP_SILENCE:
/*
* Flush the software (pre-feed) buffer, but try to minimize playback
* interruption. (I.e., record unplayed samples with intent to
* restore by SNDCTL_DSP_SKIP.) Intended for application "pause"
* functionality.
*/
if (wrch == NULL)
ret = EINVAL;
else {
struct snd_dbuf *bs;
CHN_LOCK(wrch);
while (wrch->inprog != 0)
cv_wait(&wrch->cv, wrch->lock);
bs = wrch->bufsoft;
if ((bs->shadbuf != NULL) && (sndbuf_getready(bs) > 0)) {
bs->sl = sndbuf_getready(bs);
sndbuf_dispose(bs, bs->shadbuf, sndbuf_getready(bs));
sndbuf_fillsilence(bs);
chn_start(wrch, 0);
}
CHN_UNLOCK(wrch);
}
break;
case SNDCTL_DSP_SKIP:
/*
* OSSv4 docs: "This ioctl call discards all unplayed samples in the
* playback buffer by moving the current write position immediately
* before the point where the device is currently reading the samples."
*/
if (wrch == NULL)
ret = EINVAL;
else {
struct snd_dbuf *bs;
CHN_LOCK(wrch);
while (wrch->inprog != 0)
cv_wait(&wrch->cv, wrch->lock);
bs = wrch->bufsoft;
if ((bs->shadbuf != NULL) && (bs->sl > 0)) {
sndbuf_softreset(bs);
sndbuf_acquire(bs, bs->shadbuf, bs->sl);
bs->sl = 0;
chn_start(wrch, 0);
}
CHN_UNLOCK(wrch);
}
break;
case SNDCTL_DSP_CURRENT_OPTR:
case SNDCTL_DSP_CURRENT_IPTR:
/**
* @note Changing formats resets the buffer counters, which differs
* from the 4Front drivers. However, I don't expect this to be
* much of a problem.
*
* @note In a test where @c CURRENT_OPTR is called immediately after write
* returns, this driver is about 32K samples behind whereas
* 4Front's is about 8K samples behind. Should determine source
* of discrepancy, even if only out of curiosity.
*
* @todo Actually test SNDCTL_DSP_CURRENT_IPTR.
*/
chn = (cmd == SNDCTL_DSP_CURRENT_OPTR) ? wrch : rdch;
if (chn == NULL)
ret = EINVAL;
else {
struct snd_dbuf *bs;
/* int tmp; */
oss_count_t *oc = (oss_count_t *)arg;
CHN_LOCK(chn);
bs = chn->bufsoft;
#if 0
tmp = (sndbuf_getsize(b) + chn_getptr(chn) - sndbuf_gethwptr(b)) % sndbuf_getsize(b);
oc->samples = (sndbuf_gettotal(b) + tmp) / sndbuf_getbps(b);
oc->fifo_samples = (sndbuf_getready(b) - tmp) / sndbuf_getbps(b);
#else
oc->samples = sndbuf_gettotal(bs) / sndbuf_getbps(bs);
oc->fifo_samples = sndbuf_getready(bs) / sndbuf_getbps(bs);
#endif
CHN_UNLOCK(chn);
}
break;
case SNDCTL_DSP_HALT_OUTPUT:
case SNDCTL_DSP_HALT_INPUT:
chn = (cmd == SNDCTL_DSP_HALT_OUTPUT) ? wrch : rdch;
if (chn == NULL)
ret = EINVAL;
else {
CHN_LOCK(chn);
chn_abort(chn);
CHN_UNLOCK(chn);
}
break;
case SNDCTL_DSP_LOW_WATER:
/*
* Set the number of bytes required to attract attention by
* select/poll.
*/
if (wrch != NULL) {
CHN_LOCK(wrch);
wrch->lw = (*arg_i > 1) ? *arg_i : 1;
CHN_UNLOCK(wrch);
}
if (rdch != NULL) {
CHN_LOCK(rdch);
rdch->lw = (*arg_i > 1) ? *arg_i : 1;
CHN_UNLOCK(rdch);
}
break;
case SNDCTL_DSP_GETERROR:
/*
* OSSv4 docs: "All errors and counters will automatically be
* cleared to zeroes after the call so each call will return only
* the errors that occurred after the previous invocation. ... The
* play_underruns and rec_overrun fields are the only usefull fields
* returned by OSS 4.0."
*/
{
audio_errinfo *ei = (audio_errinfo *)arg;
bzero((void *)ei, sizeof(*ei));
if (wrch != NULL) {
CHN_LOCK(wrch);
ei->play_underruns = wrch->xruns;
wrch->xruns = 0;
CHN_UNLOCK(wrch);
}
if (rdch != NULL) {
CHN_LOCK(rdch);
ei->rec_overruns = rdch->xruns;
rdch->xruns = 0;
CHN_UNLOCK(rdch);
}
}
break;
case SNDCTL_DSP_SYNCGROUP:
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
PCM_ACQUIRE_QUICK(d);
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
ret = dsp_oss_syncgroup(wrch, rdch, (oss_syncgroup *)arg);
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
PCM_RELEASE_QUICK(d);
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
break;
case SNDCTL_DSP_SYNCSTART:
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
PCM_ACQUIRE_QUICK(d);
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
ret = dsp_oss_syncstart(*arg_i);
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
PCM_RELEASE_QUICK(d);
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
break;
case SNDCTL_DSP_POLICY:
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
PCM_ACQUIRE_QUICK(d);
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
ret = dsp_oss_policy(wrch, rdch, *arg_i);
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
PCM_RELEASE_QUICK(d);
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
break;
#ifdef OSSV4_EXPERIMENT
/*
* XXX The following ioctls are not yet supported and just return
* EINVAL.
