July 2024 - Alsa-devel - mailman.alsa-project.org

[PATCH 0/2] {ASoC,media}: constify snd_soc_component_driver struct
by Javier Carrasco 26 Jul '24

26 Jul '24

The `snd_soc_component_driver` struct is never modified after its declaration, and its only direct user `devm_snd_soc_register_component()` expects a constant value anyway. Declare `snd_soc_component_driver` as const to move their declarations to read-only sections. Apart from a single case under media/, the affected drivers are members of the ASoC subsystem. Signed-off-by: Javier Carrasco <javier.carrasco.cruz(a)gmail.com> --- Javier Carrasco (2): media: i2c: tda1997x: constify snd_soc_component_driver struct ASoC: constify snd_soc_component_driver struct drivers/media/i2c/tda1997x.c | 2 +- sound/soc/au1x/dbdma2.c | 2 +- sound/soc/au1x/dma.c | 2 +- sound/soc/bcm/cygnus-pcm.c | 2 +- sound/soc/codecs/cpcap.c | 2 +- sound/soc/codecs/cs43130.c | 2 +- sound/soc/codecs/pcm186x.c | 4 ++-- sound/soc/codecs/pcm5102a.c | 2 +- sound/soc/codecs/spdif_receiver.c | 2 +- sound/soc/codecs/spdif_transmitter.c | 2 +- sound/soc/codecs/sti-sas.c | 2 +- sound/soc/codecs/tas6424.c | 2 +- sound/soc/stm/stm32_adfsdm.c | 2 +- sound/soc/uniphier/evea.c | 2 +- 14 files changed, 15 insertions(+), 15 deletions(-) --- base-commit: 864b1099d16fc7e332c3ad7823058c65f890486c change-id: 20240725-const_snd_soc_component_driver-b9629a95b948 Best regards, -- Javier Carrasco <javier.carrasco.cruz(a)gmail.com>

3 6

[GIT PULL] ASoC fixes for v6.11-merge-window
by Mark Brown 25 Jul '24

25 Jul '24

The following changes since commit c51cba4755609ad97ba97713210c16f043c73224: Fix the unbalanced pm_runtime_enable in wcd937x-sdw (2024-07-12 17:38:18 +0100) are available in the Git repository at: https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git tags/asoc-fix-v6.11-merge-window for you to fetch changes up to ab53dfdcdd1ec8df8729890aefa5b0e3c900afbb: ASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components (2024-07-25 14:20:30 +0100) ---------------------------------------------------------------- ASoC: Fixes for v6.11 A selection of routine fixes and quirks that came in since the merge window. The fsl-asoc-card change is a fix for systems with multiple cards where updating templates in place leaks data from one card to another. ---------------------------------------------------------------- Arnd Bergmann (1): ASoC: tegra: select CONFIG_SND_SIMPLE_CARD_UTILS Curtis Malainey (1): ASoC: Intel: Fix RT5650 SSP lookup Dan Carpenter (1): ASoC: TAS2781: Fix tasdev_load_calibrated_data() Daniel Baluta (1): ASoC: SOF: imx8m: Fix DSP control regmap retrieval Peter Ujfalusi (2): ASoC: SOF: ipc4-topology: Only handle dai_config with HW_PARAMS for ChainDMA ASoC: SOF: ipc4-topology: Preserve the DMA Link ID for ChainDMA on unprepare Pierre-Louis Bossart (2): ASOC: SOF: Intel: hda-loader: only wait for HDaudio IOC for IPC4 devices ASoC: Intel: use soc_intel_is_byt_cr() only when IOSF_MBI is reachable Shengjiu Wang (1): ASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components Takashi Iwai (1): ASoC: amd: yc: Support mic on Lenovo Thinkpad E16 Gen 2 Venkata Prasad Potturu (1): ASoC: sof: amd: fix for firmware reload failure in Vangogh platform sound/soc/amd/yc/acp6x-mach.c | 7 ++++ sound/soc/codecs/tas2781-fmwlib.c | 2 +- sound/soc/fsl/fsl-asoc-card.c | 46 ++++++++++++++-------- sound/soc/intel/common/soc-acpi-intel-ssp-common.c | 9 +++++ sound/soc/intel/common/soc-intel-quirks.h | 2 +- sound/soc/sof/amd/pci-vangogh.c | 1 - sound/soc/sof/imx/imx8m.c | 2 +- sound/soc/sof/intel/hda-loader.c | 20 ++++++---- sound/soc/sof/intel/hda.c | 17 ++++---- sound/soc/sof/ipc4-topology.c | 18 +++++++-- sound/soc/tegra/Kconfig | 1 + 11 files changed, 85 insertions(+), 40 deletions(-)

