diff mbox

ACPI/IORT: Make dma masks set-up IORT specific

Message ID 20161205122619.25045-1-lorenzo.pieralisi@arm.com (mailing list archive)
State New, archived
Headers show

Commit Message

Lorenzo Pieralisi Dec. 5, 2016, 12:26 p.m. UTC
The introduction of acpi_dma_configure() allows to configure DMA
and related IOMMU for any device that is DMA capable. To achieve
that goal it ensures DMA masks are set-up to sane default values
before proceeding with IOMMU and DMA ops configuration.

On x86/ia64 systems, through acpi_bind_one(), acpi_dma_configure() is
called for every device that has an ACPI companion, in that every device
is considered DMA capable on x86/ia64 systems (ie acpi_get_dma_attr() API),
which has the side effect of initializing dma masks also for
pseudo-devices (eg CPUs and memory nodes) and potentially for devices
whose dma masks were not set-up before the acpi_dma_configure() API was
introduced, which may have noxious side effects.

Therefore, in preparation for IORT firmware specific DMA masks set-up,
wrap the default DMA masks set-up in acpi_dma_configure() inside an IORT
specific wrapper that reverts to a NOP on x86/ia64 systems, restoring the
default expected behaviour on x86/ia64 systems and keeping DMA default
masks set-up on IORT based (ie ARM) arch configurations.

Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Hanjun Guo <hanjun.guo@linaro.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Tomasz Nowicki <tn@semihalf.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Sricharan R <sricharan@codeaurora.org>
---
Joerg,

pending Rafael's ACK on it, given the 4.10 release timing and that the
series is queued via the IOMMU tree please consider applying this patch to
your arm/smmu branch for 4.10, it is not fixing a bug but it is modifying
the x86/ia64 code path; I prefer preventing any issue related to default
dma masks on x86/ia64 so I hope it can get merged along with the rest of
the ACPI IORT SMMU series.

Thanks a lot and apologies,
Lorenzo

 drivers/acpi/arm64/iort.c | 22 ++++++++++++++++++++++
 drivers/acpi/scan.c       | 14 +-------------
 include/linux/acpi_iort.h |  2 ++
 3 files changed, 25 insertions(+), 13 deletions(-)

Comments

Rafael J. Wysocki Dec. 5, 2016, 10:18 p.m. UTC | #1
On Mon, Dec 5, 2016 at 1:26 PM, Lorenzo Pieralisi
<lorenzo.pieralisi@arm.com> wrote:
> The introduction of acpi_dma_configure() allows to configure DMA
> and related IOMMU for any device that is DMA capable. To achieve
> that goal it ensures DMA masks are set-up to sane default values
> before proceeding with IOMMU and DMA ops configuration.
>
> On x86/ia64 systems, through acpi_bind_one(), acpi_dma_configure() is
> called for every device that has an ACPI companion, in that every device
> is considered DMA capable on x86/ia64 systems (ie acpi_get_dma_attr() API),
> which has the side effect of initializing dma masks also for
> pseudo-devices (eg CPUs and memory nodes) and potentially for devices
> whose dma masks were not set-up before the acpi_dma_configure() API was
> introduced, which may have noxious side effects.
>
> Therefore, in preparation for IORT firmware specific DMA masks set-up,
> wrap the default DMA masks set-up in acpi_dma_configure() inside an IORT
> specific wrapper that reverts to a NOP on x86/ia64 systems, restoring the
> default expected behaviour on x86/ia64 systems and keeping DMA default
> masks set-up on IORT based (ie ARM) arch configurations.
>
> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Will Deacon <will.deacon@arm.com>
> Cc: Hanjun Guo <hanjun.guo@linaro.org>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: Robin Murphy <robin.murphy@arm.com>
> Cc: Tomasz Nowicki <tn@semihalf.com>
> Cc: Joerg Roedel <joro@8bytes.org>
> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
> Cc: Sricharan R <sricharan@codeaurora.org>

