Skip to contentSkip to conversation

irqchip/gic-v3: Assorted fixes and improvements

PATCH Applied Marc Zyngier 4 participants 18 messages Mar 15, 2022 – Apr 5, 2022

Subsystems (3)

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore

Marc Zyngier

author maintainer
#0 lore raw

cover irqchip/gic-v3: Assorted fixes and improvements

This is a small series that happens on the back of [1], which implements additional features in KVM's vgic.

The first patch fixes a pretty embarrassing bug, which is fortunately unlikely to cause any issue in real life. The next two patches are performance optimisations.

Thanks,

        M.

[1] https://lore.kernel.org/r/[email protected]

Marc Zyngier (3):
irqchip/gic-v3: Fix GICR_CTLR.RWP polling irqchip/gic-v3: Detect LPI invalidation MMIO registers irqchip/gic-v3: Relax polling of GIC{R,D}_CTLR.RWP

 drivers/irqchip/irq-gic-v3.c       | 51 +++++++++++++++---------------
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 28 insertions(+), 25 deletions(-)
Signature · 1 line
2.34.1
author maintainer
#1 lore raw

PATCH 1/3 irqchip/gic-v3: Fix GICR_CTLR.RWP polling

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Just fix the damn thing.

Fixes 021f653791ad ("irqchip: gic-v3: Initial support for GICv3")
---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +4 -4
index 5e935d97207d..736163d36b13 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
Signature · 1 line
2.34.1
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

Quoted 6 lines · On Tue, 15 Mar 2022 16:50:32 +0000 Marc Zyngier <[email protected]> wrote:

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Indeed, I wonder why we didn't see issues before. I guess it's either the UWP bit at position GICR_CTLR[31] having a similar implementation, or the MMIO access alone providing enough delay for the writes to finish.

Anyway:

Quoted 5 lines

Just fix the damn thing.

Fixes 021f653791ad ("irqchip: gic-v3: Initial support for GICv3")

Cheers,
Andre

Quoted 38 lines
---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +4 -4
index 5e935d97207d..736163d36b13 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
author maintainer
#3 lore raw

Re: [PATCH 1/3] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

Quoted 8 lines · Andre Przywara
Quoted 4 lines · On Tue, 15 Mar 2022 16:50:32 +0000 Marc Zyngier <[email protected]> wrote:

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

Quoted 7 lines
Quoted 2 lines

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Indeed, I wonder why we didn't see issues before. I guess it's either the UWP bit at position GICR_CTLR[31] having a similar implementation, or the MMIO access alone providing enough delay for the writes to finish.

Because we don't strictly need to wait. Most of the time, the write will have taken place long before we can observe any effect of it. And how often do we disable a SGI or a PPI? Almost never (the PMU is the only one that I can think of).

Quoted 9 lines

Anyway:

Quoted 5 lines

Just fix the damn thing.

Fixes 021f653791ad ("irqchip: gic-v3: Initial support for GICv3")

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Reviewed-by on v1 1/3
#4 lore raw

Re: [PATCH 1/3] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

In reply to Marc Zyngier
Quoted 7 lines · Marc Zyngier

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Quoted 8 lines

Just fix the damn thing.

Fixes 021f653791ad ("irqchip: gic-v3: Initial support for GICv3")
---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
Quoted 37 lines
drivers/irqchip/irq-gic-v3.c +4 -4
index 5e935d97207d..736163d36b13 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
Signature · 1 line
2.34.1
#5 lore raw

[irqchip: irq/irqchip-next] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

In reply to Marc Zyngier
The following commit has been merged into the irq/irqchip-next branch of irqchip:

The following commit has been merged into the irq/irqchip-next branch of irqchip:

Commit-ID:     c114741827436ad1f1d465f3719f77b996ea6eca
Gitweb:        https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms/c114741827436ad1f1d465f3719f77b996ea6eca
Author:        Marc Zyngier <[email protected]>
AuthorDate:    Tue, 15 Mar 2022 16:50:32 
Committer:     Marc Zyngier <[email protected]>
CommitterDate: Mon, 21 Mar 2022 09:17:13 

irqchip/gic-v3: Fix GICR_CTLR.RWP polling

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Just fix the damn thing.

---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +4 -4
index 0efe1a9..9b63165 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
#6 lore raw

[irqchip: irq/irqchip-next] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

In reply to Marc Zyngier
The following commit has been merged into the irq/irqchip-next branch of irqchip:

The following commit has been merged into the irq/irqchip-next branch of irqchip:

Commit-ID:     49cdcea1b0772c29abb9468d82809933c718110e
Gitweb:        https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms/49cdcea1b0772c29abb9468d82809933c718110e
Author:        Marc Zyngier <[email protected]>
AuthorDate:    Tue, 15 Mar 2022 16:50:32 
Committer:     Marc Zyngier <[email protected]>
CommitterDate: Mon, 21 Mar 2022 14:02:44 

irqchip/gic-v3: Fix GICR_CTLR.RWP polling

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Just fix the damn thing.