2 1

[REGRESSION] ALSA: firewire-lib: heavy digital distortion with Fireface 800
by edmund.raile 25 Jul '24

25 Jul '24

Bisection revealed that the bitcrushing distortion with RME FireFace 800 was caused by 1d717123bb1a7555 ("ALSA: firewire-lib: Avoid -Wflex-array-member-not-at-end warning"). Reverting this commit yields restoration of clear audio output. I will send in a patch reverting this commit for now, soonTM. #regzbot introduced: 1d717123bb1a7555

4 6

[PATCH] ASoC: fsl-asoc-card: Dynamically allocate memory for snd_soc_dai_link_components
by Shengjiu Wang 25 Jul '24

25 Jul '24

The static snd_soc_dai_link_components cause conflict for multiple instances of this generic driver. For example, when there is wm8962 and SPDIF case enabled together, the contaminated snd_soc_dai_link_components will cause another device probe fail. Fixes: 6d174cc4f224 ("ASoC: fsl-asoc-card: merge spdif support from imx-spdif.c") Signed-off-by: Shengjiu Wang <shengjiu.wang(a)nxp.com> --- sound/soc/fsl/fsl-asoc-card.c | 46 ++++++++++++++++++++++------------- 1 file changed, 29 insertions(+), 17 deletions(-) diff --git a/sound/soc/fsl/fsl-asoc-card.c b/sound/soc/fsl/fsl-asoc-card.c index 82df887b3af5..f6c3aeff0d8e 100644 --- a/sound/soc/fsl/fsl-asoc-card.c +++ b/sound/soc/fsl/fsl-asoc-card.c @@ -306,27 +306,12 @@ static int be_hw_params_fixup(struct snd_soc_pcm_runtime *rtd, return 0; } -SND_SOC_DAILINK_DEFS(hifi, - DAILINK_COMP_ARRAY(COMP_EMPTY()), - DAILINK_COMP_ARRAY(COMP_EMPTY(), COMP_EMPTY()), - DAILINK_COMP_ARRAY(COMP_EMPTY())); - -SND_SOC_DAILINK_DEFS(hifi_fe, - DAILINK_COMP_ARRAY(COMP_EMPTY()), - DAILINK_COMP_ARRAY(COMP_DUMMY()), - DAILINK_COMP_ARRAY(COMP_EMPTY())); - -SND_SOC_DAILINK_DEFS(hifi_be, - DAILINK_COMP_ARRAY(COMP_EMPTY()), - DAILINK_COMP_ARRAY(COMP_EMPTY(), COMP_EMPTY())); - static const struct snd_soc_dai_link fsl_asoc_card_dai[] = { /* Default ASoC DAI Link*/ { .name = "HiFi", .stream_name = "HiFi", .ops = &fsl_asoc_card_ops, - SND_SOC_DAILINK_REG(hifi), }, /* DPCM Link between Front-End and Back-End (Optional) */ { @@ -335,7 +320,6 @@ static const struct snd_soc_dai_link fsl_asoc_card_dai[] = { .dpcm_playback = 1, .dpcm_capture = 1, .dynamic = 1, - SND_SOC_DAILINK_REG(hifi_fe), }, { .name = "HiFi-ASRC-BE", @@ -345,7 +329,6 @@ static const struct snd_soc_dai_link fsl_asoc_card_dai[] = { .dpcm_playback = 1, .dpcm_capture = 1, .no_pcm = 1, - SND_SOC_DAILINK_REG(hifi_be), }, }; @@ -637,6 +620,7 @@ static int fsl_asoc_card_probe(struct platform_device *pdev) struct platform_device *cpu_pdev; struct fsl_asoc_card_priv *priv; struct device *codec_dev[2] = { NULL, NULL }; + struct snd_soc_dai_link_component *dlc; const char *codec_dai_name[2]; const char *codec_dev_name[2]; u32 asrc_fmt = 0; @@ -717,7 +701,35 @@ static int fsl_asoc_card_probe(struct platform_device *pdev) memcpy(priv->dai_link, fsl_asoc_card_dai, sizeof(struct snd_soc_dai_link) * ARRAY_SIZE(priv->dai_link)); + /* + * "Default ASoC DAI Link": 1 cpus, 2 codecs, 1 platforms + * "DPCM Link Front-End": 1 cpus, 1 codecs (dummy), 1 platforms + * "DPCM Link Back-End": 1 cpus, 2 codecs + * totally 10 components + */ + dlc = devm_kcalloc(&pdev->dev, 10, sizeof(*dlc), GFP_KERNEL); + if (!dlc) { + ret = -ENOMEM; + goto asrc_fail; + } + + priv->dai_link[0].cpus = &dlc[0]; + priv->dai_link[0].num_cpus = 1; + priv->dai_link[0].codecs = &dlc[1]; priv->dai_link[0].num_codecs = 1; + priv->dai_link[0].platforms = &dlc[3]; + priv->dai_link[0].num_platforms = 1; + + priv->dai_link[1].cpus = &dlc[4]; + priv->dai_link[1].num_cpus = 1; + priv->dai_link[1].codecs = &dlc[5]; + priv->dai_link[1].num_codecs = 0; /* dummy */ + priv->dai_link[1].platforms = &dlc[6]; + priv->dai_link[1].num_platforms = 1; + + priv->dai_link[2].cpus = &dlc[7]; + priv->dai_link[2].num_cpus = 1; + priv->dai_link[2].codecs = &dlc[8]; priv->dai_link[2].num_codecs = 1; priv->card.dapm_routes = audio_map; -- 2.34.1

2 1

[PATCH v5 0/9] iio: adc: dfsdm: add scaling support
by Olivier Moysan 25 Jul '24

25 Jul '24

The aim of this serie is to add scaling support to STM32 DFSDM peripheral in the analog context. The DFSDM currently operates as a consumer of IIO channels provided by a generic SD modulator. As previously discussed in RFC [1], this topology is not suitable for implementing scaling. This series brings the integration of the DFSDM driver with the new IIO backend framework [2], enabling the DFSDM IIO device to offer scaling feature based on reference voltage data obtained from the IIO SD modulator backend. This generic SD modulator backend takes the place of the former SD modulator, used with legacy implementation. The DFSDM driver has been updated to adopt the generic ADC channel binding [3]. The reasons for this include: - Reducing the use of proprietary properties - Simplifying the coexistence of legacy and new backend bindings - Prepare the support of the MDF peripheral on STM32MP25 SoC Backward compatibility is maintained through legacy support. This series extends the backend framework with the following APIs: - iio_backend_read_raw: This API is intented to retrieve channel attributes from the backend. It is based on IIO framework read_raw API. - iio_backend_read_scale / iio_backend_read_offset Dedicated APIs to read scale and offset channel values. - iio_backend_disable / iio_backend_enable: backend enable/disable to be used for PM management - devm_iio_backend_fwnode_get Intended for parsing DT subnodes to allow generic channel binding support, as generic channel DT nodes are not populated as devices. [1]: https://lore.kernel.org/lkml/20200204101008.11411-1-olivier.moysan@st.com/ [2]: https://lore.kernel.org/all/20240206-iio-backend-v9-0-df66d159c000@analog.c… [3]: devicetree/bindings/iio/adc/adc.yaml Changes in v2: - Update enable/disable backend API - Rename devm_iio_backend_subnode_get(), as devm_iio_backend_fwnode_get() - Update iio_backend_read_raw() prototype to fully match IIO framework read_raw callback prototype. - Change st,adc-channel-type property name and type in DFSDM binding - Remove sd-backend and rename ads1201 compatibles in SD binding Changes in v3: - Add iio_backend_read_scale and iio_backend_read_offset - Remove sd-backend compatible. Adapt existing sd modulator driver and binding to support backend topology. - Misc corrections in DFSDM binding & minor changes in DFSDM driver Changes in v4: - Remove allOf and rework examples in SD modulator binding - Wrap lines to 80 characters in IIO code Changes in v5: - Move iio_backend_read_scale and iio_backend_read_offset to inline and update prototype - Fix channel number & remove legacy binding example in DFSDM binding - Rearrange DFSDM patches to fix build issue in serie - Split SD modulator binding patch. Move reference voltage change in a separate patch Olivier Moysan (9): iio: add read scale and offset services to iio backend framework iio: add enable and disable services to iio backend framework iio: add child nodes support in iio backend framework dt-bindings: iio: dfsdm: move to backend framework dt-bindings: iio: add backend support to sd modulator dt-bindings: iio: add vref support to sd modulator iio: adc: stm32-dfsdm: adopt generic channels bindings iio: add iio backend support to sd modulator iio: adc: stm32-dfsdm: add scaling support to dfsdm .../iio/adc/sigma-delta-modulator.yaml | 25 +- .../bindings/iio/adc/st,stm32-dfsdm-adc.yaml | 124 +++++++- drivers/iio/adc/Kconfig | 2 + drivers/iio/adc/sd_adc_modulator.c | 92 +++++- drivers/iio/adc/stm32-dfsdm-adc.c | 293 +++++++++++++++--- drivers/iio/industrialio-backend.c | 101 +++++- include/linux/iio/backend.h | 29 +- 7 files changed, 593 insertions(+), 73 deletions(-) base-commit: 472438c7e0e2261c6737a8321f46ef176eef1c8f -- 2.25.1

3 3

[RFC v2 00/xx] ALSA: ALSA: add snd_pcm_is_playback/capture()
by Kuninori Morimoto 25 Jul '24

25 Jul '24

Hi Iwai-san, Mark Cc Sakamoto-san, Amadeusz, Pierre-Louis It seems we can use _Generic() more simply, v2 RFC is using it. I'm not 100% sure but unfortunately we can't use bit-field with _Generic() in gcc (clang seems possible to handle it ?). pci/ac97 is the only user of bit-field direction. To avoid posting patch-bomb, I will post main patch and some sample patches as RFC. Please review it. Link: https://lore.kernel.org/r/87zfqel1g7.wl-kuninori.morimoto.gx@renesas.com Thank you for your help !! Best regards --- Kuninori Morimoto

4 9

[kernel][usb audio class 1] surround channels mapped to wrong alsa channel ids
by GitHub issues - opened 25 Jul '24

25 Jul '24

alsa-project/alsa-lib issue #403 was opened from sylware: In $linux_src/sound/usb/stream.c, function "convert_chmap", in the "uac1_maps" array, the "left surround" should be SNDRV_CHMAP_RL instead of SNDRV_CHMAP_SL, and the "right surround" should be SNDRV_CHMAP_RR instead of SNDDRV_CHMAP_SR. Issue URL : https://github.com/alsa-project/alsa-lib/issues/403 Repository URL: https://github.com/alsa-project/alsa-lib

1 0

[PATCH] ALSA: firewire-lib: restore process context workqueue to prevent deadlock
by Edmund Raile 24 Jul '24