Acked -by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

> ---
> Joerg,
>
> pending Rafael's ACK on it, given the 4.10 release timing and that the
> series is queued via the IOMMU tree please consider applying this patch to
> your arm/smmu branch for 4.10, it is not fixing a bug but it is modifying
> the x86/ia64 code path; I prefer preventing any issue related to default
> dma masks on x86/ia64 so I hope it can get merged along with the rest of
> the ACPI IORT SMMU series.
>
> Thanks a lot and apologies,
> Lorenzo
>
>  drivers/acpi/arm64/iort.c | 22 ++++++++++++++++++++++
>  drivers/acpi/scan.c       | 14 +-------------
>  include/linux/acpi_iort.h |  2 ++
>  3 files changed, 25 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
> index 47bace8..e0d2e6e 100644
> --- a/drivers/acpi/arm64/iort.c
> +++ b/drivers/acpi/arm64/iort.c
> @@ -547,6 +547,28 @@ static const struct iommu_ops *iort_iommu_xlate(struct device *dev,
>  }
>
>  /**
> + * iort_set_dma_mask - Set-up dma mask for a device.
> + *
> + * @dev: device to configure
> + */
> +void iort_set_dma_mask(struct device *dev)
> +{
> +       /*
> +        * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
> +        * setup the correct supported mask.
> +        */
> +       if (!dev->coherent_dma_mask)
> +               dev->coherent_dma_mask = DMA_BIT_MASK(32);
> +
> +       /*
> +        * Set it to coherent_dma_mask by default if the architecture
> +        * code has not set it.
> +        */
> +       if (!dev->dma_mask)
> +               dev->dma_mask = &dev->coherent_dma_mask;
> +}
> +
> +/**
>   * iort_iommu_configure - Set-up IOMMU configuration for a device.
>   *
>   * @dev: device to configure
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 80698d3..93b00cf 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -1380,19 +1380,7 @@ void acpi_dma_configure(struct device *dev, enum dev_dma_attr attr)
>  {
>         const struct iommu_ops *iommu;
>
> -       /*
> -        * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
> -        * setup the correct supported mask.
> -        */
> -       if (!dev->coherent_dma_mask)
> -               dev->coherent_dma_mask = DMA_BIT_MASK(32);
> -
> -       /*
> -        * Set it to coherent_dma_mask by default if the architecture
> -        * code has not set it.
> -        */
> -       if (!dev->dma_mask)
> -               dev->dma_mask = &dev->coherent_dma_mask;
> +       iort_set_dma_mask(dev);
>
>         iommu = iort_iommu_configure(dev);
>
> diff --git a/include/linux/acpi_iort.h b/include/linux/acpi_iort.h
> index dcb2b60..77e0809 100644
> --- a/include/linux/acpi_iort.h
> +++ b/include/linux/acpi_iort.h
> @@ -35,6 +35,7 @@ bool iort_node_match(u8 type);
>  u32 iort_msi_map_rid(struct device *dev, u32 req_id);
>  struct irq_domain *iort_get_device_domain(struct device *dev, u32 req_id);
>  /* IOMMU interface */
> +void iort_set_dma_mask(struct device *dev);
>  const struct iommu_ops *iort_iommu_configure(struct device *dev);
>  #else
>  static inline void acpi_iort_init(void) { }
> @@ -45,6 +46,7 @@ static inline struct irq_domain *iort_get_device_domain(struct device *dev,
>                                                         u32 req_id)
>  { return NULL; }
>  /* IOMMU interface */
> +static inline void iort_set_dma_mask(struct device *dev) { }
>  static inline
>  const struct iommu_ops *iort_iommu_configure(struct device *dev)
>  { return NULL; }
> --
> 2.10.0
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
Hanjun Guo Dec. 6, 2016, 7:55 a.m. UTC | #2
On 2016/12/5 20:26, Lorenzo Pieralisi wrote:
> The introduction of acpi_dma_configure() allows to configure DMA
> and related IOMMU for any device that is DMA capable. To achieve
> that goal it ensures DMA masks are set-up to sane default values
> before proceeding with IOMMU and DMA ops configuration.
>
> On x86/ia64 systems, through acpi_bind_one(), acpi_dma_configure() is
> called for every device that has an ACPI companion, in that every device
> is considered DMA capable on x86/ia64 systems (ie acpi_get_dma_attr() API),
> which has the side effect of initializing dma masks also for
> pseudo-devices (eg CPUs and memory nodes) and potentially for devices
> whose dma masks were not set-up before the acpi_dma_configure() API was
> introduced, which may have noxious side effects.
>
> Therefore, in preparation for IORT firmware specific DMA masks set-up,
> wrap the default DMA masks set-up in acpi_dma_configure() inside an IORT
> specific wrapper that reverts to a NOP on x86/ia64 systems, restoring the
> default expected behaviour on x86/ia64 systems and keeping DMA default
> masks set-up on IORT based (ie ARM) arch configurations.
>
> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Will Deacon <will.deacon@arm.com>
> Cc: Hanjun Guo <hanjun.guo@linaro.org>

