July 2023 - Alsa-devel - mailman.alsa-project.org

[syzbot] [alsa?] memory leak in snd_seq_create_port
by syzbot 17 Jul '23

17 Jul '23

Hello, syzbot found the following issue on: HEAD commit: 3f01e9fed845 Merge tag 'linux-watchdog-6.5-rc2' of git://w.. git tree: upstream console output: https://syzkaller.appspot.com/x/log.txt?x=14b07344a80000 kernel config: https://syzkaller.appspot.com/x/.config?x=75da4f0a455bdbd3 dashboard link: https://syzkaller.appspot.com/bug?extid=cf8e7fa4eeec59b3d485 compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2 syz repro: https://syzkaller.appspot.com/x/repro.syz?x=15877dc2a80000 C reproducer: https://syzkaller.appspot.com/x/repro.c?x=12905004a80000 Downloadable assets: disk image: https://storage.googleapis.com/syzbot-assets/441fb7ea58b8/disk-3f01e9fe.raw… vmlinux: https://storage.googleapis.com/syzbot-assets/8fa7790ba0c3/vmlinux-3f01e9fe.… kernel image: https://storage.googleapis.com/syzbot-assets/5e7a6471dadf/bzImage-3f01e9fe.… IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: syzbot+cf8e7fa4eeec59b3d485(a)syzkaller.appspotmail.com Warning: Permanently added '10.128.1.1' (ED25519) to the list of known hosts. executing program executing program BUG: memory leak unreferenced object 0xffff888100877000 (size 512): comm "syz-executor257", pid 5012, jiffies 4294941742 (age 12.790s) hex dump (first 32 bytes): 80 ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<ffffffff8154bf94>] kmalloc_trace+0x24/0x90 mm/slab_common.c:1076 [<ffffffff83d29e28>] kmalloc include/linux/slab.h:582 [inline] [<ffffffff83d29e28>] kzalloc include/linux/slab.h:703 [inline] [<ffffffff83d29e28>] snd_seq_create_port+0x78/0x300 sound/core/seq/seq_ports.c:135 [<ffffffff83d1f681>] snd_seq_ioctl_create_port+0xe1/0x2a0 sound/core/seq/seq_clientmgr.c:1324 [<ffffffff83d20e5e>] snd_seq_ioctl+0x13e/0x290 sound/core/seq/seq_clientmgr.c:2327 [<ffffffff81685173>] vfs_ioctl fs/ioctl.c:51 [inline] [<ffffffff81685173>] __do_sys_ioctl fs/ioctl.c:870 [inline] [<ffffffff81685173>] __se_sys_ioctl fs/ioctl.c:856 [inline] [<ffffffff81685173>] __x64_sys_ioctl+0x103/0x140 fs/ioctl.c:856 [<ffffffff84a77ff9>] do_syscall_x64 arch/x86/entry/common.c:50 [inline] [<ffffffff84a77ff9>] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 [<ffffffff84c0008b>] entry_SYSCALL_64_after_hwframe+0x63/0xcd BUG: memory leak unreferenced object 0xffff888106742c00 (size 512): comm "syz-executor257", pid 5013, jiffies 4294942276 (age 7.450s) hex dump (first 32 bytes): 80 ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<ffffffff8154bf94>] kmalloc_trace+0x24/0x90 mm/slab_common.c:1076 [<ffffffff83d29e28>] kmalloc include/linux/slab.h:582 [inline] [<ffffffff83d29e28>] kzalloc include/linux/slab.h:703 [inline] [<ffffffff83d29e28>] snd_seq_create_port+0x78/0x300 sound/core/seq/seq_ports.c:135 [<ffffffff83d1f681>] snd_seq_ioctl_create_port+0xe1/0x2a0 sound/core/seq/seq_clientmgr.c:1324 [<ffffffff83d20e5e>] snd_seq_ioctl+0x13e/0x290 sound/core/seq/seq_clientmgr.c:2327 [<ffffffff81685173>] vfs_ioctl fs/ioctl.c:51 [inline] [<ffffffff81685173>] __do_sys_ioctl fs/ioctl.c:870 [inline] [<ffffffff81685173>] __se_sys_ioctl fs/ioctl.c:856 [inline] [<ffffffff81685173>] __x64_sys_ioctl+0x103/0x140 fs/ioctl.c:856 [<ffffffff84a77ff9>] do_syscall_x64 arch/x86/entry/common.c:50 [inline] [<ffffffff84a77ff9>] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 [<ffffffff84c0008b>] entry_SYSCALL_64_after_hwframe+0x63/0xcd --- This report is generated by a bot. It may contain errors. See https://goo.gl/tpsmEJ for more information about syzbot. syzbot engineers can be reached at syzkaller(a)googlegroups.com. syzbot will keep track of this issue. See: https://goo.gl/tpsmEJ#status for how to communicate with syzbot. If the bug is already fixed, let syzbot know by replying with: #syz fix: exact-commit-title If you want syzbot to run the reproducer, reply with: #syz test: git://repo/address.git branch-or-commit-hash If you attach or paste a git patch, syzbot will apply it before testing. If you want to change bug's subsystems, reply with: #syz set subsystems: new-subsystem (See the list of subsystem names on the web dashboard) If the bug is a duplicate of another bug, reply with: #syz dup: exact-subject-of-another-report If you want to undo deduplication, reply with: #syz undup

4 7

[PATCH] ASoC: SOF: ipc3-dtrace: uninitialized data in dfsentry_trace_filter_write()
by Dan Carpenter 17 Jul '23

17 Jul '23

This doesn't check how many bytes the simple_write_to_buffer() writes to the buffer. The only thing that we know is that the first byte is initialized and the last byte of the buffer is set to NUL. However the middle bytes could be uninitialized. There is no need to use simple_write_to_buffer(). This code does not support partial writes but instead passes "pos = 0" as the starting offset regardless of what the user passed as "*ppos". Just use the copy_from_user() function and initialize the whole buffer. Fixes: 671e0b90051e ("ASoC: SOF: Clone the trace code to ipc3-dtrace as fw_tracing implementation") Signed-off-by: Dan Carpenter <dan.carpenter(a)linaro.org> --- The nicer thing to do would have been to use strndup_user() but that comes with a small risk of breaking user space and I can't test this so I didn't do that. sound/soc/sof/ipc3-dtrace.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/sound/soc/sof/ipc3-dtrace.c b/sound/soc/sof/ipc3-dtrace.c index 1d3bca2d28dd..35da85a45a9a 100644 --- a/sound/soc/sof/ipc3-dtrace.c +++ b/sound/soc/sof/ipc3-dtrace.c @@ -186,7 +186,6 @@ static ssize_t dfsentry_trace_filter_write(struct file *file, const char __user struct snd_sof_dfsentry *dfse = file->private_data; struct sof_ipc_trace_filter_elem *elems = NULL; struct snd_sof_dev *sdev = dfse->sdev; - loff_t pos = 0; int num_elems; char *string; int ret; @@ -201,11 +200,11 @@ static ssize_t dfsentry_trace_filter_write(struct file *file, const char __user if (!string) return -ENOMEM; - /* assert null termination */ - string[count] = 0; - ret = simple_write_to_buffer(string, count, &pos, from, count); - if (ret < 0) + if (copy_from_user(string, from, count)) { + ret = -EFAULT; goto error; + } + string[count] = '\0'; ret = trace_filter_parse(sdev, string, &num_elems, &elems); if (ret < 0) -- 2.39.2