*/
case SNDCTL_DSP_GETOPEAKS:
case SNDCTL_DSP_GETIPEAKS:
chn = (cmd == SNDCTL_DSP_GETOPEAKS) ? wrch : rdch;
if (chn == NULL)
ret = EINVAL;
else {
oss_peaks_t *op = (oss_peaks_t *)arg;
int lpeak, rpeak;
CHN_LOCK(chn);
ret = chn_getpeaks(chn, &lpeak, &rpeak);
if (ret == -1)
ret = EINVAL;
else {
(*op)[0] = lpeak;
(*op)[1] = rpeak;
}
CHN_UNLOCK(chn);
}
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
/*
* XXX Once implemented, revisit this for proper cv protection
* (if necessary).
*/
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
case SNDCTL_DSP_COOKEDMODE:
ret = dsp_oss_cookedmode(wrch, rdch, *arg_i);
break;
case SNDCTL_DSP_GET_CHNORDER:
ret = dsp_oss_getchnorder(wrch, rdch, (unsigned long long *)arg);
break;
case SNDCTL_DSP_SET_CHNORDER:
ret = dsp_oss_setchnorder(wrch, rdch, (unsigned long long *)arg);
break;
case SNDCTL_GETLABEL:
ret = dsp_oss_getlabel(wrch, rdch, (oss_label_t *)arg);
break;
case SNDCTL_SETLABEL:
ret = dsp_oss_setlabel(wrch, rdch, (oss_label_t *)arg);
break;
case SNDCTL_GETSONG:
ret = dsp_oss_getsong(wrch, rdch, (oss_longname_t *)arg);
break;
case SNDCTL_SETSONG:
ret = dsp_oss_setsong(wrch, rdch, (oss_longname_t *)arg);
break;
case SNDCTL_SETNAME:
ret = dsp_oss_setname(wrch, rdch, (oss_longname_t *)arg);
break;
#if 0
/**
* @note The SNDCTL_CARDINFO ioctl was omitted per 4Front developer
* documentation. "The usability of this call is very limited. It's
* provided only for completeness of the API. OSS API doesn't have
* any concept of card. Any information returned by this ioctl calld
* is reserved exclusively for the utility programs included in the
* OSS package. Applications should not try to use for this
* information in any ways."
*/
case SNDCTL_CARDINFO:
ret = EINVAL;
break;
/**
* @note The S/PDIF interface ioctls, @c SNDCTL_DSP_READCTL and
* @c SNDCTL_DSP_WRITECTL have been omitted at the suggestion of
* 4Front Technologies.
*/
case SNDCTL_DSP_READCTL:
case SNDCTL_DSP_WRITECTL:
ret = EINVAL;
break;
#endif /* !0 (explicitly omitted ioctls) */
#endif /* !OSSV4_EXPERIMENT */
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SNDCTL_DSP_MAPINBUF:
case SNDCTL_DSP_MAPOUTBUF:
case SNDCTL_DSP_SETSYNCRO:
/* undocumented */
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +00:00
case SNDCTL_DSP_SUBDIVIDE:
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
case SOUND_PCM_WRITE_FILTER:
case SOUND_PCM_READ_FILTER:
/* dunno what these do, don't sound important */
Implement SNDCTL_DSP_SETDUPLEX. This may fix sound apps which want to use full duplex mode. Approved by: matk
2004-07-02 15:31:44 +00:00
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
default:
Fix code that had rotted behind debugging macros. Approved by: cg (in principle) MFC after: 2 weeks
2002-01-25 04:14:12 +00:00
DEB(printf("default ioctl fn 0x%08lx fail\n", cmd));
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
ret = EINVAL;
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
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
relchns(i_dev, rdch, wrch, 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
PCM_GIANT_LEAVE(d);
return (ret);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_poll(struct cdev *i_dev, int events, struct thread *td)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
struct snddev_info *d;
struct pcm_channel *wrch, *rdch;
add defines and ifdefs so this code will compile on 4.x add spls so this code will work on 4.x
2001-06-23 17:36:51 +00:00
int ret, e;
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
d = dsp_get_info(i_dev);
if (!DSP_REGISTERED(d, i_dev))
return (EBADF);
PCM_GIANT_ENTER(d);
wrch = NULL;
rdch = NULL;
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
ret = 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
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
getchns(i_dev, &rdch, &wrch, SD_F_PRIO_RD | SD_F_PRIO_WR);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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 (wrch != NULL && !(wrch->flags & CHN_F_DEAD)) {
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
e = (events & (POLLOUT | POLLWRNORM));
if (e)
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
ret |= chn_poll(wrch, e, td);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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 (rdch != NULL && !(rdch->flags & CHN_F_DEAD)) {
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +00:00
e = (events & (POLLIN | POLLRDNORM));
if (e)
KSE Milestone 2 Note ALL MODULES MUST BE RECOMPILED make the kernel aware that there are smaller units of scheduling than the process. (but only allow one thread per process at this time). This is functionally equivalent to teh previousl -current except that there is a thread associated with each process. Sorry john! (your next MFC will be a doosie!) Reviewed by: peter@freebsd.org, dillon@freebsd.org X-MFC after: ha ha ha ha
2001-09-12 08:38:13 +00:00
ret |= chn_poll(rdch, e, td);
various locking fixes, rework open logic and channel registration PR: kern/28084
2001-06-14 13:31:30 +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
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
relchns(i_dev, rdch, wrch, SD_F_PRIO_RD | SD_F_PRIO_WR);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
PCM_GIANT_LEAVE(d);
return (ret);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
dsp_mmap(struct cdev *i_dev, vm_offset_t offset, vm_paddr_t *paddr, int nprot)
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +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
struct snddev_info *d;
struct pcm_channel *wrch, *rdch, *c;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
* Reject PROT_EXEC by default. It just doesn't makes sense.
* Unfortunately, we have to give up this one due to linux_mmap
* changes.
*
* http://lists.freebsd.org/pipermail/freebsd-emulation/2007-June/003698.html
*
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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 ((nprot & PROT_EXEC) && dsp_mmap_allow_prot_exec == 0)
return (-1);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
d = dsp_get_info(i_dev);
if (!DSP_REGISTERED(d, i_dev))
return (-1);
mega-commit. this introduces a new buffering mechanism which results in dramatic simplification of the channel manager. as several structures have changed, we take the opportunity to move their definitions into the source files where they are used, make them private and de-typedef them. the sound drivers are updated to use snd_setup_intr instead of bus_setup_intr, and to comply with the de-typedefed structures. the ac97, mixer and channel layers have been updated with finegrained locking, as have some drivers- not all though. the rest will follow soon.
2001-03-24 23:10:29 +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
PCM_GIANT_ENTER(d);
getchns(i_dev, &rdch, &wrch, SD_F_PRIO_RD | SD_F_PRIO_WR);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
/*
* XXX The linux api uses the nprot to select read/write buffer
* our vm system doesn't allow this, so force write buffer.
*
* This is just a quack to fool full-duplex mmap, so that at
* least playback _or_ recording works. If you really got the
* urge to make _both_ work at the same time, avoid O_RDWR.
* Just open each direction separately and mmap() it.
*
* Failure is not an option due to INVARIANTS check within
* device_pager.c, which means, we have to give up one over
* another.
*/
c = (wrch != NULL) ? wrch : rdch;
if (c == NULL || (c->flags & CHN_F_MMAP_INVALID) ||
offset >= sndbuf_getsize(c->bufsoft) ||
(wrch != NULL && (wrch->flags & CHN_F_MMAP_INVALID)) ||
(rdch != NULL && (rdch->flags & CHN_F_MMAP_INVALID))) {
don't allow mmap beyond the end of the buffer Submitted by: Philippe Anel <philippe.anel@noos.fr> (partially)
2002-01-23 04:32:18 +00:00
relchns(i_dev, rdch, wrch, SD_F_PRIO_RD | SD_F_PRIO_WR);
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
PCM_GIANT_EXIT(d);
return (-1);
don't allow mmap beyond the end of the buffer Submitted by: Philippe Anel <philippe.anel@noos.fr> (partially)
2002-01-23 04:32:18 +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
/* XXX full-duplex quack. */
if (wrch != NULL)
wrch->flags |= CHN_F_MAPPED;
if (rdch != NULL)
rdch->flags |= CHN_F_MAPPED;
don't set a flag that is already set
2001-07-01 18:57:16 +00:00
Cleanup of the d_mmap_t interface. - Get rid of the useless atop() / pmap_phys_address() detour. The device mmap handlers must now give back the physical address without atop()'ing it. - Don't borrow the physical address of the mapping in the returned int. Now we properly pass a vm_offset_t * and expect it to be filled by the mmap handler when the mapping was successful. The mmap handler must now return 0 when successful, any other value is considered as an error. Previously, returning -1 was the only way to fail. This change thus accidentally fixes some devices which were bogusly returning errno constants which would have been considered as addresses by the device pager. - Garbage collect the poorly named pmap_phys_address() now that it's no longer used. - Convert all the d_mmap_t consumers to the new API. I'm still not sure wheter we need a __FreeBSD_version bump for this, since and we didn't guarantee API/ABI stability until 5.1-RELEASE. Discussed with: alc, phk, jake Reviewed by: peter Compile-tested on: LINT (i386), GENERIC (alpha and sparc64) Runtime-tested on: i386
2003-02-25 03:21:22 +00:00
*paddr = vtophys(sndbuf_getbufofs(c->bufsoft, offset));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
relchns(i_dev, rdch, wrch, SD_F_PRIO_RD | SD_F_PRIO_WR);
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
PCM_GIANT_LEAVE(d);
return (0);
say hello to newpcm. it is not yet enabled, requiring new pnp code from dfr to compile successfully. further details will be provided in the commit enabling newpcm.
1999-09-01 04:08:39 +00:00
}
add defines and ifdefs so this code will compile on 4.x add spls so this code will work on 4.x
2001-06-23 17:36:51 +00:00
#ifdef USING_DEVFS
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +00:00
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
/* So much for dev_stdclone() */
static int
dsp_stdclone(char *name, char *namep, char *sep, int use_sep, int *u, int *c)
{
size_t len;
len = strlen(namep);
if (bcmp(name, namep, len) != 0)
return (ENODEV);
name += len;
if (isdigit(*name) == 0)
return (ENODEV);
len = strlen(sep);
if (*name == '0' && !(name[1] == '\0' || bcmp(name + 1, sep, len) == 0))
return (ENODEV);
for (*u = 0; isdigit(*name) != 0; name++) {
*u *= 10;
*u += *name - '0';
if (*u > dsp_umax)
return (ENODEV);
}
if (*name == '\0')
return ((use_sep == 0) ? 0 : ENODEV);
if (bcmp(name, sep, len) != 0 || isdigit(name[len]) == 0)
return (ENODEV);
name += len;
if (*name == '0' && name[1] != '\0')
return (ENODEV);
for (*c = 0; isdigit(*name) != 0; name++) {
*c *= 10;
*c += *name - '0';
if (*c > dsp_cmax)
return (ENODEV);
}
if (*name != '\0')
return (ENODEV);
return (0);
}
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
static void
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
dsp_clone(void *arg,
#if __FreeBSD_version >= 600034
struct ucred *cred,
#endif
char *name, int namelen, struct cdev **dev)
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
{
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;
struct snd_clone_entry *ce;
struct pcm_channel *c;
int i, unit, udcmask, cunit, devtype, devhw, devcmax, tumax;
char *devname, *devsep;
KASSERT(dsp_umax >= 0 && dsp_cmax >= 0, ("Uninitialized unit!"));
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
Second half of the dev_t cleanup. The big lines are: NODEV -> NULL NOUDEV -> NODEV udev_t -> dev_t udev2dev() -> findcdev() Various minor adjustments including handling of userland access to kernel space struct cdev etc.
2004-06-17 17:16:53 +00:00
if (*dev != NULL)
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
return;
tidy up/simplify dsp_clone()
2001-06-27 19:51:02 +00:00
unit = -1;
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
cunit = -1;
devtype = -1;
devhw = 0;
devcmax = -1;
tumax = -1;
devname = NULL;
devsep = 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
for (i = 0; unit == -1 &&
i < (sizeof(dsp_cdevs) / sizeof(dsp_cdevs[0])); i++) {
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
devtype = dsp_cdevs[i].type;
devname = dsp_cdevs[i].name;
devsep = dsp_cdevs[i].sep;
devhw = dsp_cdevs[i].hw;
devcmax = dsp_cdevs[i].max - 1;
if (strcmp(name, devname) == 0)
tidy up/simplify dsp_clone()
2001-06-27 19:51:02 +00:00
unit = snd_unit;
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 (dsp_stdclone(name, devname, devsep,
dsp_cdevs[i].use_sep, &unit, &cunit) != 0) {
unit = -1;
cunit = -1;
tidy up/simplify dsp_clone()
2001-06-27 19:51:02 +00:00
}
}
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +00:00
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
d = devclass_get_softc(pcm_devclass, unit);
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 (!PCM_REGISTERED(d) || d->clones == NULL)
Fix a panic in dsp_clone when trying to access a sound device that doesn't exists. I'm using my discretion and committing without mentor approval since Seigo is away. Noticed by: Maxime Henrion <mux@freebsd.org>
2004-01-25 22:46:22 +00:00
return;
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 Need Giant magic entry ??? */
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
pcm_lock(d);
if (snd_clone_disabled(d->clones)) {
pcm_unlock(d);
return;
}
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +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
PCM_WAIT(d);
PCM_ACQUIRE(d);
pcm_unlock(d);
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
udcmask = snd_u2unit(unit) | snd_d2unit(devtype);
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +00:00
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 (devhw != 0) {
KASSERT(devcmax <= dsp_cmax,
("overflow: devcmax=%d, dsp_cmax=%d", devcmax, dsp_cmax));
if (cunit > devcmax) {
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
PCM_RELEASE_QUICK(d);
Fix a panic when kldloading a sound driver. Do this by replacing the link-list of dev_t's with named variables. Remove used code. Approved by: tanimura (mentor)
2004-01-20 03:58:57 +00:00
return;
revise dsp_clone() to return the first nonbusy channel instead of simply cycling channel numbers. remove unused fields from struct snddev_info.
2001-06-17 23:23:06 +00:00
}
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
udcmask |= snd_c2unit(cunit);
CHN_FOREACH(c, d, channels.pcm) {
CHN_LOCK(c);
if (c->unit != udcmask) {
CHN_UNLOCK(c);
continue;
}
CHN_UNLOCK(c);
udcmask &= ~snd_c2unit(cunit);
/*
* Temporarily increase clone maxunit to overcome
* vchan flexibility.
*
* # sysctl dev.pcm.0.play.vchans=256
* dev.pcm.0.play.vchans: 1 -> 256
* # cat /dev/zero > /dev/dsp0.vp255 &
* [1] 17296
* # sysctl dev.pcm.0.play.vchans=0
* dev.pcm.0.play.vchans: 256 -> 1
* # fg
* [1] + running cat /dev/zero > /dev/dsp0.vp255
* ^C
* # cat /dev/zero > /dev/dsp0.vp255
* zsh: operation not supported: /dev/dsp0.vp255
*/
tumax = snd_clone_getmaxunit(d->clones);
if (cunit > tumax)
snd_clone_setmaxunit(d->clones, cunit);
else
tumax = -1;
goto dsp_clone_alloc;
}
/*
* Ok, so we're requesting unallocated vchan, but still
* within maximum vchan limit.
*/
if (((devtype == SND_DEV_DSPHW_VPLAY && d->pvchancount > 0) ||
(devtype == SND_DEV_DSPHW_VREC && d->rvchancount > 0)) &&
cunit < snd_maxautovchans) {
udcmask &= ~snd_c2unit(cunit);
tumax = snd_clone_getmaxunit(d->clones);
if (cunit > tumax)
snd_clone_setmaxunit(d->clones, cunit);
else
tumax = -1;
goto dsp_clone_alloc;
}
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
PCM_RELEASE_QUICK(d);
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
return;
revise dsp_clone() to return the first nonbusy channel instead of simply cycling channel numbers. remove unused fields from struct snddev_info.
2001-06-17 23:23:06 +00:00
}
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
dsp_clone_alloc:
ce = snd_clone_alloc(d->clones, dev, &cunit, udcmask);
if (tumax != -1)
snd_clone_setmaxunit(d->clones, tumax);
if (ce != NULL) {
udcmask |= snd_c2unit(cunit);
*dev = make_dev(&dsp_cdevsw, unit2minor(udcmask),
UID_ROOT, GID_WHEEL, 0666, "%s%d%s%d",
devname, unit, devsep, cunit);
snd_clone_register(ce, *dev);
}
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
PCM_RELEASE_QUICK(d);
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 (*dev != NULL)
dev_ref(*dev);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
static void
dsp_sysinit(void *p)
{
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 (dsp_ehtag != NULL)
return;
/* initialize unit numbering */
snd_unit_init();
dsp_umax = PCMMAXUNIT;
dsp_cmax = PCMMAXCHAN;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
dsp_ehtag = EVENTHANDLER_REGISTER(dev_clone, dsp_clone, 0, 1000);
}
static void
dsp_sysuninit(void *p)
{
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 (dsp_ehtag == NULL)
return;
EVENTHANDLER_DEREGISTER(dev_clone, dsp_ehtag);
dsp_ehtag = NULL;
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
}
SYSINIT(dsp_sysinit, SI_SUB_DRIVERS, SI_ORDER_MIDDLE, dsp_sysinit, NULL);
SYSUNINIT(dsp_sysuninit, SI_SUB_DRIVERS, SI_ORDER_MIDDLE, dsp_sysuninit, NULL);
add defines and ifdefs so this code will compile on 4.x add spls so this code will work on 4.x
2001-06-23 17:36:51 +00:00
#endif
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +00:00
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
char *
dsp_unit2name(char *buf, size_t len, int unit)
{
int i, dtype;
Cast len to be a uintmax_t and make format in KASSERT match so as to avoid different sizes on different platforms types of complaints. Reviewed by: Ariff
2007-06-10 15:46:34 +00:00
KASSERT(buf != NULL && len != 0,
("bogus buf=%p len=%ju", buf, (uintmax_t)len));
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
dtype = snd_unit2d(unit);
for (i = 0; i < (sizeof(dsp_cdevs) / sizeof(dsp_cdevs[0])); i++) {
if (dtype != dsp_cdevs[i].type)
continue;
snprintf(buf, len, "%s%d%s%d", dsp_cdevs[i].name,
snd_unit2u(unit), dsp_cdevs[i].sep, snd_unit2c(unit));
return (buf);
}
return (NULL);
}
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 Handler for SNDCTL_AUDIOINFO.
*
* Gathers information about the audio device specified in ai->dev. If
* ai->dev == -1, then this function gathers information about the current
* device. If the call comes in on a non-audio device and ai->dev == -1,
* return EINVAL.
*
* This routine is supposed to go practically straight to the hardware,
* getting capabilities directly from the sound card driver, side-stepping
* the intermediate channel interface.
*
* Note, however, that the usefulness of this command is significantly
* decreased when requesting info about any device other than the one serving
* the request. While each snddev_channel refers to a specific device node,
* the converse is *not* true. Currently, when a sound device node is opened,
* the sound subsystem scans for an available audio channel (or channels, if
* opened in read+write) and then assigns them to the si_drv[12] private
* data fields. As a result, any information returned linking a channel to
* a specific character device isn't necessarily accurate.
*
* @note
* Calling threads must not hold any snddev_info or pcm_channel locks.
*
* @param dev device on which the ioctl was issued
* @param ai ioctl request data container
*
* @retval 0 success
* @retval EINVAL ai->dev specifies an invalid device
*
* @todo Verify correctness of Doxygen tags. ;)
*/
int
dsp_oss_audioinfo(struct cdev *i_dev, oss_audioinfo *ai)
{
struct pcmchan_caps *caps;
struct pcm_channel *ch;
struct snddev_info *d;
uint32_t fmts;
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 i, nchan, *rates, minch, maxch;
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
char *devname, buf[CHN_NAMELEN];
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
/*
* If probing the device that received the ioctl, make sure it's a
* DSP device. (Users may use this ioctl with /dev/mixer and
* /dev/midi.)
*/
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 (ai->dev == -1 && i_dev->si_devsw != &dsp_cdevsw)
return (EINVAL);
use a global devclass for all drivers - i'm not entirely sure why this worked before. mixer, dsp and sndstat are seperate devices - give them their own cdevsws instead of demuxing requests sent to a single cdevsw. use the si_drv1/si_drv2 fields in dev_t structures for holding information specific to an open instance of mixer/dsp. nuke /dev/{dsp,dspW,audio}[0-9]* links - this functionality is now provided using cloning. various locking fixes.
2001-06-16 21:25:10 +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
ch = 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
devname = NULL;
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
nchan = 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
bzero(buf, sizeof(buf));
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
/*
* Search for the requested audio device (channel). Start by
* iterating over pcm devices.
*/
Insert NULL pointer checking around devclass_get_maxunit(pcm_devclass, ..) . Things can get ugly without it due to uninitialized class. RELENG_6 need a simmilar, but different treatment as well. err.. perhaps we should teach devclass_get_maxunit() to return -1 ? MFC after: 1 day
2007-06-03 10:56:22 +00:00
for (i = 0; pcm_devclass != NULL &&
i < devclass_get_maxunit(pcm_devclass); i++) {
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
d = devclass_get_softc(pcm_devclass, i);
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 (!PCM_REGISTERED(d))
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
continue;
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 Need Giant magic entry ??? */
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
/* See the note in function docblock */
mtx_assert(d->lock, MA_NOTOWNED);
pcm_lock(d);
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, d, channels.pcm) {
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
mtx_assert(ch->lock, MA_NOTOWNED);
CHN_LOCK(ch);
if (ai->dev == -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 (DSP_REGISTERED(d, i_dev) &&
(ch == PCM_RDCH(i_dev) || /* record ch */
ch == PCM_WRCH(i_dev))) { /* playback ch */
devname = dsp_unit2name(buf,
sizeof(buf), ch->unit);
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
}
} else if (ai->dev == nchan) {
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
devname = dsp_unit2name(buf, sizeof(buf),
ch->unit);
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
if (devname != NULL)
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
CHN_UNLOCK(ch);
++nchan;
}
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 (devname != NULL) {
/*
* At this point, the following synchronization stuff
* has happened:
* - a specific PCM device is locked.
* - a specific audio channel has been locked, so be
* sure to unlock when exiting;
*/
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
caps = chn_getcaps(ch);
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
/*
* With all handles collected, zero out the user's
* container and begin filling in its fields.
*/
bzero((void *)ai, sizeof(oss_audioinfo));
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
ai->dev = nchan;
strlcpy(ai->name, ch->name, sizeof(ai->name));
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
if ((ch->flags & CHN_F_BUSY) == 0)
ai->busy = 0;
else
ai->busy = (ch->direction == PCMDIR_PLAY) ? OPEN_WRITE : OPEN_READ;
/**
* @note
* @c cmd - OSSv4 docs: "Only supported under Linux at
* this moment." Cop-out, I know, but I'll save
* running around in the process table for later.
