Skip to contentSkip to conversation

arm64/sysreg: Correct the values for GICv4.1

PATCH Applied Raghavendra Rao Ananta 5 participants 7 messages Jul 18, 2024 – Jul 23, 2024

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm64/sysreg: Correct the values for GICv4.1

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +2 -2
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0	63:32
 UnsignedEnum	31:28	GIC
 	0b0000	NI
 	0b0001	GICv3
-	0b0010	GICv4p1
+	0b0011	GICv4p1
 EndEnum
 UnsignedEnum	27:24	Virt_frac
 	0b0000	NI
@@ -903,7 +903,7 @@ EndEnum
 UnsignedEnum	27:24	GIC
 	0b0000	NI
 	0b0001	IMP
-	0b0010	V4P1
+	0b0011	V4P1
 EndEnum
 SignedEnum	23:20	AdvSIMD
 	0b0000	IMP

base-commit: 0c3836482481200ead7b416ca80c68a29cfdaabd

Signature · 1 line
2.45.2.1089.g2a221341d9-goog
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 24 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +1 -1
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0	63:32
 UnsignedEnum	31:28	GIC
 	0b0000	NI
 	0b0001	GICv3
-	0b0010	GICv4p1
+	0b0011	GICv4p1
 EndEnum
 UnsignedEnum	27:24	Virt_frac
 	0b0000	NI

arch/arm64/tools/sysreg EndEnum · line 906

Quoted 5 lines
arch/arm64/tools/sysreg +1 -1
@@ -903,7 +903,7 @@ EndEnum
 UnsignedEnum	27:24	GIC
 	0b0000	NI
 	0b0001	IMP
-	0b0010	V4P1
+	0b0011	V4P1

I wonder why we have different naming schemes for the same feature...

arch/arm64/tools/sysreg EndEnum · line 909

Quoted 4 lines
arch/arm64/tools/sysreg +0 -0
 EndEnum
 SignedEnum	23:20	AdvSIMD
 	0b0000	IMP

Yup, this looks correct and checks out against revision H.b of the GICv3 spec, revision K.a of the ARM ARM, and even I.a (which the original patches were referencing).

Once more, it shows that these dumps should be automatically generated from the XML instead of (creatively) hand-written.

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
reviewer Reviewed-by on v1 1/1
#3 lore raw
In reply to Raghavendra Rao Ananta
Quoted 23 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +1 -1
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0	63:32
 UnsignedEnum	31:28	GIC
 	0b0000	NI
 	0b0001	GICv3
-	0b0010	GICv4p1
+	0b0011	GICv4p1
 EndEnum
 UnsignedEnum	27:24	Virt_frac
 	0b0000	NI

arch/arm64/tools/sysreg EndEnum · line 909

Quoted 8 lines
arch/arm64/tools/sysreg +1 -1
@@ -903,7 +903,7 @@ EndEnum
 UnsignedEnum	27:24	GIC
 	0b0000	NI
 	0b0001	IMP
-	0b0010	V4P1
+	0b0011	V4P1
 EndEnum
 SignedEnum	23:20	AdvSIMD
 	0b0000	IMP

Fortunately there is no user for this bit inside kernel. We had checked against the correct hard-coded value (0x3) in gic_cpuif_has_vsgi().

#2 lore raw
In reply to Marc Zyngier
Quoted 25 lines · Marc Zyngier
Quoted 23 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +1 -1
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0	63:32
 UnsignedEnum	31:28	GIC
 	0b0000	NI
 	0b0001	GICv3
-	0b0010	GICv4p1
+	0b0011	GICv4p1
 EndEnum
 UnsignedEnum	27:24	Virt_frac
 	0b0000	NI
Quoted 8 lines

arch/arm64/tools/sysreg EndEnum · line 906

Quoted 6 lines
arch/arm64/tools/sysreg +1 -1
@@ -903,7 +903,7 @@ EndEnum
 UnsignedEnum	27:24	GIC
 	0b0000	NI
 	0b0001	IMP
-	0b0010	V4P1
+	0b0011	V4P1

I wonder why we have different naming schemes for the same feature...