24 Jul '24

Commit b5b519965c4c ("ALSA: firewire-lib: operate for period elapse event in process context") removed the process context workqueue from amdtp_domain_stream_pcm_pointer() and update_pcm_pointers() to remove its overhead. With RME Fireface 800, this lead to a regression since Kernels 5.14.0, causing a deadlock with eventual system freeze under ALSA operation: ? tasklet_unlock_spin_wait </NMI> <TASK> ohci_flush_iso_completions firewire_ohci amdtp_domain_stream_pcm_pointer snd_firewire_lib snd_pcm_update_hw_ptr0 snd_pcm snd_pcm_status64 snd_pcm ? native_queued_spin_lock_slowpath </NMI> <IRQ> _raw_spin_lock_irqsave snd_pcm_period_elapsed snd_pcm process_rx_packets snd_firewire_lib irq_target_callback snd_firewire_lib handle_it_packet firewire_ohci context_tasklet firewire_ohci Restore the work queue to prevent deadlock between ALSA substream process context spin_lock of snd_pcm_stream_lock_irq() in snd_pcm_status64() and OHCI 1394 IT softIRQ context spin_lock of snd_pcm_stream_lock_irqsave() in snd_pcm_period_elapsed(). to reproduce the issue: direct ALSA playback to the device: mpv --audio-device=alsa/sysdefault:CARD=Fireface800 Spor-Ignition.flac Time to occurrence: 2s to 30m Likelihood increased by: - high CPU load stress --cpu $(nproc) - switching between applications via workspaces tested with i915 in Xfce PulsaAudio / PipeWire conceal the issue as they run PCM substream without period wakeup mode, issuing less hardIRQs. Closes: https://lore.kernel.org/regressions/kwryofzdmjvzkuw6j3clftsxmoolynljztxqwg7… Fixes: 7ba5ca32fe6e ("ALSA: firewire-lib: operate for period elapse event in process context") Signed-off-by: Edmund Raile <edmund.raile(a)proton.me> --- This is the follow-up patch to the 5.14.0 regression I reported: https://lore.kernel.org/regressions/kwryofzdmjvzkuw6j3clftsxmoolynljztxqwg7… ("[REGRESSION] ALSA: firewire-lib: snd_pcm_period_elapsed deadlock with Fireface 800") Takashi Sakamoto explained the issue in his response to the regression: A. In the process context * (lock A) Acquiring spin_lock by snd_pcm_stream_lock_irq() in snd_pcm_status64() * (lock B) Then attempt to enter tasklet B. In the softIRQ context * (lock B) Enter tasklet * (lock A) Attempt to acquire spin_lock by snd_pcm_stream_lock_irqsave() in snd_pcm_period_elapsed() This leads me to believe this isn't just an issue limited to the RME Fireface driver (snd_fireface), though I can not test the other devices. sound/firewire/amdtp-stream.c | 32 +++++++++++++++++++++----------- sound/firewire/amdtp-stream.h | 1 + 2 files changed, 22 insertions(+), 11 deletions(-) diff --git a/sound/firewire/amdtp-stream.c b/sound/firewire/amdtp-stream.c index d35d0a420ee0..77b99a2117f4 100644 --- a/sound/firewire/amdtp-stream.c +++ b/sound/firewire/amdtp-stream.c @@ -77,6 +77,8 @@ // overrun. Actual device can skip more, then this module stops the packet streaming. #define IR_JUMBO_PAYLOAD_MAX_SKIP_CYCLES 5 +static void pcm_period_work(struct work_struct *work); + /** * amdtp_stream_init - initialize an AMDTP stream structure * @s: the AMDTP stream to initialize @@ -105,6 +107,7 @@ int amdtp_stream_init(struct amdtp_stream *s, struct fw_unit *unit, s->flags = flags; s->context = ERR_PTR(-1); mutex_init(&s->mutex); + INIT_WORK(&s->period_work, pcm_period_work); s->packet_index = 0; init_waitqueue_head(&s->ready_wait); @@ -347,6 +350,7 @@ EXPORT_SYMBOL(amdtp_stream_get_max_payload); */ void amdtp_stream_pcm_prepare(struct amdtp_stream *s) { + cancel_work_sync(&s->period_work); s->pcm_buffer_pointer = 0; s->pcm_period_pointer = 0; } @@ -611,19 +615,21 @@ static void update_pcm_pointers(struct amdtp_stream *s, // The program in user process should periodically check the status of intermediate // buffer associated to PCM substream to process PCM frames in the buffer, instead // of receiving notification of period elapsed by poll wait. - if (!pcm->runtime->no_period_wakeup) { - if (in_softirq()) { - // In software IRQ context for 1394 OHCI. - snd_pcm_period_elapsed(pcm); - } else { - // In process context of ALSA PCM application under acquired lock of - // PCM substream. - snd_pcm_period_elapsed_under_stream_lock(pcm); - } - } + if (!pcm->runtime->no_period_wakeup) + queue_work(system_highpri_wq, &s->period_work); } } +static void pcm_period_work(struct work_struct *work) +{ + struct amdtp_stream *s = container_of(work, struct amdtp_stream, + period_work); + struct snd_pcm_substream *pcm = READ_ONCE(s->pcm); + + if (pcm) + snd_pcm_period_elapsed(pcm); +} + static int queue_packet(struct amdtp_stream *s, struct fw_iso_packet *params, bool sched_irq) { @@ -1854,7 +1860,10 @@ unsigned long amdtp_domain_stream_pcm_pointer(struct amdtp_domain *d, if (irq_target && amdtp_stream_running(irq_target)) { // In software IRQ context, the call causes dead-lock to disable the tasklet // synchronously. - if (!in_softirq()) + // workqueue serves to prevent deadlock between process context spinlock + // of snd_pcm_stream_lock_irq() in snd_pcm_status64() and softIRQ spinlock + // of snd_pcm_stream_lock_irqsave() in snd_pcm_period_elapsed() + if ((!in_softirq()) && (current_work() != &s->period_work)) fw_iso_context_flush_completions(irq_target->context); } @@ -1910,6 +1919,7 @@ static void amdtp_stream_stop(struct amdtp_stream *s) return; } + cancel_work_sync(&s->period_work); fw_iso_context_stop(s->context); fw_iso_context_destroy(s->context); s->context = ERR_PTR(-1); diff --git a/sound/firewire/amdtp-stream.h b/sound/firewire/amdtp-stream.h index a1ed2e80f91a..775db3fc4959 100644 --- a/sound/firewire/amdtp-stream.h +++ b/sound/firewire/amdtp-stream.h @@ -191,6 +191,7 @@ struct amdtp_stream { /* For a PCM substream processing. */ struct snd_pcm_substream *pcm; + struct work_struct period_work; snd_pcm_uframes_t pcm_buffer_pointer; unsigned int pcm_period_pointer; unsigned int pcm_frame_multiplier; -- 2.45.2