Add this patch on top of your v9 acpi smmu patchset,
tested on Hisilicon D03 (ARM64), devices with SMMU
enabled work fine,

Tested-by: Hanjun Guo <hanjun.guo@linaro.org>
Reviewed-by: Hanjun Guo <hanjun.guo@linaro.org>

Thanks
Hanjun
Lorenzo Pieralisi Dec. 6, 2016, 9:37 a.m. UTC | #3
Hi Joerg,

On Mon, Dec 05, 2016 at 11:18:56PM +0100, Rafael J. Wysocki wrote:
> On Mon, Dec 5, 2016 at 1:26 PM, Lorenzo Pieralisi
> <lorenzo.pieralisi@arm.com> wrote:
> > The introduction of acpi_dma_configure() allows to configure DMA
> > and related IOMMU for any device that is DMA capable. To achieve
> > that goal it ensures DMA masks are set-up to sane default values
> > before proceeding with IOMMU and DMA ops configuration.
> >
> > On x86/ia64 systems, through acpi_bind_one(), acpi_dma_configure() is
> > called for every device that has an ACPI companion, in that every device
> > is considered DMA capable on x86/ia64 systems (ie acpi_get_dma_attr() API),
> > which has the side effect of initializing dma masks also for
> > pseudo-devices (eg CPUs and memory nodes) and potentially for devices
> > whose dma masks were not set-up before the acpi_dma_configure() API was
> > introduced, which may have noxious side effects.
> >
> > Therefore, in preparation for IORT firmware specific DMA masks set-up,
> > wrap the default DMA masks set-up in acpi_dma_configure() inside an IORT
> > specific wrapper that reverts to a NOP on x86/ia64 systems, restoring the
> > default expected behaviour on x86/ia64 systems and keeping DMA default
> > masks set-up on IORT based (ie ARM) arch configurations.
> >
> > Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> > Cc: Will Deacon <will.deacon@arm.com>
> > Cc: Hanjun Guo <hanjun.guo@linaro.org>
> > Cc: Bjorn Helgaas <bhelgaas@google.com>
> > Cc: Robin Murphy <robin.murphy@arm.com>
> > Cc: Tomasz Nowicki <tn@semihalf.com>
> > Cc: Joerg Roedel <joro@8bytes.org>
> > Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
> > Cc: Sricharan R <sricharan@codeaurora.org>
> 
> Acked -by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> 
> > ---
> > Joerg,
> >
> > pending Rafael's ACK on it, given the 4.10 release timing and that the
> > series is queued via the IOMMU tree please consider applying this patch to
> > your arm/smmu branch for 4.10, it is not fixing a bug but it is modifying
> > the x86/ia64 code path; I prefer preventing any issue related to default
> > dma masks on x86/ia64 so I hope it can get merged along with the rest of
> > the ACPI IORT SMMU series.

I can apply Rafael and Hanjun's tags and resend a v2 to you if you
prefer, it would be great if you could apply this patch to your arm/smmu
branch for v4.10 as per description above, please let me know.

