Skip to contentSkip to conversation

irqchip/gic-v3: Enable non-coherent GIC designs probing

PATCH Superseded Lorenzo Pieralisi 5 participants 15 messages Jan 23, 2024 – Jun 7, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

v5 cover irqchip/gic-v3: Enable non-coherent GIC designs probing

This series is v5 of previous series:

v4: https://lore.kernel.org/all/[email protected]
v3: https://lore.kernel.org/all/[email protected]
v2: https://lore.kernel.org/all/[email protected]
v1: https://lore.kernel.org/all/[email protected]

v4 -> v5
- ACPICA patches merged for v6.8
- Refactored ACPI parsing code according to review
- Rebased against v6.8-rc1

v3 -> v4:
- Dropped patches [1-3], already merged
- Added Linuxized ACPICA changes accepted upstream
- Rebased against v6.7-rc3

v2 -> v3:
- Added ACPICA temporary changes and ACPI changes to implement ECR https://bugzilla.tianocore.org/show_bug.cgi?id=4557
- ACPI changes are for testing purposes - subject to ECR code first approval

v1 -> v2:
- Updated DT bindings as per feedback
- Updated patch[2] to use GIC quirks infrastructure

Original cover letter
---
The GICv3 architecture specifications provide a means for the system programmer to set the shareability and cacheability attributes the GIC components (redistributors and ITSes) use to drive memory transactions.

Albeit the architecture give control over shareability/cacheability memory transactions attributes (and barriers), it is allowed to connect the GIC interconnect ports to non-coherent memory ports on the interconnect, basically tying off shareability/cacheability "wires" and de-facto making the redistributors and ITSes non-coherent memory observers.

This series aims at starting a discussion over a possible solution to this problem, by adding to the GIC device tree bindings the standard dma-noncoherent property. The GIC driver uses the property to force the redistributors and ITSes shareability attributes to non-shareable, which consequently forces the driver to use CMOs on GIC memory tables.

On ARM DT DMA is default non-coherent, so the GIC driver can't rely on the generic DT dma-coherent/non-coherent property management layer (of_dma_is_coherent()) which would default all GIC designs in the field as non-coherent; it has to rely on ad-hoc dma-noncoherent property handling.

When a consistent approach is agreed upon for DT an equivalent binding will be put forward for ACPI based systems.

Lorenzo Pieralisi (1):
irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v5 1/1 irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
drivers/acpi/processor_core.c +15 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
 	return rv;
 }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
 static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
 {
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index fec1b58470df..a60c560ce891 100644
@@ -5591,6 +5591,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
 		goto node_err;
 	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
 	err = its_probe_one(its);
 	if (!err)
 		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 98b0329b7154..8cb8dff86c12 100644
@@ -2356,6 +2356,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
 		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
 		return -ENOMEM;
 	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_request_region(redist->base_address, redist->length, "GICR");
 
 	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2380,6 +2385,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
 		return -ENOMEM;
 	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
 	return 0;
 }
include/linux/acpi.h +3 -0
index b7165e52b3c6..4eedab0e51c3 100644
@@ -284,6 +284,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
 	return phys_id == PHYS_CPUID_INVALID;
 }
 
+
+int __init acpi_get_madt_revision(void);
+
 /* Validate the processor object's proc_id */
 bool acpi_duplicate_processor_id(int proc_id);
 /* Processor _CTS control */
Signature · 1 line
2.34.1
Reviewed-by on v5 1/1
#2 lore raw

Re: [PATCH v5 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

Quoted 21 lines · Lorenzo Pieralisi

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

Quoted 7 lines

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

FWIW,

I guess the most contentious parts disappeared in Marc's refactoring, so what's left seems entirely straightforward and innocuous to me.

I'd also agree that there seems no real value in going out of our way to police the firmware for consistency - if at least one entry says to apply the quirk, then applying the quirk is the right and safest thing to do, so all we could really do on top of that is add extra complexity to be able to say "hey, firmware did a silly thing!", but if we're starting from the premise that hardware did an inadvisable thing to begin with, maybe we should have similarly realistic expectations of the corresponding firmware anyway :)

Cheers,
Robin.

