ASoC: Intel: Skylake: Fix module removal
Subsystems (3)
v2 cover ASoC: Intel: Skylake: Fix module removal
Skylake driver were crashing on module removal when we run "alsa reload".
The first issues was caused by component framework regression which Russell fixed and in in 4.5. In driver there were still some missing order which needs to be fixed. This series first update the list parsing in core, then ensure proper freeup of code objects, and then ensuring driver does proper cleanup and i915 dependency
Changes in v2:
- Add acked by Takashi
- Remove codec destructor wrapper
Vinod Koul (6):
ALSA: hda: use list macro for parsing on cleanup
ASoC: Intel: Skylake: free codec objects on removal
ASoC: Intel: Skylake: Freeup properly on skl_dsp_free
ASoC: Intel: Skylake: Unmap the address last
ASoC: Intel: Skylake: Call i915 exit last
ASoC: Intel: Skylake: remove call to pci_dev_put
sound/hda/ext/hdac_ext_stream.c | 5 ++--- sound/soc/intel/skylake/skl-sst-dsp.c | 5 +++++ sound/soc/intel/skylake/skl.c | 17 ++++++++++------- 3 files changed, 17 insertions(+), 10 deletions(-)
Signature · 1 line
1.9.1
PATCH v2 1/6 ALSA: hda: use list macro for parsing on cleanup
It is always better to use list_for_each_entry_safe() while doing cleanup. So use this instead of open coding this in list in snd_hdac_stream_free_all()
--- sound/hda/ext/hdac_ext_stream.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
sound/hda/ext/hdac_ext_stream.c +2 -3
@@ -104,12 +104,11 @@ EXPORT_SYMBOL_GPL(snd_hdac_ext_stream_init_all);*/ void snd_hdac_stream_free_all (struct hdac_ext_bus *ebus) { -struct hdac_stream *s; +struct hdac_stream *s, *_s;struct hdac_ext_stream *stream;struct hdac_bus *bus =ebus_to_hbus (ebus); -while (!list_empty (&bus->stream_list)) { - s = list_first_entry(&bus->stream_list, struct hdac_stream, list); +list_for_each_entry_safe (s, _s, &bus->stream_list, list) { stream =stream_to_hdac_ext_stream (s);snd_hdac_ext_stream_decouple (ebus, stream, false);list_del (&s->list);
Signature · 1 line
1.9.1
PATCH v2 2/6 ASoC: Intel: Skylake: free codec objects on removal
In reply to Vinod KoulOn driver removal we should ask the core to remove the device objects as well, so invoke snd_hdac_ext_bus_device_remove() in remove.
--- sound/soc/intel/skylake/skl.c | 4 ++++ 1 file changed, 4 insertions(+)
sound/soc/intel/skylake/skl.c +4 -0
@@ -725,6 +725,10 @@ static void skl_remove(struct pci_dev *pci)if (pci_dev_run_wake (pci))pm_runtime_get_noresume (&pci->dev);pci_dev_put (pci); + +/* codec removal, invoke bus_device_remove */ +snd_hdac_ext_bus_device_remove (ebus); +skl_platform_unregister (&pci->dev);skl_free_dsp (skl);skl_machine_device_unregister (skl);
Signature · 1 line
1.9.1
Mark Brown
maintainerApplied "ASoC: Intel: Skylake: free codec objects on removal" to the asoc tree
The patch
ASoC: Intel: Skylake: free codec objects on removal
has been applied to the asoc tree at
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
7373f481dc4098a844a756201e98341bc56baaa2Mon Sep 17 00:00:00 2001
From: Vinod Koul <[email protected]>
Date: Tue, 15 Mar 2016 16:39:24 +0530
Subject: [PATCH] ASoC: Intel: Skylake: free codec objects on removal
On driver removal we should ask the core to remove the device objects as well, so invoke snd_hdac_ext_bus_device_remove() in remove.
--- sound/soc/intel/skylake/skl.c | 4 ++++ 1 file changed, 4 insertions(+)
sound/soc/intel/skylake/skl.c +4 -0
@@ -725,6 +725,10 @@ static void skl_remove(struct pci_dev *pci)if (pci_dev_run_wake (pci))pm_runtime_get_noresume (&pci->dev);pci_dev_put (pci); + +/* codec removal, invoke bus_device_remove */ +snd_hdac_ext_bus_device_remove (ebus); +skl_platform_unregister (&pci->dev);skl_free_dsp (skl);skl_machine_device_unregister (skl);
Signature · 1 line
2.7.0
PATCH v2 3/6 ASoC: Intel: Skylake: Freeup properly on skl_dsp_free
In reply to Vinod KoulWe are supposed to freeup the Code loader DMA allocation and ensure all interrupts are disabled before we disable dsp cores. So invoke these to ensure DSP shuts down properly.
--- sound/soc/intel/skylake/skl-sst-dsp.c | 5 +++++ 1 file changed, 5 insertions(+)
sound/soc/intel/skylake/skl-sst-dsp.c +5 -0
@@ -336,6 +336,11 @@ void skl_dsp_free(struct sst_dsp *dsp)skl_ipc_int_disable (dsp);free_irq (dsp->irq, dsp); + dsp->cl_dev.ops.cl_cleanup_controller (dsp); +skl_cldma_int_disable (dsp); +skl_ipc_op_int_disable (dsp); +skl_ipc_int_disable (dsp); +skl_dsp_disable_core (dsp); }EXPORT_SYMBOL_GPL (skl_dsp_free);
Signature · 1 line
1.9.1
PATCH v2 4/6 ASoC: Intel: Skylake: Unmap the address last
In reply to Vinod KoulIn Skylake destructor we unmap the hardware address and then free links and streams. The stream free accesses hardware to write to registers and predictably causes oops.
So change the order and unmap last in destructor.
--- sound/soc/intel/skylake/skl.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
sound/soc/intel/skylake/skl.c +4 -3
@@ -316,12 +316,13 @@ static int skl_free(struct hdac_ext_bus *ebus)if (bus->irq >=0 )free_irq (bus->irq, (void *)bus); -if (bus->remap_addr) -iounmap (bus->remap_addr); -snd_hdac_bus_free_stream_pages (bus);snd_hdac_stream_free_all (ebus);snd_hdac_link_free_all (ebus); + +if (bus->remap_addr) +iounmap (bus->remap_addr); +pci_release_regions (skl->pci);pci_disable_device (skl->pci);
Signature · 1 line
1.9.1
PATCH v2 5/6 ASoC: Intel: Skylake: Call i915 exit last
In reply to Vinod KoulThe Skylake driver uses i915 component APIs to talk to display. On remove we should free up by invoking snd_hdac_i915_exit() but that should be last thing in remove routine, so move it to last in skl_free()
--- sound/soc/intel/skylake/skl.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
sound/soc/intel/skylake/skl.c +2 -3
@@ -328,6 +328,8 @@ static int skl_free(struct hdac_ext_bus *ebus)snd_hdac_ext_bus_exit (ebus); +if (IS_ENABLED (CONFIG_SND_SOC_HDAC_HDMI )) +snd_hdac_i915_exit (&ebus->bus);return 0 ; }
@@ -720,9 +722,6 @@ static void skl_remove(struct pci_dev *pci)if (skl->tplg)release_firmware (skl->tplg); -if (IS_ENABLED (CONFIG_SND_SOC_HDAC_HDMI )) -snd_hdac_i915_exit (&ebus->bus); -if (pci_dev_run_wake (pci))pm_runtime_get_noresume (&pci->dev);pci_dev_put (pci);
Signature · 1 line
1.9.1
PATCH v2 6/6 ASoC: Intel: Skylake: remove call to pci_dev_put
In reply to Vinod KoulIn driver remove we call pci_dev_put(), that is not required as we never call pci_dev_get() so remove that.
--- sound/soc/intel/skylake/skl.c | 1 - 1 file changed, 1 deletion(-)
sound/soc/intel/skylake/skl.c +0 -1
@@ -724,7 +724,6 @@ static void skl_remove(struct pci_dev *pci)if (pci_dev_run_wake (pci))pm_runtime_get_noresume (&pci->dev); -pci_dev_put (pci);/* codec removal, invoke bus_device_remove */ snd_hdac_ext_bus_device_remove (ebus);
Signature · 1 line
1.9.1
Mark Brown
maintainerRe: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 2 lines · Vinod Koul
In driver remove we call
pci_dev_put(), that is not required as we never callpci_dev_get()so remove that.
Why is the fix for this not to call pci_dev_get()?
Re: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 5 lines · Mark Brown
Quoted 2 lines · Vinod Koul
In driver remove we call
pci_dev_put(), that is not required as we never callpci_dev_get()so remove that.Why is the fix for this not to call
pci_dev_get()?
Why do I need either, I see no reason why driver should be doing this, so removed :)
Signature · 1 line
~Vinod
Mark Brown
maintainerRe: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 6 lines · Vinod Koul
Quoted 1 line · Mark Brown
Why is the fix for this not to call
pci_dev_get()?Why do I need either, I see no reason why driver should be doing this, so removed :)
Well, the PCI documentation says that drivers are expected to record a reference to their devices in probe(). This is a bit unusual given that normally the driver core takes a reference to the device for us but presumably there's some reason for this?
Re: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 6 lines · Mark Brown
Quoted 3 lines · Vinod Koul
Quoted 1 line · Mark Brown
Why is the fix for this not to call
pci_dev_get()?
Quoted 7 lines
Quoted 2 lines
Why do I need either, I see no reason why driver should be doing this, so removed :)
Well, the PCI documentation says that drivers are expected to record a reference to their devices in
probe(). This is a bit unusual given that normally the driver core takes a reference to the device for us but presumably there's some reason for this?
Maybe the document is obsoleted. The PCI core, at least the probe / remove via the normal PCI bus, takes pci_dev_get() and pci_dev_put() already there.
Takashi
Re: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 11 lines · Takashi Iwai
Quoted 9 lines · Mark Brown
Quoted 6 lines · Vinod Koul
Quoted 1 line · Mark Brown
Why is the fix for this not to call
pci_dev_get()?Why do I need either, I see no reason why driver should be doing this, so removed :)
Quoted 8 lines
Quoted 4 lines
Well, the PCI documentation says that drivers are expected to record a reference to their devices in
probe(). This is a bit unusual given that normally the driver core takes a reference to the device for us but presumably there's some reason for this?Maybe the document is obsoleted. The PCI core, at least the probe / remove via the normal PCI bus, takes
pci_dev_get()andpci_dev_put()already there.
Yes that is my understanding too, that is why we removed this from driver here..
Mark, is this fine now?
Signature · 1 line
~Vinod
Mark Brown
maintainerRe: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 10 lines · Vinod Koul
Quoted 3 lines · Takashi Iwai
Maybe the document is obsoleted. The PCI core, at least the probe / remove via the normal PCI bus, takes
pci_dev_get()andpci_dev_put()already there.Yes that is my understanding too, that is why we removed this from driver here..
Mark, is this fine now?
You might want to put some of this analysis in the changelog.
Re: [PATCH v2 6/6] ASoC: Intel: Skylake: remove call to pci_dev_put
Quoted 7 lines · Mark Brown
Quoted 5 lines · Vinod Koul
Quoted 3 lines · Takashi Iwai
Maybe the document is obsoleted. The PCI core, at least the probe / remove via the normal PCI bus, takes
pci_dev_get()andpci_dev_put()already there.
Quoted 6 lines
Quoted 4 lines
Yes that is my understanding too, that is why we removed this from driver here..
Mark, is this fine now?
You might want to put some of this analysis in the changelog.
Yup makes sense, v3 on the way then !
Signature · 1 line
~Vinod
Mark Brown
maintainerApplied "ASoC: Intel: Skylake: remove call to pci_dev_put" to the asoc tree
In reply to Vinod KoulThe patch
ASoC: Intel: Skylake: remove call to pci_dev_put
has been applied to the asoc tree at
git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
36e7972c0d3f8819a5d9335c36c5dcd168cd2b72Mon Sep 17 00:00:00 2001
From: Vinod Koul <[email protected]>
Date: Wed, 16 Mar 2016 21:51:31 +0530
Subject: [PATCH] ASoC: Intel: Skylake: remove call to pci_dev_put
The PCI bus takes pci_dev_get() and pci_dev_put() is also there. So no need for drivers to invoke these. In SKL driver we were calling pci_dev_put() only which is not right, so remove this
--- sound/soc/intel/skylake/skl.c | 1 - 1 file changed, 1 deletion(-)
sound/soc/intel/skylake/skl.c +0 -1
@@ -724,7 +724,6 @@ static void skl_remove(struct pci_dev *pci)if (pci_dev_run_wake (pci))pm_runtime_get_noresume (&pci->dev); -pci_dev_put (pci);/* codec removal, invoke bus_device_remove */ snd_hdac_ext_bus_device_remove (ebus);
Signature · 1 line
2.7.0