| Message ID | 20231004145540.32321-1-maarten.lankhorst@linux.intel.com |
|---|---|
| Headers | show |
| Series | sound: Use -EPROBE_DEFER instead of i915 module loading. | expand |
On 2023-10-05 12:58, Péter Ujfalusi wrote: > > > On 04/10/2023 19:59, Kai Vehmanen wrote: >> Hi, >> >> I'm good with rest of the series, but one patch requires work. >> >> On Wed, 4 Oct 2023, Maarten Lankhorst wrote: >> >>> Now that we can use -EPROBE_DEFER, it's no longer required to spin off >>> the snd_hdac_i915_init into a workqueue. >>> >>> Use the -EPROBE_DEFER mechanism instead, which must be returned in the >>> probe function. >>> >>> The previously added probe_early can be used for this, >>> and we also use the newly added remove_late for unbinding afterwards. >> [...] >>> --- a/sound/soc/sof/intel/hda-common-ops.c >>> +++ b/sound/soc/sof/intel/hda-common-ops.c >>> @@ -19,6 +19,7 @@ struct snd_sof_dsp_ops sof_hda_common_ops = { >>> .probe_early = hda_dsp_probe_early, >>> .probe = hda_dsp_probe, >>> .remove = hda_dsp_remove, >>> + .remove_late = hda_dsp_remove_late, >>> >>> /* Register IO uses direct mmio */ >>> >>> diff --git a/sound/soc/sof/intel/hda.c b/sound/soc/sof/intel/hda.c >>> index 86a2571488bc..4eb7f04b8ae1 100644 >>> --- a/sound/soc/sof/intel/hda.c >>> +++ b/sound/soc/sof/intel/hda.c >>> @@ -1160,6 +1160,7 @@ int hda_dsp_probe_early(struct snd_sof_dev *sdev) >>> return -ENOMEM; >>> sdev->pdata->hw_pdata = hdev; >>> hdev->desc = chip; >>> + ret = hda_init(sdev); >>> >>> err: >>> return ret; >> >> I don't think this works. The hda_codec_i915_init() errors are ignored in >> hda_init() so this never returns -EPROBE_DEFER. >> >> So something like this is needed on top (tested quickly on one SOF >> machine and this blocks SOF load until i915 or xe driver is loaded): >> >> --cut-- >> diff --git a/sound/soc/sof/intel/hda.c b/sound/soc/sof/intel/hda.c >> index 9025bfaf6a7e..8b17c82dcc89 100644 >> --- a/sound/soc/sof/intel/hda.c >> +++ b/sound/soc/sof/intel/hda.c >> @@ -863,13 +863,20 @@ static int hda_init(struct snd_sof_dev *sdev) >> /* init i915 and HDMI codecs */ >> ret = hda_codec_i915_init(sdev); >> if (ret < 0) >> - dev_warn(sdev->dev, "init of i915 and HDMI codec >> failed\n"); >> + dev_warn(sdev->dev, "init of i915 and HDMI codec failed >> (%d)\n", ret); > > we should not print anything or maximum dev_dbg in case of EPROBE_DEFER. There's dev_err_probe, which is dev_err on error, or sets the reason for deferred probe to the arguments if the error is -EPROBE_DEFER. ~Maarten
Hey, On 2023-10-09 08:23, Takashi Iwai wrote: > On Thu, 05 Oct 2023 13:26:18 +0200, > Maarten Lankhorst wrote: >> >> >> >> On 2023-10-05 12:58, Péter Ujfalusi wrote: >>> >>> >>> On 04/10/2023 19:59, Kai Vehmanen wrote: >>>> Hi, >>>> >>>> I'm good with rest of the series, but one patch requires work. >>>> >>>> On Wed, 4 Oct 2023, Maarten Lankhorst wrote: >>>> >>>>> Now that we can use -EPROBE_DEFER, it's no longer required to spin off >>>>> the snd_hdac_i915_init into a workqueue. >>>>> >>>>> Use the -EPROBE_DEFER mechanism instead, which must be returned in the >>>>> probe function. >>>>> >>>>> The previously added probe_early can be used for this, >>>>> and we also use the newly added remove_late for unbinding afterwards. >>>> [...] >>>>> --- a/sound/soc/sof/intel/hda-common-ops.c >>>>> +++ b/sound/soc/sof/intel/hda-common-ops.c >>>>> @@ -19,6 +19,7 @@ struct snd_sof_dsp_ops sof_hda_common_ops = { >>>>> .probe_early = hda_dsp_probe_early, >>>>> .probe = hda_dsp_probe, >>>>> .remove = hda_dsp_remove, >>>>> + .remove_late = hda_dsp_remove_late, >>>>> /* Register IO uses direct mmio */ >>>>> diff --git a/sound/soc/sof/intel/hda.c >>>>> b/sound/soc/sof/intel/hda.c >>>>> index 86a2571488bc..4eb7f04b8ae1 100644 >>>>> --- a/sound/soc/sof/intel/hda.c >>>>> +++ b/sound/soc/sof/intel/hda.c >>>>> @@ -1160,6 +1160,7 @@ int hda_dsp_probe_early(struct snd_sof_dev *sdev) >>>>> return -ENOMEM; >>>>> sdev->pdata->hw_pdata = hdev; >>>>> hdev->desc = chip; >>>>> + ret = hda_init(sdev); >>>>> err: >>>>> return ret; >>>> >>>> I