Quoted 94 lines
Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
---
  drivers/acpi/processor_core.c    | 15 +++++++++++++++
  drivers/irqchip/irq-gic-v3-its.c |  4 ++++
  drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
  include/linux/acpi.h             |  3 +++
  4 files changed, 31 insertions(+)
drivers/acpi/processor_core.c +15 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
  	return rv;
  }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
  static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
  {
  	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index fec1b58470df..a60c560ce891 100644
@@ -5591,6 +5591,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
  		goto node_err;
  	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
  	err = its_probe_one(its);
  	if (!err)
  		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 98b0329b7154..8cb8dff86c12 100644
@@ -2356,6 +2356,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
  		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
  		return -ENOMEM;
  	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
  	gic_request_region(redist->base_address, redist->length, "GICR");
 
  	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2380,6 +2385,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
  		return -ENOMEM;
  	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
  	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
  	return 0;
  }
include/linux/acpi.h +3 -0
index b7165e52b3c6..4eedab0e51c3 100644
@@ -284,6 +284,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
  	return phys_id == PHYS_CPUID_INVALID;
  }
 
+
+int __init acpi_get_madt_revision(void);
+
  /* Validate the processor object's proc_id */
  bool acpi_duplicate_processor_id(int proc_id);
  /* Processor _CTS control */
#3 lore raw

Re: [PATCH v5 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

In reply to Lorenzo Pieralisi
Quoted 34 lines · Lorenzo Pieralisi

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
Quoted 6 lines
---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)

Hi Marc, Rafael,

I would kindly ask you please what to do with this patch, it still applies to v6.9-rc5 - I can resend it if needed, ACPICA changes are already merged as-per the cover letter.

Thanks,
Lorenzo

Quoted 85 lines
drivers/acpi/processor_core.c +15 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
 	return rv;
 }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
 static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
 {
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index fec1b58470df..a60c560ce891 100644
@@ -5591,6 +5591,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
 		goto node_err;
 	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
 	err = its_probe_one(its);
 	if (!err)
 		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 98b0329b7154..8cb8dff86c12 100644
@@ -2356,6 +2356,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
 		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
 		return -ENOMEM;
 	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_request_region(redist->base_address, redist->length, "GICR");
 
 	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2380,6 +2385,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
 		return -ENOMEM;
 	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
 	return 0;
 }
include/linux/acpi.h +3 -0
index b7165e52b3c6..4eedab0e51c3 100644
@@ -284,6 +284,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
 	return phys_id == PHYS_CPUID_INVALID;
 }
 
+
+int __init acpi_get_madt_revision(void);
+
 /* Validate the processor object's proc_id */
 bool acpi_duplicate_processor_id(int proc_id);
 /* Processor _CTS control */
Signature · 1 line
2.34.1
#4 lore raw

Re: [PATCH v5 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

Quoted 42 lines · Lorenzo Pieralisi
Quoted 34 lines · Lorenzo Pieralisi

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
Quoted 6 lines
---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
Quoted 5 lines

Hi Marc, Rafael,

I would kindly ask you please what to do with this patch, it still applies to v6.9-rc5 - I can resend it if needed, ACPICA changes are already merged as-per the cover letter.

Hi Marc, Rafael,

I would kindly ask please what to do with this patch, rebased to v6.10-rc1, I can resend it if that's preferred, please let me know.

Thanks,
Lorenzo

Quoted 85 lines
Quoted 85 lines
drivers/acpi/processor_core.c +15 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
 	return rv;
 }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
 static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
 {
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index fec1b58470df..a60c560ce891 100644
@@ -5591,6 +5591,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
 		goto node_err;
 	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
 	err = its_probe_one(its);
 	if (!err)
 		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 98b0329b7154..8cb8dff86c12 100644
@@ -2356,6 +2356,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
 		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
 		return -ENOMEM;
 	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_request_region(redist->base_address, redist->length, "GICR");
 
 	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2380,6 +2385,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
 		return -ENOMEM;
 	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
 	return 0;
 }
include/linux/acpi.h +3 -0
index b7165e52b3c6..4eedab0e51c3 100644
@@ -284,6 +284,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
 	return phys_id == PHYS_CPUID_INVALID;
 }
 
+
+int __init acpi_get_madt_revision(void);
+
 /* Validate the processor object's proc_id */
 bool acpi_duplicate_processor_id(int proc_id);
 /* Processor _CTS control */
Signature · 1 line
2.34.1
maintainer Acked-by
#5 lore raw

Re: [PATCH v5 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

Quoted 45 lines · Lorenzo Pieralisi
Quoted 44 lines · Lorenzo Pieralisi
Quoted 34 lines · Lorenzo Pieralisi

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
Quoted 6 lines
---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)

Hi Marc, Rafael,

Quoted 8 lines
Quoted 3 lines

I would kindly ask you please what to do with this patch, it still applies to v6.9-rc5 - I can resend it if needed, ACPICA changes are already merged as-per the cover letter.

Hi Marc, Rafael,

I would kindly ask please what to do with this patch, rebased to v6.10-rc1, I can resend it if that's preferred, please let me know.

Please resend it with my:

and Cc'ing Thomas Gleixner.

Thanks,

        M.
Signature · 1 line
Jazz is not dead. It just smells funny...
#6 lore raw
In reply to Lorenzo Pieralisi
Quoted 11 lines · Lorenzo Pieralisi

This series is v5 of previous series:

v4: https://lore.kernel.org/all/[email protected]
v3: https://lore.kernel.org/all/[email protected]
v2: https://lore.kernel.org/all/[email protected]
v1: https://lore.kernel.org/all/[email protected]

v4 -> v5
- ACPICA patches merged for v6.8
- Refactored ACPI parsing code according to review
- Rebased against v6.8-rc1

Hi Marc, all,

this is not an urgent fix (I don't think there is any ACPI platform affected in the field so it is not even a fix), I am just asking please what should I do with it, I appreciate it is late in the cycle (and I know some fixes got merged in -rcX leading up to -rc7 that are pre-requisite for this patch to work).

Thanks,
Lorenzo

Quoted 56 lines

v3 -> v4:
- Dropped patches [1-3], already merged
- Added Linuxized ACPICA changes accepted upstream
- Rebased against v6.7-rc3

v2 -> v3:
- Added ACPICA temporary changes and ACPI changes to implement ECR https://bugzilla.tianocore.org/show_bug.cgi?id=4557
- ACPI changes are for testing purposes - subject to ECR code first approval

v1 -> v2:
- Updated DT bindings as per feedback
- Updated patch[2] to use GIC quirks infrastructure

Original cover letter
---
The GICv3 architecture specifications provide a means for the system programmer to set the shareability and cacheability attributes the GIC components (redistributors and ITSes) use to drive memory transactions.

Albeit the architecture give control over shareability/cacheability memory transactions attributes (and barriers), it is allowed to connect the GIC interconnect ports to non-coherent memory ports on the interconnect, basically tying off shareability/cacheability "wires" and de-facto making the redistributors and ITSes non-coherent memory observers.

This series aims at starting a discussion over a possible solution to this problem, by adding to the GIC device tree bindings the standard dma-noncoherent property. The GIC driver uses the property to force the redistributors and ITSes shareability attributes to non-shareable, which consequently forces the driver to use CMOs on GIC memory tables.

On ARM DT DMA is default non-coherent, so the GIC driver can't rely on the generic DT dma-coherent/non-coherent property management layer (of_dma_is_coherent()) which would default all GIC designs in the field as non-coherent; it has to rely on ad-hoc dma-noncoherent property handling.

When a consistent approach is agreed upon for DT an equivalent binding will be put forward for ACPI based systems.

Lorenzo Pieralisi (1):
irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
Signature · 1 line
2.34.1
#7 lore raw
Quoted 13 lines · Lorenzo Pieralisi
Quoted 11 lines · Lorenzo Pieralisi

This series is v5 of previous series:

v4: https://lore.kernel.org/all/[email protected]
v3: https://lore.kernel.org/all/[email protected]
v2: https://lore.kernel.org/all/[email protected]
v1: https://lore.kernel.org/all/[email protected]

v4 -> v5
- ACPICA patches merged for v6.8
- Refactored ACPI parsing code according to review
- Rebased against v6.8-rc1

Quoted 7 lines

Hi Marc, all,

this is not an urgent fix (I don't think there is any ACPI platform affected in the field so it is not even a fix), I am just asking please what should I do with it, I appreciate it is late in the cycle (and I know some fixes got merged in -rcX leading up to -rc7 that are pre-requisite for this patch to work).

Hi,

just a reminder to ask how to proceed with this patch, I know it is not urgent, just to understand how to handle it please.

The related ACPICA changes are already merged in the mainline.

Thanks,
Lorenzo

Quoted 59 lines

Thanks,
Lorenzo

Quoted 56 lines

v3 -> v4:
- Dropped patches [1-3], already merged
- Added Linuxized ACPICA changes accepted upstream
- Rebased against v6.7-rc3

v2 -> v3:
- Added ACPICA temporary changes and ACPI changes to implement ECR https://bugzilla.tianocore.org/show_bug.cgi?id=4557
- ACPI changes are for testing purposes - subject to ECR code first approval

v1 -> v2:
- Updated DT bindings as per feedback
- Updated patch[2] to use GIC quirks infrastructure

Original cover letter
---
The GICv3 architecture specifications provide a means for the system programmer to set the shareability and cacheability attributes the GIC components (redistributors and ITSes) use to drive memory transactions.

Albeit the architecture give control over shareability/cacheability memory transactions attributes (and barriers), it is allowed to connect the GIC interconnect ports to non-coherent memory ports on the interconnect, basically tying off shareability/cacheability "wires" and de-facto making the redistributors and ITSes non-coherent memory observers.

This series aims at starting a discussion over a possible solution to this problem, by adding to the GIC device tree bindings the standard dma-noncoherent property. The GIC driver uses the property to force the redistributors and ITSes shareability attributes to non-shareable, which consequently forces the driver to use CMOs on GIC memory tables.

On ARM DT DMA is default non-coherent, so the GIC driver can't rely on the generic DT dma-coherent/non-coherent property management layer (of_dma_is_coherent()) which would default all GIC designs in the field as non-coherent; it has to rely on ad-hoc dma-noncoherent property handling.

When a consistent approach is agreed upon for DT an equivalent binding will be put forward for ACPI based systems.

Lorenzo Pieralisi (1):
irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
Signature · 1 line
2.34.1
#8 lore raw

v6 cover irqchip/gic-v3: Enable non-coherent GIC designs probing

In reply to Lorenzo Pieralisi

This series is v6 of previous series:

v5: https://lore.kernel.org/all/[email protected]
v4: https://lore.kernel.org/all/[email protected]
v3: https://lore.kernel.org/all/[email protected]
v2: https://lore.kernel.org/all/[email protected]
v1: https://lore.kernel.org/all/[email protected]

v5 -> v6
- Rebased against v6.10-rc1
- Added Marc Zyngier's ACK

v4 -> v5
- ACPICA patches merged for v6.8
- Refactored ACPI parsing code according to review
- Rebased against v6.8-rc1

v3 -> v4:
- Dropped patches [1-3], already merged
- Added Linuxized ACPICA changes accepted upstream
- Rebased against v6.7-rc3

v2 -> v3:
- Added ACPICA temporary changes and ACPI changes to implement ECR https://bugzilla.tianocore.org/show_bug.cgi?id=4557
- ACPI changes are for testing purposes - subject to ECR code first approval

v1 -> v2:
- Updated DT bindings as per feedback
- Updated patch[2] to use GIC quirks infrastructure

Original cover letter
---
The GICv3 architecture specifications provide a means for the system programmer to set the shareability and cacheability attributes the GIC components (redistributors and ITSes) use to drive memory transactions.

Albeit the architecture give control over shareability/cacheability memory transactions attributes (and barriers), it is allowed to connect the GIC interconnect ports to non-coherent memory ports on the interconnect, basically tying off shareability/cacheability "wires" and de-facto making the redistributors and ITSes non-coherent memory observers.

This series aims at starting a discussion over a possible solution to this problem, by adding to the GIC device tree bindings the standard dma-noncoherent property. The GIC driver uses the property to force the redistributors and ITSes shareability attributes to non-shareable, which consequently forces the driver to use CMOs on GIC memory tables.

On ARM DT DMA is default non-coherent, so the GIC driver can't rely on the generic DT dma-coherent/non-coherent property management layer (of_dma_is_coherent()) which would default all GIC designs in the field as non-coherent; it has to rely on ad-hoc dma-noncoherent property handling.

When a consistent approach is agreed upon for DT an equivalent binding will be put forward for ACPI based systems.

Lorenzo Pieralisi (1):
irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
Signature · 1 line
2.34.1
#9 lore raw

PATCH v6 1/1 irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Thomas Gleixner <[email protected]>
Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
drivers/acpi/processor_core.c +15 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
 	return rv;
 }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
 static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
 {
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index 40ebf1726393..af5297ef2293 100644
@@ -5600,6 +5600,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
 		goto node_err;
 	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
 	err = its_probe_one(its);
 	if (!err)
 		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 6fb276504bcc..e4bc5f04226e 100644
@@ -2349,6 +2349,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
 		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
 		return -ENOMEM;
 	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_request_region(redist->base_address, redist->length, "GICR");
 
 	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2373,6 +2378,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
 		return -ENOMEM;
 	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
 	return 0;
 }
include/linux/acpi.h +3 -0
index 28c3fb2bef0d..000d339e1596 100644
@@ -279,6 +279,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
 	return phys_id == PHYS_CPUID_INVALID;
 }
 
+
+int __init acpi_get_madt_revision(void);
+
 /* Validate the processor object's proc_id */
 bool acpi_duplicate_processor_id(int proc_id);
 /* Processor _CTS control */
Signature · 1 line
2.34.1
#10 lore raw

[tip: irq/core] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

The following commit has been merged into the irq/core branch of tip:

The following commit has been merged into the irq/core branch of tip:

Commit-ID:     ababa16fd9bd0e2727a1c31c4fb68d6be053bddc
Gitweb:        https://git.kernel.org/tip/ababa16fd9bd0e2727a1c31c4fb68d6be053bddc
Author:        Lorenzo Pieralisi <[email protected]>
AuthorDate:    Thu, 06 Jun 2024 11:42:38 +02:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Thu, 06 Jun 2024 16:30:15 +02:00

irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

---
 drivers/acpi/processor_core.c    | 15 +++++++++++++++
 drivers/irqchip/irq-gic-v3-its.c |  4 ++++
 drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
 include/linux/acpi.h             |  3 +++
 4 files changed, 31 insertions(+)
drivers/acpi/processor_core.c +15 -0
index b203cfe..915713c 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
 	return rv;
 }
 
+int __init acpi_get_madt_revision(void)
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
 static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
 {
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index 40ebf17..af5297e 100644
@@ -5600,6 +5600,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
 		goto node_err;
 	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
 	err = its_probe_one(its);
 	if (!err)
 		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 6fb2765..e4bc5f0 100644
@@ -2349,6 +2349,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
 		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
 		return -ENOMEM;
 	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_request_region(redist->base_address, redist->length, "GICR");
 
 	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2373,6 +2378,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
 		return -ENOMEM;
 	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
 	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
 	return 0;
 }
include/linux/acpi.h +3 -0
index 28c3fb2..000d339 100644
@@ -279,6 +279,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
 	return phys_id == PHYS_CPUID_INVALID;
 }
 
+
+int __init acpi_get_madt_revision(void);
+
 /* Validate the processor object's proc_id */
 bool acpi_duplicate_processor_id(int proc_id);
 /* Processor _CTS control */
#11 lore raw

PATCH v6 1/1 irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

In reply to Lorenzo Pieralisi
Quoted 50 lines

----------------------------------------------------------------------
The GIC architecture specification defines a set of registers for redistributors and ITSes that control the sharebility and cacheability attributes of redistributors/ITSes initiator ports on the interconnect (GICR_[V]PROPBASER, GICR_[V]PENDBASER, GITS_BASER<n>).

Architecturally the GIC provides a means to drive shareability and cacheability attributes signals but it is not mandatory for designs to wire up the corresponding interconnect signals that control the cacheability/shareability of transactions.

Redistributors and ITSes interconnect ports can be connected to non-coherent interconnects that are not able to manage the shareability/cacheability attributes; this implicitly makes the redistributors and ITSes non-coherent observers.

To enable non-coherent GIC designs on ACPI based systems, parse the MADT GICC/GICR/ITS subtables non-coherent flags to determine whether the respective components are non-coherent observers and force the shareability attributes to be programmed into the redistributors and ITSes registers.

An ACPI global function (acpi_get_madt_revision()) is added to retrieve the MADT revision, in that it is essential to check the MADT revision before checking for flags that were added with MADT revision 7 so that if the kernel is booted with an ACPI MADT table with revision < 7 it skips parsing the newly added flags (that should be zeroed reserved values for MADT versions < 7 but they could turn out to be buggy and should be ignored).

Cc Thomas Gleixner <[email protected]>
Cc Robin Murphy <[email protected]>
Cc Mark Rutland <[email protected]>
Cc "Rafael J. Wysocki" <[email protected]>
Cc Marc Zyngier <[email protected]>
---
  drivers/acpi/processor_core.c    | 15 +++++++++++++++
  drivers/irqchip/irq-gic-v3-its.c |  4 ++++
  drivers/irqchip/irq-gic-v3.c     |  9 +++++++++
  include/linux/acpi.h             |  3 +++
  4 files changed, 31 insertions(+)
drivers/acpi/processor_core.c +0 -0
index b203cfe28550..915713c0e9b7 100644

drivers/acpi/processor_core.c phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id) · line 218

Quoted 4 lines
drivers/acpi/processor_core.c +1 -0
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
  	return rv;
  }
 
+int __init acpi_get_madt_revision(void)

Wondering, if we can have a generic function (acpi_get_tbl_revision) to obtain the revision number for any ACPI table, not just specific to MADT?

Quoted 73 lines
drivers/acpi/processor_core.c +14 -0
+{
+	struct acpi_table_header *madt = NULL;
+	int revision;
+
+	if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_MADT, 0, &madt)))
+		return -EINVAL;
+
+	revision = madt->revision;
+
+	acpi_put_table(madt);
+
+	return revision;
+}
+
  static phys_cpuid_t map_mat_entry(acpi_handle handle, int type, u32 acpi_id)
  {
  	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
drivers/irqchip/irq-gic-v3-its.c +4 -0
index 40ebf1726393..af5297ef2293 100644
@@ -5600,6 +5600,10 @@ static int __init gic_acpi_parse_madt_its(union acpi_subtable_headers *header,
  		goto node_err;
  	}
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (its_entry->flags & ACPI_MADT_ITS_NON_COHERENT))
+		its->flags |= ITS_FLAGS_FORCE_NON_SHAREABLE;
+
  	err = its_probe_one(its);
  	if (!err)
  		return 0;
drivers/irqchip/irq-gic-v3.c +9 -0
index 6fb276504bcc..e4bc5f04226e 100644
@@ -2349,6 +2349,11 @@ gic_acpi_parse_madt_redist(union acpi_subtable_headers *header,
  		pr_err("Couldn't map GICR region @%llx\n", redist->base_address);
  		return -ENOMEM;
  	}
+
+	if (acpi_get_madt_revision() >= 7 &&
+	    (redist->flags & ACPI_MADT_GICR_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
  	gic_request_region(redist->base_address, redist->length, "GICR");
 
  	gic_acpi_register_redist(redist->base_address, redist_base);
@@ -2373,6 +2378,10 @@ gic_acpi_parse_madt_gicc(union acpi_subtable_headers *header,
  		return -ENOMEM;
  	gic_request_region(gicc->gicr_base_address, size, "GICR");
 