---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +4 -4
index 0efe1a9..9b63165 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
#7 lore raw

[irqchip: irq/irqchip-fixes] irqchip/gic-v3: Fix GICR_CTLR.RWP polling

In reply to Marc Zyngier
The following commit has been merged into the irq/irqchip-fixes branch of irqchip:

The following commit has been merged into the irq/irqchip-fixes branch of irqchip:

Commit-ID:     0df6664531a12cdd8fc873f0cac0dcb40243d3e9
Gitweb:        https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms/0df6664531a12cdd8fc873f0cac0dcb40243d3e9
Author:        Marc Zyngier <[email protected]>
AuthorDate:    Tue, 15 Mar 2022 16:50:32 
Committer:     Marc Zyngier <[email protected]>
CommitterDate: Tue, 05 Apr 2022 16:33:13 +01:00

irqchip/gic-v3: Fix GICR_CTLR.RWP polling

It turns out that our polling of RWP is totally wrong when checking for it in the redistributors, as we test the *distributor* bit index, whereas it is a different bit number in the RDs... Oopsie boo.

This is embarassing. Not only because it is wrong, but also because it took *8 years* to notice the blunder...

Just fix the damn thing.

---
 drivers/irqchip/irq-gic-v3.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +4 -4
index 0efe1a9..9b63165 100644
@@ -206,11 +206,11 @@ static inline void __iomem *gic_dist_base(struct irq_data *d)
 	}
 }
 
-static void gic_do_wait_for_rwp(void __iomem *base)
+static void gic_do_wait_for_rwp(void __iomem *base, u32 bit)
 {
 	u32 count = 1000000;	/* 1s! */
 
-	while (readl_relaxed(base + GICD_CTLR) & GICD_CTLR_RWP) {
+	while (readl_relaxed(base + GICD_CTLR) & bit) {
 		count--;
 		if (!count) {
 			pr_err_ratelimited("RWP timeout, gone fishing\n");
@@ -224,13 +224,13 @@ static void gic_do_wait_for_rwp(void __iomem *base)
 /* Wait for completion of a distributor change */
 static void gic_dist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data.dist_base);
+	gic_do_wait_for_rwp(gic_data.dist_base, GICD_CTLR_RWP);
 }
 
 /* Wait for completion of a redistributor change */
 static void gic_redist_wait_for_rwp(void)
 {
-	gic_do_wait_for_rwp(gic_data_rdist_rd_base());
+	gic_do_wait_for_rwp(gic_data_rdist_rd_base(), GICR_CTLR_RWP);
 }
 
 #ifdef CONFIG_ARM64
author maintainer
#8 lore raw

PATCH 2/3 irqchip/gic-v3: Detect LPI invalidation MMIO registers

In reply to Marc Zyngier

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +11 -4
index 736163d36b13..363bfe172033 100644
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);
 
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",
+		gic_data.ppi_nr,
+		gic_data.has_rss ? "RSS " : "",
+		gic_data.rdists.has_direct_lpi ? "DirectLPI " : "");
+	
 	if (gic_data.rdists.has_vlpis)
 		pr_info("GICv4 features: %s%s%s\n",
 			gic_data.rdists.has_direct_lpi ? "DirectLPI " : "",
@@ -1797,8 +1806,6 @@ static int __init gic_init_bases(void __iomem *dist_base,
 	irq_domain_update_bus_token(gic_data.domain, DOMAIN_BUS_WIRED);
 
 	gic_data.has_rss = !!(typer & GICD_TYPER_RSS);
-	pr_info("Distributor has %sRange Selector support\n",
-		gic_data.has_rss ? "" : "no ");
 
 	if (typer & GICD_TYPER_MBIS) {
 		err = mbi_init(handle, gic_data.domain);
include/linux/irqchip/arm-gic-v3.h +2 -0
index 12d91f0dedf9..aeb8ced53880 100644
@@ -127,6 +127,8 @@
 #define GICR_PIDR2			GICD_PIDR2
 
 #define GICR_CTLR_ENABLE_LPIS		(1UL << 0)
+#define GICR_CTLR_IR			(1UL << 1)
+#define GICR_CTLR_CES			(1UL << 2)
 #define GICR_CTLR_RWP			(1UL << 3)
 