Thank you !
Lorenzo

> >
> > Thanks a lot and apologies,
> > Lorenzo
> >
> >  drivers/acpi/arm64/iort.c | 22 ++++++++++++++++++++++
> >  drivers/acpi/scan.c       | 14 +-------------
> >  include/linux/acpi_iort.h |  2 ++
> >  3 files changed, 25 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
> > index 47bace8..e0d2e6e 100644
> > --- a/drivers/acpi/arm64/iort.c
> > +++ b/drivers/acpi/arm64/iort.c
> > @@ -547,6 +547,28 @@ static const struct iommu_ops *iort_iommu_xlate(struct device *dev,
> >  }
> >
> >  /**
> > + * iort_set_dma_mask - Set-up dma mask for a device.
> > + *
> > + * @dev: device to configure
> > + */
> > +void iort_set_dma_mask(struct device *dev)
> > +{
> > +       /*
> > +        * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
> > +        * setup the correct supported mask.
> > +        */
> > +       if (!dev->coherent_dma_mask)
> > +               dev->coherent_dma_mask = DMA_BIT_MASK(32);
> > +
> > +       /*
> > +        * Set it to coherent_dma_mask by default if the architecture
> > +        * code has not set it.
> > +        */
> > +       if (!dev->dma_mask)
> > +               dev->dma_mask = &dev->coherent_dma_mask;
> > +}
> > +
> > +/**
> >   * iort_iommu_configure - Set-up IOMMU configuration for a device.
> >   *
> >   * @dev: device to configure
> > diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> > index 80698d3..93b00cf 100644
> > --- a/drivers/acpi/scan.c
> > +++ b/drivers/acpi/scan.c
> > @@ -1380,19 +1380,7 @@ void acpi_dma_configure(struct device *dev, enum dev_dma_attr attr)
> >  {
> >         const struct iommu_ops *iommu;
> >
> > -       /*
> > -        * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
> > -        * setup the correct supported mask.
> > -        */
> > -       if (!dev->coherent_dma_mask)
> > -               dev->coherent_dma_mask = DMA_BIT_MASK(32);
> > -
> > -       /*
> > -        * Set it to coherent_dma_mask by default if the architecture
> > -        * code has not set it.
> > -        */
> > -       if (!dev->dma_mask)
> > -               dev->dma_mask = &dev->coherent_dma_mask;
> > +       iort_set_dma_mask(dev);
> >
> >         iommu = iort_iommu_configure(dev);
> >
> > diff --git a/include/linux/acpi_iort.h b/include/linux/acpi_iort.h
> > index dcb2b60..77e0809 100644
> > --- a/include/linux/acpi_iort.h
> > +++ b/include/linux/acpi_iort.h
> > @@ -35,6 +35,7 @@ bool iort_node_match(u8 type);
> >  u32 iort_msi_map_rid(struct device *dev, u32 req_id);
> >  struct irq_domain *iort_get_device_domain(struct device *dev, u32 req_id);
> >  /* IOMMU interface */
> > +void iort_set_dma_mask(struct device *dev);
> >  const struct iommu_ops *iort_iommu_configure(struct device *dev);
> >  #else
> >  static inline void acpi_iort_init(void) { }
> > @@ -45,6 +46,7 @@ static inline struct irq_domain *iort_get_device_domain(struct device *dev,
> >                                                         u32 req_id)
> >  { return NULL; }
> >  /* IOMMU interface */
> > +static inline void iort_set_dma_mask(struct device *dev) { }
> >  static inline
> >  const struct iommu_ops *iort_iommu_configure(struct device *dev)
> >  { return NULL; }
> > --
> > 2.10.0
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
Will Deacon Dec. 6, 2016, 9:43 a.m. UTC | #4
On Mon, Dec 05, 2016 at 12:26:19PM +0000, Lorenzo Pieralisi wrote:
> The introduction of acpi_dma_configure() allows to configure DMA
> and related IOMMU for any device that is DMA capable. To achieve
> that goal it ensures DMA masks are set-up to sane default values
> before proceeding with IOMMU and DMA ops configuration.
> 
> On x86/ia64 systems, through acpi_bind_one(), acpi_dma_configure() is
> called for every device that has an ACPI companion, in that every device
> is considered DMA capable on x86/ia64 systems (ie acpi_get_dma_attr() API),
> which has the side effect of initializing dma masks also for
> pseudo-devices (eg CPUs and memory nodes) and potentially for devices
> whose dma masks were not set-up before the acpi_dma_configure() API was
> introduced, which may have noxious side effects.
> 
> Therefore, in preparation for IORT firmware specific DMA masks set-up,
> wrap the default DMA masks set-up in acpi_dma_configure() inside an IORT
> specific wrapper that reverts to a NOP on x86/ia64 systems, restoring the
> default expected behaviour on x86/ia64 systems and keeping DMA default
> masks set-up on IORT based (ie ARM) arch configurations.
> 
> Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Will Deacon <will.deacon@arm.com>
> Cc: Hanjun Guo <hanjun.guo@linaro.org>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> Cc: Robin Murphy <robin.murphy@arm.com>
> Cc: Tomasz Nowicki <tn@semihalf.com>
> Cc: Joerg Roedel <joro@8bytes.org>
> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
> Cc: Sricharan R <sricharan@codeaurora.org>
> ---
> Joerg,
> 
> pending Rafael's ACK on it, given the 4.10 release timing and that the
> series is queued via the IOMMU tree please consider applying this patch to
> your arm/smmu branch for 4.10, it is not fixing a bug but it is modifying
> the x86/ia64 code path; I prefer preventing any issue related to default
> dma masks on x86/ia64 so I hope it can get merged along with the rest of
> the ACPI IORT SMMU series.
> 
> Thanks a lot and apologies,
> Lorenzo
> 
>  drivers/acpi/arm64/iort.c | 22 ++++++++++++++++++++++
>  drivers/acpi/scan.c       | 14 +-------------
>  include/linux/acpi_iort.h |  2 ++
>  3 files changed, 25 insertions(+), 13 deletions(-)

Looks straightforward to me:

Acked-by: Will Deacon <will.deacon@arm.com>

Joerg can probably just pick this on top of his queue.