+	if (acpi_get_madt_revision() >= 7 &&
+	    (gicc->flags & ACPI_MADT_GICC_NON_COHERENT))
+		gic_data.rdists.flags |= RDIST_FLAGS_FORCE_NON_SHAREABLE;
+
  	gic_acpi_register_redist(gicc->gicr_base_address, redist_base);
  	return 0;
  }
include/linux/acpi.h +3 -0
index 28c3fb2bef0d..000d339e1596 100644
@@ -279,6 +279,9 @@ static inline bool invalid_phys_cpuid(phys_cpuid_t phys_id)
  	return phys_id == PHYS_CPUID_INVALID;
  }
 
+
+int __init acpi_get_madt_revision(void);
+
  /* Validate the processor object's proc_id */
  bool acpi_duplicate_processor_id(int proc_id);
  /* Processor _CTS control */
#12 lore raw

Re: [PATCH v6 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

On Fri, Jun 07, 2024 at 12:21:54AM +0530, Amit Singh Tomar wrote:

[...]

Quoted 11 lines

drivers/acpi/processor_core.c phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id) · line 217

Quoted 8 lines
drivers/acpi/processor_core.c +1 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
  	return rv;
  }
+int __init acpi_get_madt_revision(void)

Wondering, if we can have a generic function (acpi_get_tbl_revision) to obtain the revision number for any ACPI table, not just specific to MADT?

We could - I don't think there would be users other than code in this patch though so I thought it would not be necessary.

Lorenzo

#13 lore raw

Re: [PATCH v6 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

Quoted 16 lines · Amit Singh Tomar

[...]

Quoted 11 lines

drivers/acpi/processor_core.c phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id) · line 217

Quoted 8 lines
drivers/acpi/processor_core.c +1 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 acpi_id)
   	return rv;
   }
+int __init acpi_get_madt_revision(void)

Wondering, if we can have a generic function (acpi_get_tbl_revision) to obtain the revision number for any ACPI table, not just specific to MADT?

Quoted 3 lines

We could - I don't think there would be users other than code in this patch though so I thought it would not be necessary.

Right, it might not be essential now but I see that MPAM will be another user of it once the MPAM patches are out.

https://git.kernel.org/pub/scm/linux/kernel/git/morse/linux.git/tree/drivers/acpi/arm64/mpam.c?h=mpam/snapshot/v6.7-rc2#n299

Thanks
-Amit

#14 lore raw

Re: [PATCH v6 1/1] irqchip/gic-v3: Enable non-coherent redistributors/ITSes ACPI probing

Quoted 17 lines · Amit Singh Tomar
Quoted 17 lines · Amit Singh Tomar

[...]

Quoted 12 lines

drivers/acpi/processor_core.c phys_cpuid_t __init acpi_map_madt_entry(u32 · line 218

Quoted 9 lines
drivers/acpi/processor_core.c +1 -0
index b203cfe28550..915713c0e9b7 100644
@@ -215,6 +215,21 @@ phys_cpuid_t __init acpi_map_madt_entry(u32 
 acpi_id)
        return rv;
    }
+int __init acpi_get_madt_revision(void)

Wondering, if we can have a generic function (acpi_get_tbl_revision) to obtain the revision number for any ACPI table, not just specific to MADT?

Quoted 7 lines
Quoted 3 lines

We could - I don't think there would be users other than code in this patch though so I thought it would not be necessary.

Right, it might not be essential now but I see that MPAM will be another user of it once the MPAM patches are out.

https://git.kernel.org/pub/scm/linux/kernel/git/morse/linux.git/tree/drivers/acpi/arm64/mpam.c?h=mpam/snapshot/v6.7-rc2#n299

Not really; there's already plenty of ACPI code which checks the revision of a table *while* also parsing other information from it, and that MPAM code is doing the same. Using a standalone function to look up the table, check one thing and throw it away, and then immediately have to look it up again to do the rest would be needlessly overcomplicated.

The thing in the GIC case is that doing this semi-redundant lookup to re-retrieve the top-level MADT header while we're already deep into parsing its subtables is still the least-worst option, because the alternative would be invasively churning the whole common MADT abstraction to pass that information all the way down just for this one slightly niche thing.

Thanks,
Robin.