2 1

[PATCH] ASoC: cs42l51: fix driver to properly autoload with automatic module loading
by Thomas Petazzoni 17 Jul '23

17 Jul '23

In commit 2cb1e0259f50 ("ASoC: cs42l51: re-hook of_match_table pointer"), 9 years ago, some random guy fixed the cs42l51 after it was split into a core part and an I2C part to properly match based on a Device Tree compatible string. However, the fix in this commit is wrong: the MODULE_DEVICE_TABLE(of, ....) is in the core part of the driver, not the I2C part. Therefore, automatic module loading based on module.alias, based on matching with the DT compatible string, loads the core part of the driver, but not the I2C part. And threfore, the i2c_driver is not registered, and the codec is not known to the system, nor matched with a DT node with the corresponding compatible string. In order to fix that, we move the MODULE_DEVICE_TABLE(of, ...) into the I2C part of the driver. The cs42l51_of_match[] array is also moved as well, as it is not possible to have this definition in one file, and the MODULE_DEVICE_TABLE(of, ...) invocation in another file, due to how MODULE_DEVICE_TABLE works. Thanks to this commit, the I2C part of the driver now properly autoloads, and thanks to its dependency on the core part, the core part gets autoloaded as well, resulting in a functional sound card without having to manually load kernel modules. Fixes: 2cb1e0259f50 ("ASoC: cs42l51: re-hook of_match_table pointer") Cc: stable(a)vger.kernel.org Signed-off-by: Thomas Petazzoni <thomas.petazzoni(a)bootlin.com> --- sound/soc/codecs/cs42l51-i2c.c | 6 ++++++ sound/soc/codecs/cs42l51.c | 7 ------- sound/soc/codecs/cs42l51.h | 1 - 3 files changed, 6 insertions(+), 8 deletions(-) diff --git a/sound/soc/codecs/cs42l51-i2c.c b/sound/soc/codecs/cs42l51-i2c.c index b2106ff6a7cb..e7db7bcd0296 100644 --- a/sound/soc/codecs/cs42l51-i2c.c +++ b/sound/soc/codecs/cs42l51-i2c.c @@ -19,6 +19,12 @@ static struct i2c_device_id cs42l51_i2c_id[] = { }; MODULE_DEVICE_TABLE(i2c, cs42l51_i2c_id); +const struct of_device_id cs42l51_of_match[] = { + { .compatible = "cirrus,cs42l51", }, + { } +}; +MODULE_DEVICE_TABLE(of, cs42l51_of_match); + static int cs42l51_i2c_probe(struct i2c_client *i2c) { struct regmap_config config; diff --git a/sound/soc/codecs/cs42l51.c b/sound/soc/codecs/cs42l51.c index a67cd3ee84e0..a7079ae0ca09 100644 --- a/sound/soc/codecs/cs42l51.c +++ b/sound/soc/codecs/cs42l51.c @@ -823,13 +823,6 @@ int __maybe_unused cs42l51_resume(struct device *dev) } EXPORT_SYMBOL_GPL(cs42l51_resume); -const struct of_device_id cs42l51_of_match[] = { - { .compatible = "cirrus,cs42l51", }, - { } -}; -MODULE_DEVICE_TABLE(of, cs42l51_of_match); -EXPORT_SYMBOL_GPL(cs42l51_of_match); - MODULE_AUTHOR("Arnaud Patard <arnaud.patard(a)rtp-net.org>"); MODULE_DESCRIPTION("Cirrus Logic CS42L51 ALSA SoC Codec Driver"); MODULE_LICENSE("GPL"); diff --git a/sound/soc/codecs/cs42l51.h b/sound/soc/codecs/cs42l51.h index a79343e8a54e..125703ede113 100644 --- a/sound/soc/codecs/cs42l51.h +++ b/sound/soc/codecs/cs42l51.h @@ -16,7 +16,6 @@ int cs42l51_probe(struct device *dev, struct regmap *regmap); void cs42l51_remove(struct device *dev); int __maybe_unused cs42l51_suspend(struct device *dev); int __maybe_unused cs42l51_resume(struct device *dev); -extern const struct of_device_id cs42l51_of_match[]; #define CS42L51_CHIP_ID 0x1B #define CS42L51_CHIP_REV_A 0x00 -- 2.41.0

2 1

[PATCH v2 0/8] ALSA: emu10k1: add support for high-bitrate modes of E-MU cards
by Oswald Buddenhagen 17 Jul '23