don't think this works. The hda_codec_i915_init() errors are ignored in >>>> hda_init() so this never returns -EPROBE_DEFER. >>>> >>>> So something like this is needed on top (tested quickly on one SOF >>>> machine and this blocks SOF load until i915 or xe driver is loaded): >>>> >>>> --cut-- >>>> diff --git a/sound/soc/sof/intel/hda.c b/sound/soc/sof/intel/hda.c >>>> index 9025bfaf6a7e..8b17c82dcc89 100644 >>>> --- a/sound/soc/sof/intel/hda.c >>>> +++ b/sound/soc/sof/intel/hda.c >>>> @@ -863,13 +863,20 @@ static int hda_init(struct snd_sof_dev *sdev) >>>> /* init i915 and HDMI codecs */ >>>> ret = hda_codec_i915_init(sdev); >>>> if (ret < 0) >>>> - dev_warn(sdev->dev, "init of i915 and HDMI codec >>>> failed\n"); >>>> + dev_warn(sdev->dev, "init of i915 and HDMI codec failed >>>> (%d)\n", ret); >>> >>> we should not print anything or maximum dev_dbg in case of EPROBE_DEFER. >> There's dev_err_probe, which is dev_err on error, or sets the reason >> for deferred probe to the arguments if the error is -EPROBE_DEFER. > > I expect you'll respin v7 for addressing this? > > I'd love to merge the series for 6.7, and the time ticks... Done, added the error handling early in the series as a bugfix. Cheers, ~Maarten
From: Maarten Lankhorst <dev@lankhorst.se> Explicitly loading i915 becomes a problem when upstreaming the new intel driver for Tiger Lake and higher graphics (xe). By loading i915, it doesn't wait for driver load of xe, and will fail completely before it loads. -EPROBE_DEFER has to be returned before any device is created in probe(), otherwise the removal of the device will cause EPROBE_DEFER to try again in an infinite loop. The conversion is done in gradual steps. First I add an argument to snd_hdac_i915_init to allow for -EPROBE_DEFER so I can convert each driver separately. Then I convert each driver to move snd_hdac_i915_init out of the workqueue. Finally I drop the ability to choose modprobe behavior after the last user is converted. Compared to previous version, I added a fix for sof_ops_free() missing call, renamed probe_no_wq and remove_no_wq to probe_early/probe_late, and fixed the resulting fallout. Cc: Jaroslav Kysela <perex@perex.cz> Cc: Takashi Iwai <tiwai@suse.com> Cc: Cezary Rojewski <cezary.rojewski@intel.com> Cc: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Cc: Liam Girdwood <liam.r.girdwood@linux.intel.com> Cc: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Cc: Bard Liao <yung-chuan.liao@linux.intel.com> Cc: Ranjani Sridharan <ranjani.sridharan@linux.intel.com> Cc: Kai Vehmanen <kai.vehmanen@linux.intel.com> Cc: Mark Brown <broonie@kernel.org> Cc: Daniel Baluta <daniel.baluta@nxp.com> Cc: alsa-devel@alsa-project.org Cc: linux-kernel@vger.kernel.org Cc: sound-open-firmware@alsa-project.org Maarten Lankhorst (10): ASoC: SOF: core: Ensure sof_ops_free() is still called when probe never ran. ALSA: hda: Intel: Fix error handling in azx_probe() ALSA: hda: i915: Allow override of gpu binding. ALSA: hda: i915: Add an allow_modprobe argument to snd_hdac_i915_init ALSA: hda: i915: Allow xe as match for i915_component_master_match ASoC: Intel: avs: Move snd_hdac_i915_init to before probe_work. ALSA: hda: Intel: Move snd_hdac_i915_init to before probe_work. ASoC: Intel: Skylake: Move snd_hdac_i915_init to before probe_work. ASoC: SOF: Intel: Move binding to display driver outside of deferred probe ALSA: hda: i915: Remove extra argument from snd_hdac_i915_init Pierre-Louis Bossart (2): ASoC: SOF: core: Add probe_early and remove_late callbacks ASoC: SOF: Intel: hda: start splitting the probe sound/hda/hdac_i915.c | 24 ++++++----- sound/pci/hda/hda_intel.c | 60 ++++++++++++++-------------- sound/soc/intel/avs/core.c | 13 ++++-- sound/soc/intel/skylake/skl.c | 31 +++++--------- sound/soc/sof/core.c | 17 +++++++- sound/soc/sof/intel/hda-common-ops.c | 2 + sound/soc/sof/intel/hda.c | 32 ++++++++------- sound/soc/sof/intel/hda.h | 2 + sound/soc/sof/ops.h | 16 ++++++++ sound/soc/sof/sof-priv.h | 2 + 10 files changed, 118 insertions(+), 81 deletions(-)