* Is there a risk of leaking information?
*/
ai->pid = ch->pid;
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
/*
* These flags stolen from SNDCTL_DSP_GETCAPS handler.
* Note, however, that a single channel operates in
* only one direction, so DSP_CAP_DUPLEX is out.
*/
/**
* @todo @c SNDCTL_AUDIOINFO::caps - Make drivers keep
* these in pcmchan::caps?
*/
ai->caps = DSP_CAP_REALTIME | DSP_CAP_MMAP | DSP_CAP_TRIGGER;
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
/*
* Collect formats supported @b natively by the
* device. Also determine min/max channels. (I.e.,
* mono, stereo, or both?)
*
* If any channel is stereo, maxch = 2;
* if all channels are stereo, minch = 2, too;
* if any channel is mono, minch = 1;
* and if all channels are mono, maxch = 1.
*/
minch = 0;
maxch = 0;
fmts = 0;
for (i = 0; caps->fmtlist[i]; i++) {
fmts |= caps->fmtlist[i];
if (caps->fmtlist[i] & AFMT_STEREO) {
minch = (minch == 0) ? 2 : minch;
maxch = 2;
} else {
minch = 1;
maxch = (maxch == 0) ? 1 : maxch;
}
}
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 (ch->direction == PCMDIR_PLAY)
ai->oformats = fmts;
else
ai->iformats = fmts;
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
/**
* @note
* @c magic - OSSv4 docs: "Reserved for internal use
* by OSS."
*
* @par
* @c card_number - OSSv4 docs: "Number of the sound
* card where this device belongs or -1 if this
* information is not available. Applications
* should normally not use this field for any
* purpose."
*/
ai->card_number = -1;
/**
* @todo @c song_name - depends first on
* SNDCTL_[GS]ETSONG @todo @c label - depends
* on SNDCTL_[GS]ETLABEL
* @todo @c port_number - routing information?
*/
ai->port_number = -1;
ai->mixer_dev = (d->mixer_dev != NULL) ? PCMUNIT(d->mixer_dev) : -1;
/**
* @note
* @c real_device - OSSv4 docs: "Obsolete."
*/
ai->real_device = -1;
strlcpy(ai->devnode, devname, sizeof(ai->devnode));
ai->enabled = device_is_attached(d->dev) ? 1 : 0;
/**
* @note
* @c flags - OSSv4 docs: "Reserved for future use."
*
* @note
* @c binding - OSSv4 docs: "Reserved for future use."
*
* @todo @c handle - haven't decided how to generate
* this yet; bus, vendor, device IDs?
*/
ai->min_rate = caps->minspeed;
ai->max_rate = caps->maxspeed;
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
ai->min_channels = minch;
ai->max_channels = maxch;
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
ai->nrates = chn_getrates(ch, &rates);
if (ai->nrates > OSS_MAX_SAMPLE_RATES)
ai->nrates = OSS_MAX_SAMPLE_RATES;
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
for (i = 0; i < ai->nrates; i++)
ai->rates[i] = rates[i];
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
CHN_UNLOCK(ch);
}
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
pcm_unlock(d);
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
if (devname != NULL)
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
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
/* Exhausted the search -- nothing is locked, so return. */
return (EINVAL);
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 Assigns a PCM channel to a sync group.
*
* Sync groups are used to enable audio operations on multiple devices
* simultaneously. They may be used with any number of devices and may
* span across applications. Devices are added to groups with
* the SNDCTL_DSP_SYNCGROUP ioctl, and operations are triggered with the
* SNDCTL_DSP_SYNCSTART ioctl.
*
* If the @c id field of the @c group parameter is set to zero, then a new
* sync group is created. Otherwise, wrch and rdch (if set) are added to
* the group specified.
*
* @todo As far as memory allocation, should we assume that things are
* okay and allocate with M_WAITOK before acquiring channel locks,
* freeing later if not?
*
* @param wrch output channel associated w/ device (if any)
* @param rdch input channel associated w/ device (if any)
* @param group Sync group parameters
*
* @retval 0 success
* @retval non-zero error to be propagated upstream
*/
static int
dsp_oss_syncgroup(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_syncgroup *group)
{
struct pcmchan_syncmember *smrd, *smwr;
struct pcmchan_syncgroup *sg;
int ret, sg_ids[3];
smrd = NULL;
smwr = NULL;
sg = NULL;
ret = 0;
/*
* Free_unr() may sleep, so store released syncgroup IDs until after
* all locks are released.
*/
sg_ids[0] = sg_ids[1] = sg_ids[2] = 0;
PCM_SG_LOCK();
/*
* - Insert channel(s) into group's member list.
* - Set CHN_F_NOTRIGGER on channel(s).
* - Stop channel(s).
*/
/*
* If device's channels are already mapped to a group, unmap them.
*/
if (wrch) {
CHN_LOCK(wrch);
sg_ids[0] = chn_syncdestroy(wrch);
}
if (rdch) {
CHN_LOCK(rdch);
sg_ids[1] = chn_syncdestroy(rdch);
}
/*
* Verify that mode matches character device properites.
* - Bail if PCM_ENABLE_OUTPUT && wrch == NULL.
* - Bail if PCM_ENABLE_INPUT && rdch == NULL.
*/
if (((wrch == NULL) && (group->mode & PCM_ENABLE_OUTPUT)) ||
((rdch == NULL) && (group->mode & PCM_ENABLE_INPUT))) {
ret = EINVAL;
goto out;
}
/*
* An id of zero indicates the user wants to create a new
* syncgroup.