Will
Joerg Roedel Dec. 6, 2016, 1:24 p.m. UTC | #5
Hi Lorenzo,

On Tue, Dec 06, 2016 at 09:37:10AM +0000, Lorenzo Pieralisi wrote:
> I can apply Rafael and Hanjun's tags and resend a v2 to you if you
> prefer, it would be great if you could apply this patch to your arm/smmu
> branch for v4.10 as per description above, please let me know.

Yes, please collect all the tags and send me a v2 please. I'll apply it
the right away.


Thanks,

	Joerg
Lorenzo Pieralisi Dec. 6, 2016, 3:18 p.m. UTC | #6
On Tue, Dec 06, 2016 at 02:24:48PM +0100, Joerg Roedel wrote:
> Hi Lorenzo,
> 
> On Tue, Dec 06, 2016 at 09:37:10AM +0000, Lorenzo Pieralisi wrote:
> > I can apply Rafael and Hanjun's tags and resend a v2 to you if you
> > prefer, it would be great if you could apply this patch to your arm/smmu
> > branch for v4.10 as per description above, please let me know.
> 
> Yes, please collect all the tags and send me a v2 please. I'll apply it
> the right away.

Thank you very much, I have just sent it.

Thanks !
Lorenzo
diff mbox

Patch

diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
index 47bace8..e0d2e6e 100644
--- a/drivers/acpi/arm64/iort.c
+++ b/drivers/acpi/arm64/iort.c
@@ -547,6 +547,28 @@  static const struct iommu_ops *iort_iommu_xlate(struct device *dev,
 }
 
 /**
+ * iort_set_dma_mask - Set-up dma mask for a device.
+ *
+ * @dev: device to configure
+ */
+void iort_set_dma_mask(struct device *dev)
+{
+	/*
+	 * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
+	 * setup the correct supported mask.
+	 */
+	if (!dev->coherent_dma_mask)
+		dev->coherent_dma_mask = DMA_BIT_MASK(32);
+
+	/*
+	 * Set it to coherent_dma_mask by default if the architecture
+	 * code has not set it.
+	 */
+	if (!dev->dma_mask)
+		dev->dma_mask = &dev->coherent_dma_mask;
+}
+
+/**
  * iort_iommu_configure - Set-up IOMMU configuration for a device.
  *
  * @dev: device to configure
diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 80698d3..93b00cf 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -1380,19 +1380,7 @@  void acpi_dma_configure(struct device *dev, enum dev_dma_attr attr)
 {
 	const struct iommu_ops *iommu;
 
-	/*
-	 * Set default coherent_dma_mask to 32 bit.  Drivers are expected to
-	 * setup the correct supported mask.
-	 */
-	if (!dev->coherent_dma_mask)
-		dev->coherent_dma_mask = DMA_BIT_MASK(32);
-
-	/*
-	 * Set it to coherent_dma_mask by default if the architecture
-	 * code has not set it.
-	 */
-	if (!dev->dma_mask)
-		dev->dma_mask = &dev->coherent_dma_mask;
+	iort_set_dma_mask(dev);
 
 	iommu = iort_iommu_configure(dev);
 
diff --git a/include/linux/acpi_iort.h b/include/linux/acpi_iort.h
index dcb2b60..77e0809 100644
--- a/include/linux/acpi_iort.h
+++ b/include/linux/acpi_iort.h
@@ -35,6 +35,7 @@  bool iort_node_match(u8 type);
 u32 iort_msi_map_rid(struct device *dev, u32 req_id);
 struct irq_domain *iort_get_device_domain(struct device *dev, u32 req_id);
 /* IOMMU interface */
+void iort_set_dma_mask(struct device *dev);
 const struct iommu_ops *iort_iommu_configure(struct device *dev);
 #else
 static inline void acpi_iort_init(void) { }
@@ -45,6 +46,7 @@  static inline struct irq_domain *iort_get_device_domain(struct device *dev,
 							u32 req_id)
 { return NULL; }
 /* IOMMU interface */
+static inline void iort_set_dma_mask(struct device *dev) { }
 static inline
 const struct iommu_ops *iort_iommu_configure(struct device *dev)
 { return NULL; }