 #define GICR_TYPER_CPU_NUMBER(r)	(((r) >> 8) & 0xffff)
Signature · 1 line
2.34.1
author maintainer
#9 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

Quoted 6 lines · Marc Zyngier
include/linux/irqchip/arm-gic-v3.h +0 -0
index 12d91f0dedf9..aeb8ced53880 100644

include/linux/irqchip/arm-gic-v3.h line 134

Quoted 8 lines
include/linux/irqchip/arm-gic-v3.h +2 -0
@@ -127,6 +127,8 @@
 #define GICR_PIDR2			GICD_PIDR2
 
 #define GICR_CTLR_ENABLE_LPIS		(1UL << 0)
+#define GICR_CTLR_IR			(1UL << 1)
+#define GICR_CTLR_CES			(1UL << 2)
 #define GICR_CTLR_RWP			(1UL << 3)
 
 #define GICR_TYPER_CPU_NUMBER(r)	(((r) >> 8) & 0xffff)

As Oliver pointed out in [1], this is bollocks, and the two new bits are swapped. I've now fixed that locally.

Thanks,

        M.

[1] https://lore.kernel.org/r/[email protected]

Signature · 1 line
Jazz is not dead. It just smells funny...
#10 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

In reply to Marc Zyngier

On Tue, 15 Mar 2022 16:50:33 +0000
Marc Zyngier <[email protected]> wrote:

Hi,

Quoted 22 lines

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +0 -0
index 736163d36b13..363bfe172033 100644

drivers/irqchip/irq-gic-v3.c static int gic_populate_rdist(void) · line 925

Quoted 9 lines
drivers/irqchip/irq-gic-v3.c +5 -1
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);

Is there any reason you didn't keep this together? I thought this was recommended, in general?

drivers/irqchip/irq-gic-v3.c static int __gic_update_rdist_properties(struct redist_region *region, · line 948

Quoted 8 lines
drivers/irqchip/irq-gic-v3.c +1 -0
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |

So this means that has_direct_lpi is not really correct anymore, as the IR bit only covers the INVL and SYNCR registers, not the GICR_SETLPIR and GICR_CLRLPIR registers, if I understand the spec correctly?

But I guess this is nitpicking, as we don't use direct LPIs at all in Linux? And I guess the target is lpi_update_config(), which now doesn't need the command queue anymore?

Maybe this could be clarified in the commit message?

drivers/irqchip/irq-gic-v3.c static void gic_update_rdist_properties(void) · line 970

Quoted 9 lines
drivers/irqchip/irq-gic-v3.c +1 -1
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",

I like having that on one line, but it looks a bit odd with the trailing comma when we have neither RSS nor DirectLPI. What about:

	pr_info("GICv3 features: %d PPIs%s%s\n",
	gic_data.ppi_nr,
	gic_data.has_rss ? ", RSS" : "",
	gic_data.rdists.has_direct_lpi ? ", DirectLPI" : "");
Quoted 21 lines
drivers/irqchip/irq-gic-v3.c +4 -2
+		gic_data.ppi_nr,
+		gic_data.has_rss ? "RSS " : "",
+		gic_data.rdists.has_direct_lpi ? "DirectLPI " : "");
+	
 	if (gic_data.rdists.has_vlpis)
 		pr_info("GICv4 features: %s%s%s\n",
 			gic_data.rdists.has_direct_lpi ? "DirectLPI " : "",
@@ -1797,8 +1806,6 @@ static int __init gic_init_bases(void __iomem *dist_base,
 	irq_domain_update_bus_token(gic_data.domain, DOMAIN_BUS_WIRED);
 
 	gic_data.has_rss = !!(typer & GICD_TYPER_RSS);
-	pr_info("Distributor has %sRange Selector support\n",
-		gic_data.has_rss ? "" : "no ");
 
 	if (typer & GICD_TYPER_MBIS) {
 		err = mbi_init(handle, gic_data.domain);
include/linux/irqchip/arm-gic-v3.h +0 -0
index 12d91f0dedf9..aeb8ced53880 100644

include/linux/irqchip/arm-gic-v3.h line 131

Quoted 5 lines
include/linux/irqchip/arm-gic-v3.h +2 -0
@@ -127,6 +127,8 @@
 #define GICR_PIDR2			GICD_PIDR2
 
 #define GICR_CTLR_ENABLE_LPIS		(1UL << 0)
+#define GICR_CTLR_IR			(1UL << 1)
+#define GICR_CTLR_CES			(1UL << 2)

As Oliver (and you) already pointed out, this is swapped.

Cheers,
Andre

Quoted 3 lines
include/linux/irqchip/arm-gic-v3.h +0 -0
 #define GICR_CTLR_RWP			(1UL << 3)
 
 #define GICR_TYPER_CPU_NUMBER(r)	(((r) >> 8) & 0xffff)
author maintainer
#11 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

Quoted 32 lines · Andre Przywara

On Tue, 15 Mar 2022 16:50:33 +0000
Marc Zyngier <[email protected]> wrote:

Hi,

Quoted 26 lines

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +0 -1
index 736163d36b13..363bfe172033 100644
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
Quoted 8 lines

drivers/irqchip/irq-gic-v3.c static int gic_populate_rdist(void) · line 925

Quoted 5 lines
drivers/irqchip/irq-gic-v3.c +5 -0
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);

Is there any reason you didn't keep this together? I thought this was recommended, in general?

Sorry, keep what together with what?

Quoted 14 lines

drivers/irqchip/irq-gic-v3.c static int __gic_update_rdist_properties(struct redist_region *region, · line 948

Quoted 9 lines
drivers/irqchip/irq-gic-v3.c +1 -0
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |

So this means that has_direct_lpi is not really correct anymore, as the IR bit only covers the INVL and SYNCR registers, not the GICR_SETLPIR and GICR_CLRLPIR registers, if I understand the spec correctly?

Quoted 3 lines

But I guess this is nitpicking, as we don't use direct LPIs at all in Linux? And I guess the target is lpi_update_config(), which now doesn't need the command queue anymore?

Exactly. The history of this crap is convoluted:

The canonical goal of DirectLPI was to support LPIs without an ITS. Thankfully, this was never implemented. What was implemented by our HiSi friends was DirectLPI *with* an ITS, which was illegal at the time, but also the only way to make GICv4.0 work at a reasonable speed. That's where the direct_lpi boolean comes from.

RVPEI added some more confusion by offering a subset of DirectLPI for invalidation of vlpis. And then IR was introduced because there is really no reason not to offer the same service on GICv3.

Quoted 2 lines

Maybe this could be clarified in the commit message?

Sure, can do.

Quoted 14 lines

drivers/irqchip/irq-gic-v3.c static void gic_update_rdist_properties(void) · line 970

Quoted 10 lines
drivers/irqchip/irq-gic-v3.c +1 -1
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",

I like having that on one line, but it looks a bit odd with the trailing comma when we have neither RSS nor DirectLPI. What about:

Quoted 4 lines
	pr_info("GICv3 features: %d PPIs%s%s\n",
	gic_data.ppi_nr,
	gic_data.has_rss ? ", RSS" : "",
	gic_data.rdists.has_direct_lpi ? ", DirectLPI" : "");

Yeah, looks better.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
#12 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

On Wed, 16 Mar 2022 15:36:54 +0000
Marc Zyngier <[email protected]> wrote:

Hi Marc,

Quoted 43 lines
Quoted 32 lines · Andre Przywara

On Tue, 15 Mar 2022 16:50:33 +0000
Marc Zyngier <[email protected]> wrote:

Hi,

Quoted 26 lines

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +0 -1
index 736163d36b13..363bfe172033 100644
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
Quoted 8 lines

drivers/irqchip/irq-gic-v3.c static int gic_populate_rdist(void) · line 925

Quoted 5 lines
drivers/irqchip/irq-gic-v3.c +5 -0
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);  

Is there any reason you didn't keep this together? I thought this was recommended, in general?

Quoted 1 line

Sorry, keep what together with what?

Sorry, I meant the variable declaration with the initialisation:

	u64 typer = gic_read_typer(ptr + GICR_TYPER);
	u32 ctlr = readl_relaxed(ptr + GICR_CTLR);

I see this a lot (especially in KVM code), so was just wondering if this is not cool anymore.

Quoted 25 lines
Quoted 13 lines

drivers/irqchip/irq-gic-v3.c static int __gic_update_rdist_properties(struct redist_region *region, · line 948

Quoted 8 lines
drivers/irqchip/irq-gic-v3.c +1 -0
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |  

So this means that has_direct_lpi is not really correct anymore, as the IR bit only covers the INVL and SYNCR registers, not the GICR_SETLPIR and GICR_CLRLPIR registers, if I understand the spec correctly?

Quoted 3 lines

But I guess this is nitpicking, as we don't use direct LPIs at all in Linux? And I guess the target is lpi_update_config(), which now doesn't need the command queue anymore?

Exactly. The history of this crap is convoluted:

The canonical goal of DirectLPI was to support LPIs without an ITS. Thankfully, this was never implemented. What was implemented by our HiSi friends was DirectLPI *with* an ITS, which was illegal at the time, but also the only way to make GICv4.0 work at a reasonable speed. That's where the direct_lpi boolean comes from.

Quoted 3 lines

RVPEI added some more confusion by offering a subset of DirectLPI for invalidation of vlpis. And then IR was introduced because there is really no reason not to offer the same service on GICv3.

Ah, I was hoping for this kind of answer ;-) , so many thanks!

Cheers,
Andre

Quoted 31 lines
Quoted 3 lines

Maybe this could be clarified in the commit message?

Sure, can do.

Quoted 14 lines

drivers/irqchip/irq-gic-v3.c static void gic_update_rdist_properties(void) · line 970

Quoted 10 lines
drivers/irqchip/irq-gic-v3.c +1 -1
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",  

I like having that on one line, but it looks a bit odd with the trailing comma when we have neither RSS nor DirectLPI. What about:

Quoted 4 lines
	pr_info("GICv3 features: %d PPIs%s%s\n",
	gic_data.ppi_nr,
	gic_data.has_rss ? ", RSS" : "",
	gic_data.rdists.has_direct_lpi ? ", DirectLPI" : "");  

Yeah, looks better.

Thanks,

	M.
#13 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

In reply to Marc Zyngier
Quoted 43 lines · Marc Zyngier

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +6 -1
index 736163d36b13..363bfe172033 100644
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);
 
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 

drivers/irqchip/irq-gic-v3.c static void gic_update_rdist_properties(void) · line 973

Quoted 8 lines
drivers/irqchip/irq-gic-v3.c +4 -1
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",
+		gic_data.ppi_nr,
+		gic_data.has_rss ? "RSS " : "",
+		gic_data.rdists.has_direct_lpi ? "DirectLPI " : "");

I understand GICR_CTLR.IR detection (which is v4.1 feature) - I don't get why in this patch we are adding a GICv3 DirectLPI info dump (hunk above), it is probably nitpicking but the hunk above does not seem to belong in this patch - it is a separate print info refactoring or I am reading it wrongly.

Lorenzo

Quoted 29 lines
drivers/irqchip/irq-gic-v3.c +1 -2
+	
 	if (gic_data.rdists.has_vlpis)
 		pr_info("GICv4 features: %s%s%s\n",
 			gic_data.rdists.has_direct_lpi ? "DirectLPI " : "",
@@ -1797,8 +1806,6 @@ static int __init gic_init_bases(void __iomem *dist_base,
 	irq_domain_update_bus_token(gic_data.domain, DOMAIN_BUS_WIRED);
 
 	gic_data.has_rss = !!(typer & GICD_TYPER_RSS);
-	pr_info("Distributor has %sRange Selector support\n",
-		gic_data.has_rss ? "" : "no ");
 
 	if (typer & GICD_TYPER_MBIS) {
 		err = mbi_init(handle, gic_data.domain);
include/linux/irqchip/arm-gic-v3.h +2 -0
index 12d91f0dedf9..aeb8ced53880 100644
@@ -127,6 +127,8 @@
 #define GICR_PIDR2			GICD_PIDR2
 
 #define GICR_CTLR_ENABLE_LPIS		(1UL << 0)
+#define GICR_CTLR_IR			(1UL << 1)
+#define GICR_CTLR_CES			(1UL << 2)
 #define GICR_CTLR_RWP			(1UL << 3)
 
 #define GICR_TYPER_CPU_NUMBER(r)	(((r) >> 8) & 0xffff)
Signature · 1 line
2.34.1
author maintainer
#14 lore raw

Re: [PATCH 2/3] irqchip/gic-v3: Detect LPI invalidation MMIO registers

Quoted 44 lines · Lorenzo Pieralisi
Quoted 42 lines · Marc Zyngier

Since GICv4.1, an implementation can offer the same MMIO-based implementation as DirectLPI, only with an ITS. Given that this can be hugely beneficial for workloads that are very LPI masking heavy (although these workloads are admitedly a bit odd).

Interestingly, this is independent of RVPEI, which only *implies* the functionnality.

So let's detect whether the implementation has GICR_CTLR.IR set, and propagate this as DirectLPI to the ITS driver.

---
 drivers/irqchip/irq-gic-v3.c       | 15 +++++++++++----
 include/linux/irqchip/arm-gic-v3.h |  2 ++
 2 files changed, 13 insertions(+), 4 deletions(-)
drivers/irqchip/irq-gic-v3.c +6 -1
index 736163d36b13..363bfe172033 100644
@@ -918,7 +918,11 @@ static int gic_populate_rdist(void)
 static int __gic_update_rdist_properties(struct redist_region *region,
 					 void __iomem *ptr)
 {
-	u64 typer = gic_read_typer(ptr + GICR_TYPER);
+	u64 typer;
+	u32 ctlr;
+
+	typer = gic_read_typer(ptr + GICR_TYPER);
+	ctlr = readl_relaxed(ptr + GICR_CTLR);
 
 	/* Boot-time cleanip */
 	if ((typer & GICR_TYPER_VLPIS) && (typer & GICR_TYPER_RVPEID)) {
@@ -941,6 +945,7 @@ static int __gic_update_rdist_properties(struct redist_region *region,
 	/* RVPEID implies some form of DirectLPI, no matter what the doc says... :-/ */
 	gic_data.rdists.has_rvpeid &= !!(typer & GICR_TYPER_RVPEID);
 	gic_data.rdists.has_direct_lpi &= (!!(typer & GICR_TYPER_DirectLPIS) |
+					   !!(ctlr & GICR_CTLR_IR) |
 					   gic_data.rdists.has_rvpeid);
 	gic_data.rdists.has_vpend_valid_dirty &= !!(typer & GICR_TYPER_DIRTY);
 
Quoted 11 lines

drivers/irqchip/irq-gic-v3.c static void gic_update_rdist_properties(void) · line 973

Quoted 9 lines
drivers/irqchip/irq-gic-v3.c +4 -1
@@ -962,7 +967,11 @@ static void gic_update_rdist_properties(void)
 	gic_iterate_rdists(__gic_update_rdist_properties);
 	if (WARN_ON(gic_data.ppi_nr == UINT_MAX))
 		gic_data.ppi_nr = 0;
-	pr_info("%d PPIs implemented\n", gic_data.ppi_nr);
+	pr_info("GICv3 features: %d PPIs, %s%s\n",
+		gic_data.ppi_nr,
+		gic_data.has_rss ? "RSS " : "",
+		gic_data.rdists.has_direct_lpi ? "DirectLPI " : "");

I understand GICR_CTLR.IR detection (which is v4.1 feature) - I don't

No, it is *also* a GICv3 feature. RVPEI implies IR, but IR is a feature on its own (see my reply to Andre on the same subject). Nothing restrict IR to a GICv4.1+ implementation, and KVM is about to expose these registers to the GICv*3* guest.

Quoted 4 lines

get why in this patch we are adding a GICv3 DirectLPI info dump (hunk above), it is probably nitpicking but the hunk above does not seem to belong in this patch - it is a separate print info refactoring or I am reading it wrongly.

It is indeed just refactoring the kernel messages so that we can see that we enable DirectLPI for GICv3 as well. I honestly don't think this deserves a separate patch.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
author maintainer
#15 lore raw

PATCH 3/3 irqchip/gic-v3: Relax polling of GIC{R,D}_CTLR.RWP

In reply to Marc Zyngier

Recent work on the KVM GIC emulation has revealed that the GICv3 driver is a bit RWP-happy, as it polls this bit for each and every write MMIO access involving a single interrupt.

As it turns out, polling RWP is only required when:
- Disabling an SGI, PPI or SPI
- Disabling LPIs at the redistributor level
- Disabling groups
- Enabling ARE
- Dealing with DPG*

Simplify the driver by removing all the other instances of RWP polling, and add the one that was missing when enabling the distributor (as that's where we set ARE).

---
 drivers/irqchip/irq-gic-v3.c | 28 +++++++++++-----------------
 1 file changed, 11 insertions(+), 17 deletions(-)
drivers/irqchip/irq-gic-v3.c +11 -17
index 363bfe172033..05ff7fef64cb 100644
@@ -352,28 +352,27 @@ static int gic_peek_irq(struct irq_data *d, u32 offset)
 
 static void gic_poke_irq(struct irq_data *d, u32 offset)
 {
-	void (*rwp_wait)(void);
 	void __iomem *base;
 	u32 index, mask;
 
 	offset = convert_offset_index(d, offset, &index);
 	mask = 1 << (index % 32);
 
-	if (gic_irq_in_rdist(d)) {
+	if (gic_irq_in_rdist(d))
 		base = gic_data_rdist_sgi_base();
-		rwp_wait = gic_redist_wait_for_rwp;
-	} else {
+	else
 		base = gic_data.dist_base;
-		rwp_wait = gic_dist_wait_for_rwp;
-	}
 
 	writel_relaxed(mask, base + offset + (index / 32) * 4);
-	rwp_wait();
 }
 
 static void gic_mask_irq(struct irq_data *d)
 {
 	gic_poke_irq(d, GICD_ICENABLER);
+	if (gic_irq_in_rdist(d))
+		gic_redist_wait_for_rwp();
+	else
+		gic_dist_wait_for_rwp();
 }
 
 static void gic_eoimode1_mask_irq(struct irq_data *d)
@@ -574,7 +573,6 @@ static int gic_set_type(struct irq_data *d, unsigned int type)
 {
 	enum gic_intid_range range;
 	unsigned int irq = gic_irq(d);
-	void (*rwp_wait)(void);
 	void __iomem *base;
 	u32 offset, index;
 	int ret;
@@ -590,17 +588,14 @@ static int gic_set_type(struct irq_data *d, unsigned int type)
 	    type != IRQ_TYPE_LEVEL_HIGH && type != IRQ_TYPE_EDGE_RISING)
 		return -EINVAL;
 
-	if (gic_irq_in_rdist(d)) {
+	if (gic_irq_in_rdist(d))
 		base = gic_data_rdist_sgi_base();
-		rwp_wait = gic_redist_wait_for_rwp;
-	} else {
+	else
 		base = gic_data.dist_base;
-		rwp_wait = gic_dist_wait_for_rwp;
-	}
 
 	offset = convert_offset_index(d, GICD_ICFGR, &index);
 
-	ret = gic_configure_irq(index, type, base + offset, rwp_wait);
+	ret = gic_configure_irq(index, type, base + offset, NULL);
 	if (ret && (range == PPI_RANGE || range == EPPI_RANGE)) {
 		/* Misconfigured PPIs are usually not fatal */
 		pr_warn("GIC: PPI INTID%d is secure or misconfigured\n", irq);
@@ -808,7 +803,7 @@ static void __init gic_dist_init(void)
 		writel_relaxed(GICD_INT_DEF_PRI_X4, base + GICD_IPRIORITYRnE + i);
 
 	/* Now do the common stuff, and wait for the distributor to drain */
-	gic_dist_config(base, GIC_LINE_NR, gic_dist_wait_for_rwp);
+	gic_dist_config(base, GIC_LINE_NR, NULL);
 
 	val = GICD_CTLR_ARE_NS | GICD_CTLR_ENABLE_G1A | GICD_CTLR_ENABLE_G1;
 	if (gic_data.rdists.gicd_typer2 & GICD_TYPER2_nASSGIcap) {
@@ -818,6 +813,7 @@ static void __init gic_dist_init(void)
 
 	/* Enable distributor with ARE, Group1 */
 	writel_relaxed(val, base + GICD_CTLR);
+	gic_dist_wait_for_rwp();
 
 	/*
 	 * Set all global interrupts to the boot CPU only. ARE must be
@@ -1293,8 +1289,6 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 	 */
 	if (enabled)
 		gic_unmask_irq(d);
-	else
-		gic_dist_wait_for_rwp();
 
 	irq_data_update_effective_affinity(d, cpumask_of(cpu));
 
Signature · 1 line
2.34.1
#16 lore raw

Re: [PATCH 3/3] irqchip/gic-v3: Relax polling of GIC{R,D}_CTLR.RWP

On Tue, 15 Mar 2022 16:50:34 +0000
Marc Zyngier <[email protected]> wrote:

Hi,

Quoted 11 lines

Recent work on the KVM GIC emulation has revealed that the GICv3 driver is a bit RWP-happy, as it polls this bit for each and every write MMIO access involving a single interrupt.

As it turns out, polling RWP is only required when:
- Disabling an SGI, PPI or SPI
- Disabling LPIs at the redistributor level
- Disabling groups
- Enabling ARE
- Dealing with DPG*

Quoted 3 lines

Simplify the driver by removing all the other instances of RWP polling, and add the one that was missing when enabling the distributor (as that's where we set ARE).

Don't we need an explicit call to wait_for_rwp() now for:
gic_irq_set_irqchip_state(IRQCHIP_STATE_MASKED, true) ?

The rest looks fine to me.

Thanks,
Andre

Quoted 92 lines
---
 drivers/irqchip/irq-gic-v3.c | 28 +++++++++++-----------------
 1 file changed, 11 insertions(+), 17 deletions(-)
drivers/irqchip/irq-gic-v3.c +11 -15
index 363bfe172033..05ff7fef64cb 100644
@@ -352,28 +352,27 @@ static int gic_peek_irq(struct irq_data *d, u32 offset)
 
 static void gic_poke_irq(struct irq_data *d, u32 offset)
 {
-     void (*rwp_wait)(void);
      void __iomem *base;
      u32 index, mask;
 
      offset = convert_offset_index(d, offset, &index);
      mask = 1 << (index % 32);
 
-     if (gic_irq_in_rdist(d)) {
+     if (gic_irq_in_rdist(d))
              base = gic_data_rdist_sgi_base();
-             rwp_wait = gic_redist_wait_for_rwp;
-     } else {
+     else
              base = gic_data.dist_base;
-             rwp_wait = gic_dist_wait_for_rwp;
-     }
 
      writel_relaxed(mask, base + offset + (index / 32) * 4);
-     rwp_wait();
 }
 
 static void gic_mask_irq(struct irq_data *d)
 {
      gic_poke_irq(d, GICD_ICENABLER);
+     if (gic_irq_in_rdist(d))
+             gic_redist_wait_for_rwp();
+     else
+             gic_dist_wait_for_rwp();
 }
 
 static void gic_eoimode1_mask_irq(struct irq_data *d)
@@ -574,7 +573,6 @@ static int gic_set_type(struct irq_data *d, unsigned int type)
 {
      enum gic_intid_range range;
      unsigned int irq = gic_irq(d);
-     void (*rwp_wait)(void);
      void __iomem *base;
      u32 offset, index;
      int ret;
@@ -590,17 +588,14 @@ static int gic_set_type(struct irq_data *d, unsigned int type)
          type != IRQ_TYPE_LEVEL_HIGH && type != IRQ_TYPE_EDGE_RISING)
              return -EINVAL;
 
-     if (gic_irq_in_rdist(d)) {
+     if (gic_irq_in_rdist(d))
              base = gic_data_rdist_sgi_base();
-             rwp_wait = gic_redist_wait_for_rwp;
-     } else {
+     else
              base = gic_data.dist_base;
-             rwp_wait = gic_dist_wait_for_rwp;
-     }
 
      offset = convert_offset_index(d, GICD_ICFGR, &index);
 
-     ret = gic_configure_irq(index, type, base + offset, rwp_wait);
+     ret = gic_configure_irq(index, type, base + offset, NULL);
      if (ret && (range == PPI_RANGE || range == EPPI_RANGE)) {
              /* Misconfigured PPIs are usually not fatal */
              pr_warn("GIC: PPI INTID%d is secure or misconfigured\n", irq);
@@ -808,7 +803,7 @@ static void __init gic_dist_init(void)
              writel_relaxed(GICD_INT_DEF_PRI_X4, base + GICD_IPRIORITYRnE + i);
 
      /* Now do the common stuff, and wait for the distributor to drain */
-     gic_dist_config(base, GIC_LINE_NR, gic_dist_wait_for_rwp);
+     gic_dist_config(base, GIC_LINE_NR, NULL);
 
      val = GICD_CTLR_ARE_NS | GICD_CTLR_ENABLE_G1A | GICD_CTLR_ENABLE_G1;
      if (gic_data.rdists.gicd_typer2 & GICD_TYPER2_nASSGIcap) {
@@ -818,6 +813,7 @@ static void __init gic_dist_init(void)
 
      /* Enable distributor with ARE, Group1 */
      writel_relaxed(val, base + GICD_CTLR);
+     gic_dist_wait_for_rwp();
 
      /*
       * Set all global interrupts to the boot CPU only. ARE must be

drivers/irqchip/irq-gic-v3.c static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val, · line 1294

Quoted 8 lines
drivers/irqchip/irq-gic-v3.c +0 -2
@@ -1293,8 +1289,6 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
       */
      if (enabled)
              gic_unmask_irq(d);
-     else
-             gic_dist_wait_for_rwp();
 
      irq_data_update_effective_affinity(d, cpumask_of(cpu));
 

IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

author maintainer
#17 lore raw

Re: [PATCH 3/3] irqchip/gic-v3: Relax polling of GIC{R,D}_CTLR.RWP

Quoted 17 lines · Andre Przywara

On Tue, 15 Mar 2022 16:50:34 +0000
Marc Zyngier <[email protected]> wrote:

Hi,

Quoted 11 lines

Recent work on the KVM GIC emulation has revealed that the GICv3 driver is a bit RWP-happy, as it polls this bit for each and every write MMIO access involving a single interrupt.

As it turns out, polling RWP is only required when:
- Disabling an SGI, PPI or SPI
- Disabling LPIs at the redistributor level
- Disabling groups
- Enabling ARE
- Dealing with DPG*

Quoted 6 lines
Quoted 3 lines

Simplify the driver by removing all the other instances of RWP polling, and add the one that was missing when enabling the distributor (as that's where we set ARE).

Don't we need an explicit call to wait_for_rwp() now for:
gic_irq_set_irqchip_state(IRQCHIP_STATE_MASKED, true) ?

Ah, yes, I missed that one. Thanks.

Quoted 1 line

IMPORTANT NOTICE: The contents of this email [...]

The ARM IT crap is firing again. Wrong SMTP server?

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.