*/
if (group->id == 0) {
sg = (struct pcmchan_syncgroup *)malloc(sizeof(*sg), M_DEVBUF, M_NOWAIT);
if (sg != NULL) {
SLIST_INIT(&sg->members);
sg->id = alloc_unr(pcmsg_unrhdr);
group->id = sg->id;
SLIST_INSERT_HEAD(&snd_pcm_syncgroups, sg, link);
} else
ret = ENOMEM;
} else {
SLIST_FOREACH(sg, &snd_pcm_syncgroups, link) {
if (sg->id == group->id)
break;
}
if (sg == NULL)
ret = EINVAL;
}
/* Couldn't create or find a syncgroup. Fail. */
if (sg == NULL)
goto out;
/*
* Allocate a syncmember, assign it and a channel together, and
* insert into syncgroup.
*/
if (group->mode & PCM_ENABLE_INPUT) {
smrd = (struct pcmchan_syncmember *)malloc(sizeof(*smrd), M_DEVBUF, M_NOWAIT);
if (smrd == NULL) {
ret = ENOMEM;
goto out;
}
SLIST_INSERT_HEAD(&sg->members, smrd, link);
smrd->parent = sg;
smrd->ch = rdch;
chn_abort(rdch);
rdch->flags |= CHN_F_NOTRIGGER;
rdch->sm = smrd;
}
if (group->mode & PCM_ENABLE_OUTPUT) {
smwr = (struct pcmchan_syncmember *)malloc(sizeof(*smwr), M_DEVBUF, M_NOWAIT);
if (smwr == NULL) {
ret = ENOMEM;
goto out;
}
SLIST_INSERT_HEAD(&sg->members, smwr, link);
smwr->parent = sg;
smwr->ch = wrch;
chn_abort(wrch);
wrch->flags |= CHN_F_NOTRIGGER;
wrch->sm = smwr;
}
out:
if (ret != 0) {
if (smrd != NULL)
free(smrd, M_DEVBUF);
if ((sg != NULL) && SLIST_EMPTY(&sg->members)) {
sg_ids[2] = sg->id;
SLIST_REMOVE(&snd_pcm_syncgroups, sg, pcmchan_syncgroup, link);
free(sg, M_DEVBUF);
}
if (wrch)
wrch->sm = NULL;
if (rdch)
rdch->sm = NULL;
}
if (wrch)
CHN_UNLOCK(wrch);
if (rdch)
CHN_UNLOCK(rdch);
PCM_SG_UNLOCK();
if (sg_ids[0])
free_unr(pcmsg_unrhdr, sg_ids[0]);
if (sg_ids[1])
free_unr(pcmsg_unrhdr, sg_ids[1]);
if (sg_ids[2])
free_unr(pcmsg_unrhdr, sg_ids[2]);
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);
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 Launch a sync group into action
*
* Sync groups are established via SNDCTL_DSP_SYNCGROUP. This function
* iterates over all members, triggering them along the way.
*
* @note Caller must not hold any channel locks.
*
* @param sg_id sync group identifier
*
* @retval 0 success
* @retval non-zero error worthy of propagating upstream to user
*/
static int
dsp_oss_syncstart(int sg_id)
{
struct pcmchan_syncmember *sm, *sm_tmp;
struct pcmchan_syncgroup *sg;
struct pcm_channel *c;
int ret, needlocks;
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
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
/* Get the synclists lock */
PCM_SG_LOCK();
do {
ret = 0;
needlocks = 0;
/* Search for syncgroup by ID */
SLIST_FOREACH(sg, &snd_pcm_syncgroups, link) {
if (sg->id == sg_id)
break;
}
/* Return EINVAL if not found */
if (sg == NULL) {
ret = EINVAL;
break;
}
/* Any removals resulting in an empty group should've handled this */
KASSERT(!SLIST_EMPTY(&sg->members), ("found empty syncgroup"));
/*
* Attempt to lock all member channels - if any are already
* locked, unlock those acquired, sleep for a bit, and try
* again.
*/
SLIST_FOREACH(sm, &sg->members, link) {
if (CHN_TRYLOCK(sm->ch) == 0) {
int timo = hz * 5/1000;
if (timo < 1)
timo = 1;
/* Release all locked channels so far, retry */
SLIST_FOREACH(sm_tmp, &sg->members, link) {
/* sm is the member already locked */
if (sm == sm_tmp)
break;
CHN_UNLOCK(sm_tmp->ch);
}
/** @todo Is PRIBIO correct/ */
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 = msleep(sm, &snd_pcm_syncgroups_mtx,
PRIBIO | PCATCH, "pcmsg", timo);
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
if (ret == EINTR || ret == ERESTART)
break;
needlocks = 1;
[stage: 7/9] EWOULDBLOCK -> EAGAIN.
2007-03-16 17:17:25 +00:00
ret = 0; /* Assumes ret == EAGAIN... */
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
}
}
} while (needlocks && ret == 0);
/* Proceed only if no errors encountered. */
if (ret == 0) {
/* Launch channels */
while((sm = SLIST_FIRST(&sg->members)) != NULL) {
SLIST_REMOVE_HEAD(&sg->members, link);
c = sm->ch;
c->sm = NULL;
chn_start(c, 1);
c->flags &= ~CHN_F_NOTRIGGER;
CHN_UNLOCK(c);
free(sm, M_DEVBUF);
}
SLIST_REMOVE(&snd_pcm_syncgroups, sg, pcmchan_syncgroup, link);
free(sg, M_DEVBUF);
}
PCM_SG_UNLOCK();
/*
* Free_unr() may sleep, so be sure to give up the syncgroup lock
* first.
*/
if (ret == 0)
free_unr(pcmsg_unrhdr, sg_id);
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);
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 Handler for SNDCTL_DSP_POLICY
*
* The SNDCTL_DSP_POLICY ioctl is a simpler interface to control fragment
* size and count like with SNDCTL_DSP_SETFRAGMENT. Instead of the user
* specifying those two parameters, s/he simply selects a number from 0..10
* which corresponds to a buffer size. Smaller numbers request smaller
* buffers with lower latencies (at greater overhead from more frequent
* interrupts), while greater numbers behave in the opposite manner.