17 Jul '23

This series is what all the work was about: support the "dual-/quad-pumped" modes of the E-MU cards. Oswald Buddenhagen (8): ALSA: emu10k1: introduce alternative E-MU D.A.S. mode ALSA: emu10k1: improve mixer control naming in E-MU D.A.S. mode ALSA: emu10k1: set the "no filtering" bits on PCM voices ALSA: emu10k1: make playback in E-MU D.A.S. mode 32-bit ALSA: add snd_ctl_add_locked() ALSA: emu10k1: add support for 2x/4x word clocks in E-MU D.A.S. mode ALSA: emu10k1: add high-rate capture in E-MU D.A.S. mode ALSA: emu10k1: add high-rate playback in E-MU D.A.S. mode include/sound/control.h | 1 + include/sound/emu10k1.h | 11 + sound/core/control.c | 31 ++ sound/pci/emu10k1/emu10k1.c | 29 +- sound/pci/emu10k1/emu10k1_main.c | 19 +- sound/pci/emu10k1/emufx.c | 109 +++- sound/pci/emu10k1/emumixer.c | 901 +++++++++++++++++++++++++++---- sound/pci/emu10k1/emupcm.c | 422 +++++++++++++-- sound/pci/emu10k1/emuproc.c | 5 + sound/pci/emu10k1/io.c | 30 +- sound/pci/emu10k1/voice.c | 6 + 11 files changed, 1383 insertions(+), 181 deletions(-) Range-diff against v1: 1: c8c1fbba5e52 = 1: 7cb49147164f ALSA: emu10k1: introduce alternative E-MU D.A.S. mode 2: 879066428bee = 2: aa9ead00e045 ALSA: emu10k1: improve mixer control naming in E-MU D.A.S. mode 3: b663229a0f46 = 3: ac830e67322f ALSA: emu10k1: set the "no filtering" bits on PCM voices 4: a9de9a73571e = 4: 26e0b7c02c1d ALSA: emu10k1: make playback in E-MU D.A.S. mode 32-bit 5: 0f1950b03193 ! 5: 94b64e6c123d ALSA: add snd_ctl_add_locked() @@ Commit message This is in fact more symmetrical to snd_ctl_remove() than snd_ctl_add() is - the former could be named snd_ctl_remove_locked() just as well. + One may argue that this is going in the wrong direction, as drivers have + no business managing the lock. This may be true in principle, but in + practice the vast majority of controls is created even before the device + was registered, and therefore before any locking is necessary at all. + That means that an even more radical approach of changing snd_ctl_add() + do do no locking, and converting the call sites that actually need + locking to a new function, would better match reality, and would be + somewhat more efficient at that. However, that seems a bit risky and way + too much work. + This will be used to dynamically change the available controls from another control's put() callback, which is already locked. One might want to add snd_ctl_replace_locked() for completeness, but I have no use for it now. Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> + --- + v2: + - extended commit message ## include/sound/control.h ## @@ include/sound/control.h: void snd_ctl_notify_one(struct snd_card * card, unsigned int mask, struct snd_kc 6: 511efe4ac7ad ! 6: 7bc314bae7f2 ALSA: emu10k1: add support for 2x/4x word clocks in E-MU D.A.S. mode @@ Commit message on playback, and throw away the excess ones on capture. Input-to-output monitoring does actually use the full sample rate, though. - Notably, add_ctls() now uses snd_ctl_add_locked(), so it doesn't - deadlock when called from snd_emu1010_clock_shift_put(). This also - affects the initial creation of the controls, which is OK, as that is - done before the card is registered, so no concurrent access can occur. + Due to hardware constraints, changing the clock multiplier (CM) changes + the available audio ports and the number of available channels. This has + an impact on the channel routing mixer controls. One way to deal with + this would be presenting a union of all possibilities, and simply + ignoring currently inapplicable settings. However, this would be a + terrible user experience, and go against the spirit of prior patches + aimed at decluttering the mixer. Therefore, we do dynamic + reconfiguration (DR) of the mixer in response to changing the CM. + + DR is somewhat controversial, as it has the potential to crash poorly + programmed applications. But that in itself isn't a very convincing + argument against it, as by that logic we'd have to ban all hot-plugging. + Such crashes would also not really qualify as regressions, as the D.A.S. + mode is a new opt-in feature, and therefore no previously stable setups + would be impacted. Also, pendantically, the driver already had DR via + SNDRV_EMU10K1_IOCTL_CODE_POKE. A somewhat valid concern is that changing + mixer settings is a non-privileged operation and therefore potential + crashes could be exploited for a somewhat more impactful nuisance attack + on another user than messing with the mixer per se. However, systemd & + co. limit device access to the user currently logged in on the seat + owning the device. + + There is a specific concern about doing DR in response to changing a + mixer control's value, as an application may legitimately react to DR by + updating all mixer settings in turn. However, that update should write + the same value to the clock multiplier, thus terminating the recursion. + + One may limit DR to merely disabling inapplicable controls, in the hope + that this would be better handled than completely tearing down and + re-creating controls as we do. However, there is no guarantee for that. + And because it is impossible to disable particular enum values within a + control, it would be necessary to have three complete sets of + per-channel controls. This would yield an extremely cluttered and + confusing UI if the application (reasonably) chose to merely visually + disable inactive controls rather than hiding them. + + We do the DR synchronously from within snd_emu1010_clock_shift_put(). + This was enabled by commit 5bbb1ab5bd ("control: use counting semaphore + as write lock for ELEM_WRITE operation"); we merely need to make + add_ctls() use snd_ctl_add_locked() instead of snd_ctl_add(), so it + doesn't deadlock. That also affects the initial creation of the + controls, which is OK, as that is done before the card is registered, so + no concurrent access can occur. + + It would be possible to do the DR in a tasklet after the ioctl finishes. + However, it is not obvious what actual problem that would solve, and the + added asynchronicity would significantly complicate matters, esp. wrt. + the batch updates expected during mixer state restoration. Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> + --- + v2: + - expanded commit message ## include/sound/emu10k1.h ## @@ include/sound/emu10k1.h: struct snd_emu1010 { 7: d5cb50ca707f = 7: 72a156fb32cd ALSA: emu10k1: add high-rate capture in E-MU D.A.S. mode 8: 319425a4ccb6 ! 8: 6d35891832b3 ALSA: emu10k1: add high-rate playback in E-MU D.A.S. mode @@ Commit message d948035a92 ("Remove PCM xfer_align sw params"). Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> + --- + v2: + - fixed `sparse` warning re missing __user annotation ## sound/pci/emu10k1/emufx.c ## @@ sound/pci/emu10k1/emufx.c: static int _snd_emu10k1_das_init_efx(struct snd_emu10k1 *emu) @@ sound/pci/emu10k1/emupcm.c: static int snd_emu10k1_efx_playback_trigger(struct s + for (i = 0; i < channels; i++) { + for (j = 0; j < subchans; j++) { + u32 *dst = get_dma_ptr_x(runtime, shift, i, j, hwoff); -+ u32 *src = (u32 *)buf + j * channels + i; ++ u32 __user *src = (u32 __user *)buf + j * channels + i; + for (k = 0; k < frames; k++, dst++, src += voices) + unsafe_get_user(*dst, src, faulted); + } -- 2.40.0.152.g15d061e6df

2 14

[PATCH 0/2] Fix CSC3551 speaker sound problem for machines without a valid ACPI _DSD
by David Xu 17 Jul '23

17 Jul '23

As the comments added in commit 4d4c4bff4f8ed79d95e05 ("ALSA: hda: cs35l41: Clarify support for CSC3551 without _DSD Properties"), CSC3551 requires a valid _DSD to work and the current implementation just fails when no _DSD can be found for CSC3551. However it is a fact that many OEMs hardcoded the configurations needed by CSC3551 into their proprietary software for various 2022 and later laptop models, and this makes the Linux installations on these models cannot make any speaker sound. Meanwhile, at this point of time, we see no hope that these OEMs would ever fix this problem via a BIOS update. To address the problem, this patch series contains two patches: Patch 1 for cs35l41 hda driver: a fixup mechanism is introduced that when the driver found there is no valid _DSD that contains the configurations, a fixup function would try to find a fixup entry that contains a proper cs35l41 configuration from a pre-defined fixup table by matching the CSC3551 ACPI _SUB id. If found, the fixup function would apply the cs35l41 configurations retrived from the entry. In this patch the fixup table only contains some entries for three Lenovo laptop models: namely 16IAH7, 16IAX7 and 16ARHA7. However as is known, several other laptop models from ASUS and HP also suffer from this no valid _DSD problem and could have it addressed with this fixup mechanism when proper fixup entries are inserted. Patch 2 for realtek hda driver: add quirks for Lenovo 16IAH7, 16IAX7 and 16ARHA7 so that cs35l41 playback hook could be registered. Please note that for these quirks to work patch 1 has to be applied. David Xu (2): ALSA: hda: cs35l41: Add fixups for machines without a valid ACPI _DSD ALSA: hda/realtek: Add quirks for Lenovo 16IAH7, 16IAX7 and 16ARHA7 sound/pci/hda/cs35l41_hda.c | 160 ++++++++++++++++++++++++++++++++++ sound/pci/hda/patch_realtek.c | 5 ++ 2 files changed, 165 insertions(+) -- 2.41.0

3 4

[PATCH] ALSA: emu10k1: do away with extra voices in E-MU D.A.S. mode
by Oswald Buddenhagen 17 Jul '23

17 Jul '23

Maybe somewhat unsurprisingly, Creative Inc. learned from the period interrupt disaster of the EMU10K1, and equipped Audigy with a way to delay the interrupts by a configurable amount. So make use of that. This allows us to do away with the extra voices for driving interrupts. The most salient advantage of that is that we can maximize the number of playback channels in 24-bit 192 kHz mode, bringing it up to 8. A minor point is that no bandwidth is wasted on fetching data that is discarded anyway. The removal of the extra voices has one disadvantage: the hardware is limited to two periods, thus increasing the likelihood of underruns. Therefore, we do this only in D.A.S. mode, which aims at high-quality low-latency multi-channel recording, and therefore can be expected to be operated in an optimal environment. Also, prior to a few commits back, the multi-channel playback was actually limited to two periods anyway (for no good reason). Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> --- this is obviously meaningless without the d.a.s. series. i'm posting it only so it's on record. --- This patch needs to be applied on top of the patch titled "ALSA: emu10k1: add high-rate playback in E-MU D.A.S. mode". --- include/sound/emu10k1.h | 11 +++++++ sound/pci/emu10k1/emu10k1_main.c | 16 ++++++++++ sound/pci/emu10k1/emupcm.c | 52 +++++++++++++++++++++++--------- sound/pci/emu10k1/irq.c | 8 +++++ 4 files changed, 73 insertions(+), 14 deletions(-) diff --git a/include/sound/emu10k1.h b/include/sound/emu10k1.h index fd4cf7d6eb3f..0931b1e65e56 100644 --- a/include/sound/emu10k1.h +++ b/include/sound/emu10k1.h @@ -838,7 +838,18 @@ SUB_REG(FXIDX, IDX, 0x0000ffff) #define A_SPRA 0x6b /* S/PDIF Host Record Address */ #define A_SPRC 0x6c /* S/PDIF Host Record Control */ +// NOTE: 64-bit +// Voice 0 holds the counter, while voice 1 holds the enable bits. +// Bits in IPR are set according to incoming IRQs as usual. +// When a corresponding bit is set in the delay enable register, +// the counter starts, and the IRQ isn't delivered until timeout. +// Additional delayed IRQs arriving during the countdown do not +// retrigger it, so they will be delayed less or not at all. +// Non-delayed IRQs arriving during the countdown are delivered +// immediately, while the countdown continues. #define A_DICE 0x6d /* Delayed Interrupt Counter & Enable */ +#define A_DICE_DELAY_MASK 0x000003ff /* Interrupt delay minus 1 */ +#define A_DICE_COUNTER_MASK 0x03ff0000 /* Counter's current value */ #define A_TTB 0x6e /* Tank Table Base */ #define A_TDOF 0x6f /* Tank Delay Offset */ diff --git a/sound/pci/emu10k1/emu10k1_main.c b/sound/pci/emu10k1/emu10k1_main.c index 2125925c2d5e..97a4be6eaf7a 100644 --- a/sound/pci/emu10k1/emu10k1_main.c +++ b/sound/pci/emu10k1/emu10k1_main.c @@ -159,6 +159,8 @@ static int snd_emu10k1_init(struct snd_emu10k1 *emu, int enable_ir) /* disable channel interrupt */ CLIEL, 0, CLIEH, 0, + HLIEL, 0, + HLIEH, 0, /* disable stop on loop end */ SOLEL, 0, @@ -174,6 +176,12 @@ static int snd_emu10k1_init(struct snd_emu10k1 *emu, int enable_ir) AC97SLOT_REAR_LEFT); } + if (emu->das_mode) { + // Enable delayed period interrupts + snd_emu10k1_ptr_write(emu, A_DICE, 0, 64 - 3 + 1); + snd_emu10k1_ptr_write(emu, A_DICE, 1, IPR_CHANNELLOOP); + } + /* init envelope engine */ for (ch = 0; ch < NUM_G; ch++) snd_emu10k1_voice_init(emu, ch); @@ -414,6 +422,12 @@ int snd_emu10k1_done(struct snd_emu10k1 *emu) else snd_emu10k1_ptr_write(emu, DBG, 0, EMU10K1_DBG_SINGLE_STEP); + if (emu->das_mode) { + // Disable delayed period interrupts + snd_emu10k1_ptr_write(emu, A_DICE, 0, 0); + snd_emu10k1_ptr_write(emu, A_DICE, 1, 0); + } + snd_emu10k1_ptr_write_multiple(emu, 0, /* reset recording buffers */ MICBS, 0, @@ -429,6 +443,8 @@ int snd_emu10k1_done(struct snd_emu10k1 *emu) /* disable channel interrupt */ CLIEL, 0, CLIEH, 0, + HLIEL, 0, + HLIEH, 0, SOLEL, 0, SOLEH, 0, diff --git a/sound/pci/emu10k1/emupcm.c b/sound/pci/emu10k1/emupcm.c index ccb0e13da3dd..b7841076f019 100644 --- a/sound/pci/emu10k1/emupcm.c +++ b/sound/pci/emu10k1/emupcm.c @@ -96,7 +96,9 @@ static int snd_emu10k1_pcm_channel_alloc(struct snd_emu10k1_pcm *epcm, if (err < 0) return err; - if (epcm->extra == NULL) { + if (epcm->emu->das_mode) { + epcm->voices[0]->interrupt = snd_emu10k1_pcm_interrupt; + } else if (epcm->extra == NULL) { // The hardware supports only (half-)loop interrupts, so to support an // arbitrary number of periods per buffer, we use an extra voice with a // period-sized loop as the interrupt source. Additionally, the interrupt @@ -500,13 +502,8 @@ static int snd_emu10k1_efx_playback_prepare(struct snd_pcm_substream *substream) epcm->pitch_target = PITCH_48000; start_addr = epcm->start_addr >> 1; // 16-bit voices - - extra_size = runtime->period_size >> shift; channel_size = runtime->buffer_size >> shift; - snd_emu10k1_pcm_init_extra_voice(emu, epcm->extra, true, - start_addr, start_addr + extra_size); - if (das_mode) { unsigned count = 1 << shift; start_addr >>= 1; @@ -520,6 +517,10 @@ static int snd_emu10k1_efx_playback_prepare(struct snd_pcm_substream *substream) } } } else { + extra_size = runtime->period_size >> shift; + snd_emu10k1_pcm_init_extra_voice(emu, epcm->extra, true, + start_addr, start_addr + extra_size); + epcm->ccca_start_addr = start_addr; for (i = 0; i < runtime->channels; i++) { snd_emu10k1_pcm_init_voices(emu, epcm->voices[i], true, false, @@ -547,7 +548,7 @@ static const struct snd_pcm_hardware snd_emu10k1_efx_playback = .buffer_bytes_max = (128*1024), .period_bytes_max = (128*1024), .periods_min = 2, - .periods_max = 1024, + .periods_max = 2, .fifo_size = 0, }; @@ -698,8 +699,16 @@ static void snd_emu10k1_playback_prepare_voices(struct snd_emu10k1 *emu, // This is why all other (open) drivers for these chips use timer-based // interrupts. // - eloop_start += (epcm->resume_pos + eloop_size - 3) % eloop_size; - snd_emu10k1_ptr_write(emu, CCCA_CURRADDR, epcm->extra->number, eloop_start); + // But then, Audigy introduced delayed interrupt functionality, so we + // use it in D.A.S. mode, where we may need all voices. We can't use + // it in regular mode due to supporting multiple substreams, which the + // delayed IRQs cannot really handle; also, we prefer supporting an + // arbitrary number of periods there, for which we need the extra voice. + // + if (!emu->das_mode) { + eloop_start += (epcm->resume_pos + eloop_size - 3) % eloop_size; + snd_emu10k1_ptr_write(emu, CCCA_CURRADDR, epcm->extra->number, eloop_start); + } // It takes a moment until the cache fills complete, // but the unmuting takes long enough for that. @@ -793,13 +802,25 @@ static void snd_emu10k1_playback_set_running(struct snd_emu10k1 *emu, struct snd_emu10k1_pcm *epcm) { epcm->running = 1; - snd_emu10k1_voice_intr_enable(emu, epcm->extra->number); + if (emu->das_mode) { + unsigned int voice = epcm->voices[0]->number; + snd_emu10k1_voice_half_loop_intr_enable(emu, voice); + snd_emu10k1_voice_intr_enable(emu, voice); + } else { + snd_emu10k1_voice_intr_enable(emu, epcm->extra->number); + } } static void snd_emu10k1_playback_set_stopped(struct snd_emu10k1 *emu, struct snd_emu10k1_pcm *epcm) { - snd_emu10k1_voice_intr_disable(emu, epcm->extra->number); + if (emu->das_mode) { + unsigned int voice = epcm->voices[0]->number; + snd_emu10k1_voice_half_loop_intr_disable(emu, voice); + snd_emu10k1_voice_intr_disable(emu, voice); + } else { + snd_emu10k1_voice_intr_disable(emu, epcm->extra->number); + } epcm->running = 0; } @@ -1064,7 +1085,8 @@ static int snd_emu10k1_efx_playback_trigger(struct snd_pcm_substream *substream, result = snd_emu10k1_voice_clear_loop_stop_multiple_atomic(emu, mask); if (result == 0) { // The extra voice is allowed to lag a bit - snd_emu10k1_playback_trigger_voice(emu, epcm->extra); + if (!emu->das_mode) + snd_emu10k1_playback_trigger_voice(emu, epcm->extra); goto leave; } @@ -1080,7 +1102,8 @@ static int snd_emu10k1_efx_playback_trigger(struct snd_pcm_substream *substream, case SNDRV_PCM_TRIGGER_SUSPEND: case SNDRV_PCM_TRIGGER_STOP: case SNDRV_PCM_TRIGGER_PAUSE_PUSH: - snd_emu10k1_playback_stop_voice(emu, epcm->extra); + if (!emu->das_mode) + snd_emu10k1_playback_stop_voice(emu, epcm->extra); snd_emu10k1_efx_playback_stop_voices( emu, epcm, das_mode, count, runtime->channels); @@ -1428,15 +1451,16 @@ static int snd_emu10k1_efx_playback_open(struct snd_pcm_substream *substream) SNDRV_PCM_INFO_RESUME | SNDRV_PCM_INFO_PAUSE; if (shift == 2) - runtime->hw.channels_max = 7; // FIXME: should be 8, but extra voice ... + runtime->hw.channels_max = 8; err = snd_pcm_hw_constraint_step( runtime, 0, SNDRV_PCM_HW_PARAM_PERIOD_SIZE, 1 << shift); if (err < 0) { kfree(epcm); return err; } } runtime->hw.formats = SNDRV_PCM_FMTBIT_S32_LE; + runtime->hw.periods_max = 2; // Not using an extra voice } } err = snd_emu10k1_playback_set_constraints(emu, runtime); diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index 71aa90b9cc88..be2dbed12e74 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -23,6 +23,14 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) "Suspected sound card removal\n"); break; } + if ((status & IPR_CHANNELLOOP) && emu->das_mode && + (snd_emu10k1_ptr_read(emu, A_DICE, 0) & A_DICE_COUNTER_MASK)) { + // If another IRQ "punched through" the delay, + // we mask away the delayed IRQ. + status &= ~(IPR_CHANNELLOOP | IPR_CHANNELNUMBERMASK); + if (!status) // Will usually happen in 2nd iteration + break; + } if (++timeout == 1000) { dev_info(emu->card->dev, "emu10k1 irq routine failure\n"); break; base-commit: 3c04ccbdf86e7e93e06e68f1915e39eb55afb0d0 prerequisite-patch-id: 45677e4f73ea654bda3258c1143c8027173e4b5c prerequisite-patch-id: 358302d653684999fa1d7b7c0b9c643d6daf768d prerequisite-patch-id: e5ad839d4411bacb9ed5d639a0bf151a3b9dfeb1 prerequisite-patch-id: aa2b589f883089a61a979b76897a60865ab2fd8a prerequisite-patch-id: 86710b3c986270dcdbabbbbad9f0d86ed634a3ad prerequisite-patch-id: 7d3ce3deda4c9a317868cc13150b3ece13f0096e prerequisite-patch-id: 3bd887e1980015ed352cc20eea63b69f3cf8a9c2 -- 2.40.0.152.g15d061e6df

1 0

[PATCH 1/2] ALSA: emu10k1: clean up driver status comments
by Oswald Buddenhagen 17 Jul '23

17 Jul '23

Empty BUGS and TODO sections don't really help anyone, so remove them. Version information is chronically outdated, and not really useful in a git world anyway, so remove it as well. Also remove duplicated (and outdated, of course) status section from p16v.h (the one in p16v.c is in better shape). Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> --- sound/pci/emu10k1/emu10k1_main.c | 6 ---- sound/pci/emu10k1/emufx.c | 6 ---- sound/pci/emu10k1/emumixer.c | 6 ---- sound/pci/emu10k1/emupcm.c | 6 ---- sound/pci/emu10k1/emuproc.c | 6 ---- sound/pci/emu10k1/io.c | 6 ---- sound/pci/emu10k1/irq.c | 6 ---- sound/pci/emu10k1/p16v.h | 56 -------------------------------- sound/pci/emu10k1/p17v.h | 1 - sound/pci/emu10k1/timer.c | 6 ---- sound/pci/emu10k1/tina2.h | 1 - sound/pci/emu10k1/voice.c | 6 ---- 12 files changed, 112 deletions(-) diff --git a/sound/pci/emu10k1/emu10k1_main.c b/sound/pci/emu10k1/emu10k1_main.c index a11fcba4b9af..529d0bba87da 100644 --- a/sound/pci/emu10k1/emu10k1_main.c +++ b/sound/pci/emu10k1/emu10k1_main.c @@ -8,12 +8,6 @@ * Added support for Audigy 2 Value. * Added EMU 1010 support. * General bug fixes and enhancements. - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/sched.h> diff --git a/sound/pci/emu10k1/emufx.c b/sound/pci/emu10k1/emufx.c index 9904bcfee106..e0ad339c5bbf 100644 --- a/sound/pci/emu10k1/emufx.c +++ b/sound/pci/emu10k1/emufx.c @@ -6,12 +6,6 @@ * * Copyright (c) by James Courtier-Dutton <James(a)superbug.co.uk> * Added EMU 1010 support. - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/pci.h> diff --git a/sound/pci/emu10k1/emumixer.c b/sound/pci/emu10k1/emumixer.c index 9a94f08f2463..6ff59c44da36 100644 --- a/sound/pci/emu10k1/emumixer.c +++ b/sound/pci/emu10k1/emumixer.c @@ -8,12 +8,6 @@ * * Copyright (c) by James Courtier-Dutton <James(a)superbug.co.uk> * Added EMU 1010 support. - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/time.h> diff --git a/sound/pci/emu10k1/emupcm.c b/sound/pci/emu10k1/emupcm.c index 8b3d1b35d6e7..7a1d448f0656 100644 --- a/sound/pci/emu10k1/emupcm.c +++ b/sound/pci/emu10k1/emupcm.c @@ -4,12 +4,6 @@ * Creative Labs, Inc. * Routines for control of EMU10K1 chips / PCM routines * Multichannel PCM support Copyright (c) Lee Revell <rlrevell(a)joe-job.com> - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/pci.h> diff --git a/sound/pci/emu10k1/emuproc.c b/sound/pci/emu10k1/emuproc.c index 5533277e4d47..39b422297c70 100644 --- a/sound/pci/emu10k1/emuproc.c +++ b/sound/pci/emu10k1/emuproc.c @@ -6,12 +6,6 @@ * * Copyright (c) by James Courtier-Dutton <James(a)superbug.co.uk> * Added EMU 1010 support. - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/slab.h> diff --git a/sound/pci/emu10k1/io.c b/sound/pci/emu10k1/io.c index c695cb863e5e..7e4483c5bd2a 100644 --- a/sound/pci/emu10k1/io.c +++ b/sound/pci/emu10k1/io.c @@ -3,12 +3,6 @@ * Copyright (c) by Jaroslav Kysela <perex(a)perex.cz> * Creative Labs, Inc. * Routines for control of EMU10K1 chips - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/time.h> diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index 8573248dd799..71aa90b9cc88 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -3,12 +3,6 @@ * Copyright (c) by Jaroslav Kysela <perex(a)perex.cz> * Creative Labs, Inc. * Routines for IRQ control of EMU10K1 chips - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/time.h> diff --git a/sound/pci/emu10k1/p16v.h b/sound/pci/emu10k1/p16v.h index 9d429ad1feff..95ab8071751b 100644 --- a/sound/pci/emu10k1/p16v.h +++ b/sound/pci/emu10k1/p16v.h @@ -2,62 +2,6 @@ /* * Copyright (c) by James Courtier-Dutton <James(a)superbug.demon.co.uk> * Driver p16v chips - * Version: 0.21 - * - * FEATURES currently supported: - * Output fixed at S32_LE, 2 channel to hw:0,0 - * Rates: 44.1, 48, 96, 192. - * - * Changelog: - * 0.8 - * Use separate card based buffer for periods table. - * 0.9 - * Use 2 channel output streams instead of 8 channel. - * (8 channel output streams might be good for ASIO type output) - * Corrected speaker output, so Front -> Front etc. - * 0.10 - * Fixed missed interrupts. - * 0.11 - * Add Sound card model number and names. - * Add Analog volume controls. - * 0.12 - * Corrected playback interrupts. Now interrupt per period, instead of half period. - * 0.13 - * Use single trigger for multichannel. - * 0.14 - * Mic capture now works at fixed: S32_LE, 96000Hz, Stereo. - * 0.15 - * Force buffer_size / period_size == INTEGER. - * 0.16 - * Update p16v.c to work with changed alsa api. - * 0.17 - * Update p16v.c to work with changed alsa api. Removed boot_devs. - * 0.18 - * Merging with snd-emu10k1 driver. - * 0.19 - * One stereo channel at 24bit now works. - * 0.20 - * Added better register defines. - * 0.21 - * Split from p16v.c - * - * BUGS: - * Some stability problems when unloading the snd-p16v kernel module. - * -- - * - * TODO: - * SPDIF out. - * Find out how to change capture sample rates. E.g. To record SPDIF at 48000Hz. - * Currently capture fixed at 48000Hz. - * - * -- - * GENERAL INFO: - * Model: SB0240 - * P16V Chip: CA0151-DBS - * Audigy 2 Chip: CA0102-IAT - * AC97 Codec: STAC 9721 - * ADC: Philips 1361T (Stereo 24bit) - * DAC: CS4382-K (8-channel, 24bit, 192Khz) * * This code was initially based on code from ALSA's emu10k1x.c which is: * Copyright (c) by Francisco Moraes <fmoraes(a)nc.rr.com> diff --git a/sound/pci/emu10k1/p17v.h b/sound/pci/emu10k1/p17v.h index d4ada1c430c8..ee4f4ab4b79c 100644 --- a/sound/pci/emu10k1/p17v.h +++ b/sound/pci/emu10k1/p17v.h @@ -2,7 +2,6 @@ /* * Copyright (c) by James Courtier-Dutton <James(a)superbug.demon.co.uk> * Driver p17v chips - * Version: 0.01 */ /******************************************************************************/ diff --git a/sound/pci/emu10k1/timer.c b/sound/pci/emu10k1/timer.c index f3c78adf3248..8798604e7f98 100644 --- a/sound/pci/emu10k1/timer.c +++ b/sound/pci/emu10k1/timer.c @@ -3,12 +3,6 @@ * Copyright (c) by Lee Revell <rlrevell(a)joe-job.com> * Clemens Ladisch <clemens(a)ladisch.de> * Routines for control of EMU10K1 chips - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/time.h> diff --git a/sound/pci/emu10k1/tina2.h b/sound/pci/emu10k1/tina2.h index 7fd235345292..e3fcb290271c 100644 --- a/sound/pci/emu10k1/tina2.h +++ b/sound/pci/emu10k1/tina2.h @@ -2,7 +2,6 @@ /* * Copyright (c) by James Courtier-Dutton <James(a)superbug.demon.co.uk> * Driver tina2 chips - * Version: 0.1 */ /********************************************************************************************************/ diff --git a/sound/pci/emu10k1/voice.c b/sound/pci/emu10k1/voice.c index 6939498e26f0..ffe87f359a0e 100644 --- a/sound/pci/emu10k1/voice.c +++ b/sound/pci/emu10k1/voice.c @@ -6,12 +6,6 @@ * Routines for control of EMU10K1 chips - voice manager * * Rewrote voice allocator for multichannel support - rlrevell 12/2004 - * - * BUGS: - * -- - * - * TODO: - * -- */ #include <linux/time.h> -- 2.40.0.152.g15d061e6df

2 2

[PATCH v3] ALSA: emu10k1: set the "no filtering" bits on PCM voices on Audigy
by Oswald Buddenhagen 17 Jul '23

17 Jul '23

Given that the filter is already set to neutral for PCM voices, the only observable effect is that the Z1/Z2/FXBUS registers don't have a stray bit set for negative numbers anymore. The bit is below the ones significant for output, but it would mess with 32-bit sample recombination, which we intend to add. kX-project does that, but I had to figure out myself why. Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> --- v3: - simplify, as this is actually unrelated to the interpolator --- include/sound/emu10k1.h | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/include/sound/emu10k1.h b/include/sound/emu10k1.h index 43c097952c3c..ef948e45b4c6 100644 --- a/include/sound/emu10k1.h +++ b/include/sound/emu10k1.h @@ -902,6 +902,11 @@ SUB_REG_NC(A_EHC, A_I2S_CAPTURE_RATE, 0x00000e00) /* This sets the capture PCM #define A_TTDA 0x7a /* Tank Table DMA Address */ #define A_TTDD 0x7b /* Tank Table DMA Data */ +// In A_FXRT1 & A_FXRT2, the 0x80 bit of each byte completely disables the +// filter (CVCF_CURRENTFILTER) for the corresponding channel. There is no +// effect on the volume (CVCF_CURRENTVOLUME) or the interpolator's filter +// (CCCA_INTERPROM_MASK). + #define A_FXRT2 0x7c #define A_FXRT_CHANNELE 0x0000003f /* Effects send bus number for channel's effects send E */ #define A_FXRT_CHANNELF 0x00003f00 /* Effects send bus number for channel's effects send F */ @@ -914,8 +919,6 @@ SUB_REG_NC(A_EHC, A_I2S_CAPTURE_RATE, 0x00000e00) /* This sets the capture PCM #define A_FXSENDAMOUNT_G_MASK 0x0000FF00 #define A_FXSENDAMOUNT_H_MASK 0x000000FF -/* 0x7c, 0x7e "high bit is used for filtering" */ - /* The send amounts for this one are the same as used with the emu10k1 */ #define A_FXRT1 0x7e #define A_FXRT_CHANNELA 0x0000003f @@ -1523,10 +1526,10 @@ struct snd_emu10k1_pcm_mixer { ((route[0] | (route[1] << 4) | (route[2] << 8) | (route[3] << 12)) << 16) #define snd_emu10k1_compose_audigy_fxrt1(route) \ -((unsigned int)route[0] | ((unsigned int)route[1] << 8) | ((unsigned int)route[2] << 16) | ((unsigned int)route[3] << 24)) +((unsigned int)route[0] | ((unsigned int)route[1] << 8) | ((unsigned int)route[2] << 16) | ((unsigned int)route[3] << 24) | 0x80808080) #define snd_emu10k1_compose_audigy_fxrt2(route) \ -((unsigned int)route[4] | ((unsigned int)route[5] << 8) | ((unsigned int)route[6] << 16) | ((unsigned int)route[7] << 24)) +((unsigned int)route[4] | ((unsigned int)route[5] << 8) | ((unsigned int)route[6] << 16) | ((unsigned int)route[7] << 24) | 0x80808080) #define snd_emu10k1_compose_audigy_sendamounts(vol) \ (((unsigned int)vol[4] << 24) | ((unsigned int)vol[5] << 16) | ((unsigned int)vol[6] << 8) | (unsigned int)vol[7]) -- 2.40.0.152.g15d061e6df

2 1

[PATCH v2] ALSA: emu10k1: track loss of external clock on E-MU cards
by Oswald Buddenhagen 17 Jul '23

17 Jul '23

This uses IRQs to track spontaneous changes to the word clock source register. FWIW, that this can happen in the first place is the reason why it is futile to lock the clock source mixer setting while the device is open - we can't consistently control the rate anyway. Though arguably, we should reset any open streams when that happens, as they become corrupted anyway. Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen(a)gmx.de> --- v2: - use driver-local lock instead of core's control_rwsem --- FIXME? while i'm not sure, i think this won't notice seamless switches between 44.1 and 48 kHz. assuming that's the case, this seems like a minor issue: firstly, just about nothing actually produces such a seamless switch - my only device that can even do that is the e-mu card itself, but the driver disrupts that by temporarily muting the output. secondly, the user is unlikely to select an external source before setting it up properly. and the easy workaround is actually never doing that. regardless, to actually test that i'd need a second e-mu card. --- include/sound/emu10k1.h | 5 +++++ sound/pci/emu10k1/emu10k1.c | 1 + sound/pci/emu10k1/emu10k1_main.c | 30 +++++++++++++++++++++++++++++- sound/pci/emu10k1/emumixer.c | 12 ++++++++---- 4 files changed, 43 insertions(+), 5 deletions(-) diff --git a/include/sound/emu10k1.h b/include/sound/emu10k1.h index 43c097952c3c..7c55a8244747 100644 --- a/include/sound/emu10k1.h +++ b/include/sound/emu10k1.h @@ -992,6 +992,9 @@ SUB_REG_NC(A_EHC, A_I2S_CAPTURE_RATE, 0x00000e00) /* This sets the capture PCM #define EMU_HANA_WCLOCK_4X 0x10 #define EMU_HANA_WCLOCK_MULT_RESERVED 0x18 +// If the selected external clock source is/becomes invalid or incompatible +// with the clock multiplier, the clock source is reset to this value, and +// a WCLK_CHANGED interrupt is raised. #define EMU_HANA_DEFCLOCK 0x06 /* 000000x 1 bits Default Word Clock */ #define EMU_HANA_DEFCLOCK_48K 0x00 #define EMU_HANA_DEFCLOCK_44_1K 0x01 @@ -1679,6 +1682,7 @@ struct snd_emu1010 { unsigned int optical_in; /* 0:SPDIF, 1:ADAT */ unsigned int optical_out; /* 0:SPDIF, 1:ADAT */ struct work_struct firmware_work; + struct work_struct clock_work; }; struct snd_emu10k1 { @@ -1753,6 +1757,7 @@ struct snd_emu10k1 { struct snd_kcontrol *ctl_efx_send_routing; struct snd_kcontrol *ctl_efx_send_volume; struct snd_kcontrol *ctl_efx_attn; + struct snd_kcontrol *ctl_clock_source; void (*hwvol_interrupt)(struct snd_emu10k1 *emu, unsigned int status); void (*capture_interrupt)(struct snd_emu10k1 *emu, unsigned int status); diff --git a/sound/pci/emu10k1/emu10k1.c b/sound/pci/emu10k1/emu10k1.c index 1a13c086e86c..421053569aa0 100644 --- a/sound/pci/emu10k1/emu10k1.c +++ b/sound/pci/emu10k1/emu10k1.c @@ -192,6 +192,7 @@ static int snd_emu10k1_suspend(struct device *dev) emu->suspend = 1; cancel_work_sync(&emu->emu1010.firmware_work); + cancel_work_sync(&emu->emu1010.clock_work); snd_ac97_suspend(emu->ac97); diff --git a/sound/pci/emu10k1/emu10k1_main.c b/sound/pci/emu10k1/emu10k1_main.c index a11fcba4b9af..604645bfc908 100644 --- a/sound/pci/emu10k1/emu10k1_main.c +++ b/sound/pci/emu10k1/emu10k1_main.c @@ -789,6 +789,30 @@ static void emu1010_firmware_work(struct work_struct *work) } } +static void emu1010_clock_work(struct work_struct *work) +{ + struct snd_emu10k1 *emu; + struct snd_ctl_elem_id id; + + emu = container_of(work, struct snd_emu10k1, + emu1010.clock_work); + if (emu->card->shutdown) + return; +#ifdef CONFIG_PM_SLEEP + if (emu->suspend) + return; +#endif + + spin_lock_irq(&emu->reg_lock); + // This is the only thing that can actually happen. + emu->emu1010.clock_source = emu->emu1010.clock_fallback; + emu->emu1010.wclock = 1 - emu->emu1010.clock_source; + snd_emu1010_update_clock(emu); + spin_unlock_irq(&emu->reg_lock); + snd_ctl_build_ioff(&id, emu->ctl_clock_source, 0); + snd_ctl_notify(emu->card, SNDRV_CTL_EVENT_MASK_VALUE, &id); +} + static void emu1010_interrupt(struct snd_emu10k1 *emu) { u32 sts; @@ -802,6 +826,8 @@ static void emu1010_interrupt(struct snd_emu10k1 *emu) } else if (sts & EMU_HANA_IRQ_DOCK) { schedule_work(&emu->emu1010.firmware_work); } + if (sts & EMU_HANA_IRQ_WCLK_CHANGED) + schedule_work(&emu->emu1010.clock_work); } /* @@ -901,7 +927,7 @@ static int snd_emu10k1_emu1010_init(struct snd_emu10k1 *emu) emu->gpio_interrupt = emu1010_interrupt; // Note: The Audigy INTE is set later snd_emu1010_fpga_write(emu, EMU_HANA_IRQ_ENABLE, - EMU_HANA_IRQ_DOCK | EMU_HANA_IRQ_DOCK_LOST); + EMU_HANA_IRQ_DOCK | EMU_HANA_IRQ_DOCK_LOST | EMU_HANA_IRQ_WCLK_CHANGED); snd_emu1010_fpga_read(emu, EMU_HANA_IRQ_STATUS, &reg); // Clear pending IRQs emu->emu1010.clock_source = 1; /* 48000 */ @@ -943,6 +969,7 @@ static void snd_emu10k1_free(struct snd_card *card) snd_emu1010_fpga_write(emu, EMU_HANA_DOCK_PWR, 0); } cancel_work_sync(&emu->emu1010.firmware_work); + cancel_work_sync(&emu->emu1010.clock_work); release_firmware(emu->firmware); release_firmware(emu->dock_fw); snd_util_memhdr_free(emu->memhdr); @@ -1522,6 +1549,7 @@ int snd_emu10k1_create(struct snd_card *card, emu->synth = NULL; emu->get_synth_voice = NULL; INIT_WORK(&emu->emu1010.firmware_work, emu1010_firmware_work); + INIT_WORK(&emu->emu1010.clock_work, emu1010_clock_work); /* read revision & serial */ emu->revision = pci->revision; pci_read_config_dword(pci, PCI_SUBSYSTEM_VENDOR_ID, &emu->serial); diff --git a/sound/pci/emu10k1/emumixer.c b/sound/pci/emu10k1/emumixer.c index 9a94f08f2463..f39025748072 100644 --- a/sound/pci/emu10k1/emumixer.c +++ b/sound/pci/emu10k1/emumixer.c @@ -986,17 +986,21 @@ static int snd_emu1010_clock_source_put(struct snd_kcontrol *kcontrol, val = ucontrol->value.enumerated.item[0] ; if (val >= emu_ci->num) return -EINVAL; + spin_lock_irq(&emu->reg_lock); change = (emu->emu1010.clock_source != val); if (change) { emu->emu1010.clock_source = val; emu->emu1010.wclock = emu_ci->vals[val]; + snd_emu1010_update_clock(emu); snd_emu1010_fpga_write(emu, EMU_HANA_UNMUTE, EMU_MUTE); snd_emu1010_fpga_write(emu, EMU_HANA_WCLOCK, emu->emu1010.wclock); + spin_unlock_irq(&emu->reg_lock); + msleep(10); // Allow DLL to settle snd_emu1010_fpga_write(emu, EMU_HANA_UNMUTE, EMU_UNMUTE); - - snd_emu1010_update_clock(emu); + } else { + spin_unlock_irq(&emu->reg_lock); } return change; } @@ -2336,8 +2340,8 @@ int snd_emu10k1_mixer(struct snd_emu10k1 *emu, emu1010_map_source(emu_ri, emu_ri->out_dflts[i]); snd_emu1010_apply_sources(emu); - err = snd_ctl_add(card, - snd_ctl_new1(&snd_emu1010_clock_source, emu)); + kctl = emu->ctl_clock_source = snd_ctl_new1(&snd_emu1010_clock_source, emu); + err = snd_ctl_add(card, kctl); if (err < 0) return err; err = snd_ctl_add(card, -- 2.40.0.152.g15d061e6df

2 1

[PATCH] soundwire: Explicitly include correct DT includes
by Rob Herring 17 Jul '23

17 Jul '23

The DT of_device.h and of_platform.h date back to the separate of_platform_bus_type before it as merged into the regular platform bus. As part of that merge prepping Arm DT support 13 years ago, they "temporarily" include each other. They also include platform_device.h and of.h. As a result, there's a pretty much random mix of those include files used throughout the tree. In order to detangle these headers and replace the implicit includes with struct declarations, users need to explicitly include the correct includes. Signed-off-by: Rob Herring <robh(a)kernel.org> --- drivers/soundwire/qcom.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/soundwire/qcom.c b/drivers/soundwire/qcom.c index 7970fdb27ba0..d178a0dc0918 100644 --- a/drivers/soundwire/qcom.c +++ b/drivers/soundwire/qcom.c @@ -10,7 +10,6 @@ #include <linux/debugfs.h> #include <linux/of.h> #include <linux/of_irq.h> -#include <linux/of_device.h> #include <linux/pm_runtime.h> #include <linux/regmap.h> #include <linux/reset.h> -- 2.40.1

2 1