2 1

[bug report] ASoC: Intel: cht_bsw_rt5645: Set card.components string
by Dan Carpenter 23 Jul '24

23 Jul '24

Hello Hans de Goede, Commit f87b4402163b ("ASoC: Intel: cht_bsw_rt5645: Set card.components string") from Nov 26, 2023 (linux-next), leads to the following Smatch static checker warning: sound/soc/intel/boards/cht_bsw_rt5645.c:587 snd_cht_mc_probe() error: we previously assumed 'adev' could be null (see line 581) sound/soc/intel/boards/cht_bsw_rt5645.c 570 /* set correct codec name */ 571 for (i = 0; i < ARRAY_SIZE(cht_dailink); i++) 572 if (cht_dailink[i].codecs->name && 573 !strcmp(cht_dailink[i].codecs->name, 574 "i2c-10EC5645:00")) { 575 dai_index = i; 576 break; 577 } 578 579 /* fixup codec name based on HID */ 580 adev = acpi_dev_get_first_match_dev(mach->id, NULL, -1); 581 if (adev) { ^^^^ The old code assumes adev can be NULL 582 snprintf(cht_rt5645_codec_name, sizeof(cht_rt5645_codec_name), 583 "i2c-%s", acpi_dev_name(adev)); 584 cht_dailink[dai_index].codecs->name = cht_rt5645_codec_name; 585 } 586 /* acpi_get_first_physical_node() returns a borrowed ref, no need to deref */ --> 587 codec_dev = acpi_get_first_physical_node(adev); ^^^^ Unchecked dereference 588 acpi_dev_put(adev); 589 if (!codec_dev) 590 return -EPROBE_DEFER; 591 592 snd_soc_card_chtrt5645.components = rt5645_components(codec_dev); 593 snd_soc_card_chtrt5650.components = rt5645_components(codec_dev); 594 regards, dan carpenter

2 2

[PATCH for-6.10 0/2] ASoC: cs35l56: Set correct upper volume limit
by Richard Fitzgerald 23 Jul '24

23 Jul '24

Patch series to limit the upper range of the CS35L56 volume control to +12 dB. These commits were not marked 'Fixes' because they were thought to be only a cosmetic issue. The user could reduce the volume to a usable value. But for some complex audio topologies with SOF Audio DSP + CS42L43 + multiple CS35L56 it has turned out to be not obvious to the user what the problem actually is and what to do to fix it. As support for these topologies went into 6.10 we would like this series to be applied to 6.10. Richard Fitzgerald (2): ASoC: cs35l56: Use header defines for Speaker Volume control definition ASoC: cs35l56: Limit Speaker Volume to +12dB maximum include/sound/cs35l56.h | 2 +- sound/soc/codecs/cs35l56.c | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-) -- 2.39.2

2 3