*
* The 4Front spec states that a value of 5 should be the default. However,
* this implementation deviates slightly by using a linear scale without
* consulting drivers. I.e., even though drivers may have different default
* buffer sizes, a policy argument of 5 will have the same result across
* all drivers.
*
* See http://manuals.opensound.com/developer/SNDCTL_DSP_POLICY.html for
* more information.
*
* @todo When SNDCTL_DSP_COOKEDMODE is supported, it'll be necessary to
* work with hardware drivers directly.
*
* @note PCM channel arguments must not be locked by caller.
*
* @param wrch Pointer to opened playback channel (optional; may be NULL)
* @param rdch " recording channel (optional; may be NULL)
* @param policy Integer from [0:10]
*
* @retval 0 constant (for now)
*/
static int
dsp_oss_policy(struct pcm_channel *wrch, struct pcm_channel *rdch, int policy)
{
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;
if (policy < CHN_POLICY_MIN || policy > CHN_POLICY_MAX)
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 (EIO);
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
/* Default: success */
ret = 0;
if (rdch) {
CHN_LOCK(rdch);
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
ret = chn_setlatency(rdch, policy);
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
CHN_UNLOCK(rdch);
}
if (wrch && ret == 0) {
CHN_LOCK(wrch);
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
ret = chn_setlatency(wrch, policy);
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
CHN_UNLOCK(wrch);
}
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)
ret = EIO;
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);
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
/**
* @brief Enable or disable "cooked" mode
*
* This is a handler for @c SNDCTL_DSP_COOKEDMODE. When in cooked mode, which
* is the default, the sound system handles rate and format conversions
* automatically (ex: user writing 11025Hz/8 bit/unsigned but card only
* operates with 44100Hz/16bit/signed samples).
*
* Disabling cooked mode is intended for applications wanting to mmap()
* a sound card's buffer space directly, bypassing the FreeBSD 2-stage
* feeder architecture, presumably to gain as much control over audio
* hardware as possible.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_DSP_COOKEDMODE.html
* for more details.
*
* @note Currently, this function is just a stub that always returns EINVAL.
*
* @todo Figure out how to and actually implement this.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param enabled 0 = raw mode, 1 = cooked mode
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_cookedmode(struct pcm_channel *wrch, struct pcm_channel *rdch, int enabled)
{
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);
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 Retrieve channel interleaving order
*
* This is the handler for @c SNDCTL_DSP_GET_CHNORDER.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_DSP_GET_CHNORDER.html
* for more details.
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support SNDCTL_DSP_GET_CHNORDER.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param map channel map (result will be stored there)
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_getchnorder(struct pcm_channel *wrch, struct pcm_channel *rdch, unsigned long long *map)
{
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);
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 Specify channel interleaving order
*
* This is the handler for @c SNDCTL_DSP_SET_CHNORDER.
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support @c SNDCTL_DSP_SET_CHNORDER.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param map channel map
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_setchnorder(struct pcm_channel *wrch, struct pcm_channel *rdch, unsigned long long *map)
{
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);
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 Retrieve an audio device's label
*
* This is a handler for the @c SNDCTL_GETLABEL ioctl.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_GETLABEL.html
* for more details.
*
* From Hannu@4Front: "For example ossxmix (just like some HW mixer
* consoles) can show variable "labels" for certain controls. By default
* the application name (say quake) is shown as the label but
* applications may change the labels themselves."
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support @c SNDCTL_GETLABEL.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param label label gets copied here
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_getlabel(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_label_t *label)
{
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);
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 Specify an audio device's label
*
* This is a handler for the @c SNDCTL_SETLABEL ioctl. Please see the
* comments for @c dsp_oss_getlabel immediately above.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_GETLABEL.html
* for more details.
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support SNDCTL_SETLABEL.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param label label gets copied from here
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_setlabel(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_label_t *label)
{
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);
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 Retrieve name of currently played song
*
* This is a handler for the @c SNDCTL_GETSONG ioctl. Audio players could
* tell the system the name of the currently playing song, which would be
* visible in @c /dev/sndstat.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_GETSONG.html
* for more details.
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support SNDCTL_GETSONG.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param song song name gets copied here
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_getsong(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *song)
{
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);
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 Retrieve name of currently played song
*
* This is a handler for the @c SNDCTL_SETSONG ioctl. Audio players could
* tell the system the name of the currently playing song, which would be
* visible in @c /dev/sndstat.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_SETSONG.html
* for more details.
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support SNDCTL_SETSONG.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param song song name gets copied from here
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_setsong(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *song)
{
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);
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 Rename a device
*
* This is a handler for the @c SNDCTL_SETNAME ioctl.
*
* See @c http://manuals.opensound.com/developer/SNDCTL_SETNAME.html for
* more details.
*
* From Hannu@4Front: "This call is used to change the device name
* reported in /dev/sndstat and ossinfo. So instead of using some generic
* 'OSS loopback audio (MIDI) driver' the device may be given a meaningfull
* name depending on the current context (for example 'OSS virtual wave table
* synth' or 'VoIP link to London')."
*
* @note As the ioctl definition is still under construction, FreeBSD
* does not currently support SNDCTL_SETNAME.
*
* @param wrch playback channel (optional; may be NULL)
* @param rdch recording channel (optional; may be NULL)
* @param name new device name gets copied from here
*
* @retval EINVAL Operation not yet supported.
*/
static int
dsp_oss_setname(struct pcm_channel *wrch, struct pcm_channel *rdch, oss_longname_t *name)
{
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);
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
}
#endif /* !OSSV4_EXPERIMENT */
Reference in New Issue Copy Permalink