Both definitions were added via different commits and different developers who might just have interpreted the following common description bit differently.

"System register interface to version 4.1 of the GIC CPU interface is supported"

1224308075f1 ("arm64/sysreg: Convert ID_PFR1_EL1 to automatic generation") cea08f2bf406 ("arm64/sysreg: Convert ID_AA64PFR0_EL1 to automatic generation")

But I agree that same fields should be named exactly the same both in their 32 bit and 64 bit variants.

Quoted 17 lines

arch/arm64/tools/sysreg EndEnum · line 909

Quoted 5 lines
arch/arm64/tools/sysreg +0 -0
 EndEnum
 SignedEnum	23:20	AdvSIMD
 	0b0000	IMP

Yup, this looks correct and checks out against revision H.b of the GICv3 spec, revision K.a of the ARM ARM, and even I.a (which the original patches were referencing).

Once more, it shows that these dumps should be automatically generated from the XML instead of (creatively) hand-written.

	M.
#4 lore raw
In reply to Zenghui Yu
Quoted 27 lines · Zenghui Yu
Quoted 26 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--

 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +0 -0
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0    63:32

 UnsignedEnum    31:28    GIC
     0b0000    NI
     0b0001    GICv3
-    0b0010    GICv4p1
+    0b0011    GICv4p1
 EndEnum
 UnsignedEnum    27:24    Virt_frac
     0b0000    NI

arch/arm64/tools/sysreg +0 -0
@@ -903,7 +903,7 @@ EndEnum

 UnsignedEnum    27:24    GIC
     0b0000    NI
     0b0001    IMP

Quoted 8 lines
Quoted 5 lines

-    0b0010    V4P1
+    0b0011    V4P1
 EndEnum
 SignedEnum    23:20    AdvSIMD
     0b0000    IMP

Fortunately there is no user for this bit inside kernel. We had checked against the correct hard-coded value (0x3) in gic_cpuif_has_vsgi().

which probably helped this problem to be left unnoticed till now :) but I guess it would be better to use ID_AA64PFR0_EL1_GIC_V4P1 there instead.

Quoted 3 lines
Reviewed-by on v1 1/1
#5 lore raw
In reply to Raghavendra Rao Ananta
Quoted 32 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

---
 arch/arm64/tools/sysreg | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/tools/sysreg +2 -2
index a4c1dd4741a47..7ceaa1e0b4bc2 100644
@@ -149,7 +149,7 @@ Res0	63:32
 UnsignedEnum	31:28	GIC
 	0b0000	NI
 	0b0001	GICv3
-	0b0010	GICv4p1
+	0b0011	GICv4p1
 EndEnum
 UnsignedEnum	27:24	Virt_frac
 	0b0000	NI
@@ -903,7 +903,7 @@ EndEnum
 UnsignedEnum	27:24	GIC
 	0b0000	NI
 	0b0001	IMP
-	0b0010	V4P1
+	0b0011	V4P1
 EndEnum
 SignedEnum	23:20	AdvSIMD
 	0b0000	IMP
Quoted 1 line

base-commit: 0c3836482481200ead7b416ca80c68a29cfdaabd

This checks against ARM DDI 0487K.a and ddi0601/2024-06 XML.

maintainer Applied
#6 lore raw
In reply to Raghavendra Rao Ananta
Quoted 5 lines · Raghavendra Rao Ananta

Currently, sysreg has value as 0b0010 for the presence of GICv4.1 in ID_PFR1_EL1 and ID_AA64PFR0_EL1, instead of 0b0011 as per ARM ARM. Hence, correct them to reflect ARM ARM.

Applied to arm64 (for-next/core), thanks!

[1/1] arm64/sysreg: Correct the values for GICv4.1

      https://git.kernel.org/arm64/c/f3dfcd25455b

Cheers,

Signature · 5 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev