[PATCH v2 00/16] clocksource/arm_arch_timer: Add basic ARMv8.6 support

[PATCH v2 00/16] clocksource/arm_arch_timer: Add basic ARMv8.6 support

[Marc Zyngier]

This is v2 of the series enabling ARMv8.6 support for timer subsystem,
and was prompted by a discussion with Oliver around the fact that an
ARMv8.6 implementation must have a 1GHz counter, which leads to a
number of things to break in the timer code:

- the counter rollover can come pretty quickly as we only advertise a
  56bit counter,
- the maximum timer delta can be remarkably small, as we use the
  countdown interface which is limited to 32bit...

Thankfully, there is a way out: we can compute the minimal width of
the counter based on the guarantees that the architecture gives us,
and we can use the 64bit comparator interface instead of the countdown
to program the timer.

Finally, we start making use of the ARMv8.6 ECV features by switching
accesses to the counters to a self-synchronising register, removing
the need for an ISB. Hopefully, implementations will *not* just stick
an invisible ISB there...

A side effect of the switch to CVAL is that XGene-1 breaks. I have
added a workaround to keep it alive.

I have added Oliver's original patch[0] to the series and tweaked a
couple of things. Blame me if I broke anything.

The whole things has been tested on Juno (sysreg + MMIO timers),
XGene-1 (broken sysreg timers), FVP (FEAT_ECV, CNT*CTSS_EL0).

* From v1:
  - New patch adding a bunch of BUILD_BUG()s for register accesses we
    don't expect. This makes subsequent patches much simpler.
  - New patch moving the ISBs for workaround in a way that makes
    more sense for the self-synchronising accessors.
  - Rework the XGene-1 workaround to rely solely on MIDR.
  - Split the CNTVCTSS trap handling in its own patch.
  - Rebased on 5.15-rc2
  - Collected RBs, with thanks.

[0] https://lore.kernel.org/r/20210807191428.3488948-1-oupton@google.com
[1] https://lore.kernel.org/r/20210809152651.2297337-2-maz@kernel.org

Marc Zyngier (15):
  clocksource/arm_arch_timer: Add build-time guards for unhandled
    register accesses
  clocksource/arm_arch_timer: Drop CNT*_TVAL read accessors
  clocksource/arm_arch_timer: Extend write side of timer register
    accessors to u64
  clocksource/arm_arch_timer: Move system register timer programming
    over to CVAL
  clocksource/arm_arch_timer: Move drop _tval from erratum function
    names
  clocksource/arm_arch_timer: Fix MMIO base address vs callback ordering
    issue
  clocksource/arm_arch_timer: Move MMIO timer programming over to CVAL
  clocksource/arm_arch_timer: Advertise 56bit timer to the core code
  clocksource/arm_arch_timer: Work around broken CVAL implementations
  clocksource/arm_arch_timer: Remove any trace of the TVAL programming
    interface
  clocksource/arm_arch_timer: Drop unnecessary ISB on CVAL programming
  clocksource/arch_arm_timer: Move workaround synchronisation around
  arm64: Add a capability for FEAT_ECV
  arm64: Add CNT{P,V}CTSS_EL0 alternatives to cnt{p,v}ct_el0
  arm64: Add handling of CNTVCTSS traps

Oliver Upton (1):
  clocksource/arm_arch_timer: Fix masking for high freq counters

 arch/arm/include/asm/arch_timer.h    |  37 ++--
 arch/arm64/include/asm/arch_timer.h  |  78 +++++----
 arch/arm64/include/asm/esr.h         |   6 +
 arch/arm64/include/asm/sysreg.h      |   3 +
 arch/arm64/kernel/cpufeature.c       |  10 ++
 arch/arm64/kernel/traps.c            |  11 ++
 arch/arm64/tools/cpucaps             |   1 +
 drivers/clocksource/arm_arch_timer.c | 243 ++++++++++++++++-----------
 include/clocksource/arm_arch_timer.h |   2 +-
 9 files changed, 245 insertions(+), 146 deletions(-)

-- 
2.30.2

[PATCH v2 01/16] clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses

[Marc Zyngier]

As we are about to change the registers that are used by the driver,
start by adding build-time checks to ensure that we always handle
all registers and access modes.

Suggested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h    | 12 ++++++++++++
 arch/arm64/include/asm/arch_timer.h  | 13 ++++++++++++-
 drivers/clocksource/arm_arch_timer.c |  8 ++++++++
 3 files changed, 32 insertions(+), 1 deletion(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index 99175812d903..a5d27cff28fa 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -34,6 +34,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
 		case ARCH_TIMER_REG_TVAL:
 			asm volatile("mcr p15, 0, %0, c14, c2, 0" : : "r" (val));
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_VIRT_ACCESS) {
 		switch (reg) {
@@ -43,7 +45,11 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
 		case ARCH_TIMER_REG_TVAL:
 			asm volatile("mcr p15, 0, %0, c14, c3, 0" : : "r" (val));
 			break;
+		default:
+			BUILD_BUG();
 		}
+	} else {
+		BUILD_BUG();
 	}
 
 	isb();
@@ -62,6 +68,8 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		case ARCH_TIMER_REG_TVAL:
 			asm volatile("mrc p15, 0, %0, c14, c2, 0" : "=r" (val));
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_VIRT_ACCESS) {
 		switch (reg) {
@@ -71,7 +79,11 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		case ARCH_TIMER_REG_TVAL:
 			asm volatile("mrc p15, 0, %0, c14, c3, 0" : "=r" (val));
 			break;
+		default:
+			BUILD_BUG();
 		}
+	} else {
+		BUILD_BUG();
 	}
 
 	return val;
diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index 88d20f04c64a..fa12ea4a9812 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -112,6 +112,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
 		case ARCH_TIMER_REG_TVAL:
 			write_sysreg(val, cntp_tval_el0);
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_VIRT_ACCESS) {
 		switch (reg) {
@@ -121,7 +123,11 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
 		case ARCH_TIMER_REG_TVAL:
 			write_sysreg(val, cntv_tval_el0);
 			break;
+		default:
+			BUILD_BUG();
 		}
+	} else {
+		BUILD_BUG();
 	}
 
 	isb();
@@ -136,6 +142,8 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 			return read_sysreg(cntp_ctl_el0);
 		case ARCH_TIMER_REG_TVAL:
 			return arch_timer_reg_read_stable(cntp_tval_el0);
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_VIRT_ACCESS) {
 		switch (reg) {
@@ -143,10 +151,13 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 			return read_sysreg(cntv_ctl_el0);
 		case ARCH_TIMER_REG_TVAL:
 			return arch_timer_reg_read_stable(cntv_tval_el0);
+		default:
+			BUILD_BUG();
 		}
 	}
 
-	BUG();
+	BUILD_BUG();
+	unreachable();
 }
 
 static inline u32 arch_timer_get_cntfrq(void)
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index be6d741d404c..3b7d46d9db73 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -112,6 +112,8 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u32 val,
 		case ARCH_TIMER_REG_TVAL:
 			writel_relaxed(val, timer->base + CNTP_TVAL);
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_MEM_VIRT_ACCESS) {
 		struct arch_timer *timer = to_arch_timer(clk);
@@ -122,6 +124,8 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u32 val,
 		case ARCH_TIMER_REG_TVAL:
 			writel_relaxed(val, timer->base + CNTV_TVAL);
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else {
 		arch_timer_reg_write_cp15(access, reg, val);
@@ -143,6 +147,8 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
 		case ARCH_TIMER_REG_TVAL:
 			val = readl_relaxed(timer->base + CNTP_TVAL);
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else if (access == ARCH_TIMER_MEM_VIRT_ACCESS) {
 		struct arch_timer *timer = to_arch_timer(clk);
@@ -153,6 +159,8 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
 		case ARCH_TIMER_REG_TVAL:
 			val = readl_relaxed(timer->base + CNTV_TVAL);
 			break;
+		default:
+			BUILD_BUG();
 		}
 	} else {
 		val = arch_timer_reg_read_cp15(access, reg);
-- 
2.30.2

[PATCH v2 02/16] clocksource/arm_arch_timer: Drop CNT*_TVAL read accessors

[Marc Zyngier]

The arch timer driver never reads the various TVAL registers, only
writes to them. It is thus pointless to provide accessors
for them and to implement errata workarounds.

Drop these read-side accessors, and add a couple of BUG() statements
for the time being. These statements will be removed further down
the line.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h    |  6 ----
 arch/arm64/include/asm/arch_timer.h  | 17 -----------
 drivers/clocksource/arm_arch_timer.c | 44 ----------------------------
 3 files changed, 67 deletions(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index a5d27cff28fa..7d757085c61a 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -65,9 +65,6 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mrc p15, 0, %0, c14, c2, 1" : "=r" (val));
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mrc p15, 0, %0, c14, c2, 0" : "=r" (val));
-			break;
 		default:
 			BUILD_BUG();
 		}
@@ -76,9 +73,6 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mrc p15, 0, %0, c14, c3, 1" : "=r" (val));
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mrc p15, 0, %0, c14, c3, 0" : "=r" (val));
-			break;
 		default:
 			BUILD_BUG();
 		}
diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index fa12ea4a9812..8332fcfb08e8 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -52,8 +52,6 @@ struct arch_timer_erratum_workaround {
 	enum arch_timer_erratum_match_type match_type;
 	const void *id;
 	const char *desc;
-	u32 (*read_cntp_tval_el0)(void);
-	u32 (*read_cntv_tval_el0)(void);
 	u64 (*read_cntpct_el0)(void);
 	u64 (*read_cntvct_el0)(void);
 	int (*set_next_event_phys)(unsigned long, struct clock_event_device *);
@@ -64,17 +62,6 @@ struct arch_timer_erratum_workaround {
 DECLARE_PER_CPU(const struct arch_timer_erratum_workaround *,
 		timer_unstable_counter_workaround);
 
-/* inline sysreg accessors that make erratum_handler() work */
-static inline notrace u32 arch_timer_read_cntp_tval_el0(void)
-{
-	return read_sysreg(cntp_tval_el0);
-}
-
-static inline notrace u32 arch_timer_read_cntv_tval_el0(void)
-{
-	return read_sysreg(cntv_tval_el0);
-}
-
 static inline notrace u64 arch_timer_read_cntpct_el0(void)
 {
 	return read_sysreg(cntpct_el0);
@@ -140,8 +127,6 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			return read_sysreg(cntp_ctl_el0);
-		case ARCH_TIMER_REG_TVAL:
-			return arch_timer_reg_read_stable(cntp_tval_el0);
 		default:
 			BUILD_BUG();
 		}
@@ -149,8 +134,6 @@ u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			return read_sysreg(cntv_ctl_el0);
-		case ARCH_TIMER_REG_TVAL:
-			return arch_timer_reg_read_stable(cntv_tval_el0);
 		default:
 			BUILD_BUG();
 		}
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 3b7d46d9db73..67bdc7288f59 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -144,9 +144,6 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
 		case ARCH_TIMER_REG_CTRL:
 			val = readl_relaxed(timer->base + CNTP_CTL);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			val = readl_relaxed(timer->base + CNTP_TVAL);
-			break;
 		default:
 			BUILD_BUG();
 		}
@@ -156,9 +153,6 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
 		case ARCH_TIMER_REG_CTRL:
 			val = readl_relaxed(timer->base + CNTV_CTL);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			val = readl_relaxed(timer->base + CNTV_TVAL);
-			break;
 		default:
 			BUILD_BUG();
 		}
@@ -247,16 +241,6 @@ struct ate_acpi_oem_info {
 	_new;						\
 })
 
-static u32 notrace fsl_a008585_read_cntp_tval_el0(void)
-{
-	return __fsl_a008585_read_reg(cntp_tval_el0);
-}
-
-static u32 notrace fsl_a008585_read_cntv_tval_el0(void)
-{
-	return __fsl_a008585_read_reg(cntv_tval_el0);
-}
-
 static u64 notrace fsl_a008585_read_cntpct_el0(void)
 {
 	return __fsl_a008585_read_reg(cntpct_el0);
@@ -293,16 +277,6 @@ static u64 notrace fsl_a008585_read_cntvct_el0(void)
 	_new;							\
 })
 
-static u32 notrace hisi_161010101_read_cntp_tval_el0(void)
-{
-	return __hisi_161010101_read_reg(cntp_tval_el0);
-}
-
-static u32 notrace hisi_161010101_read_cntv_tval_el0(void)
-{
-	return __hisi_161010101_read_reg(cntv_tval_el0);
-}
-
 static u64 notrace hisi_161010101_read_cntpct_el0(void)
 {
 	return __hisi_161010101_read_reg(cntpct_el0);
@@ -387,16 +361,6 @@ static u64 notrace sun50i_a64_read_cntvct_el0(void)
 {
 	return __sun50i_a64_read_reg(cntvct_el0);
 }
-
-static u32 notrace sun50i_a64_read_cntp_tval_el0(void)
-{
-	return read_sysreg(cntp_cval_el0) - sun50i_a64_read_cntpct_el0();
-}
-
-static u32 notrace sun50i_a64_read_cntv_tval_el0(void)
-{
-	return read_sysreg(cntv_cval_el0) - sun50i_a64_read_cntvct_el0();
-}
 #endif
 
 #ifdef CONFIG_ARM_ARCH_TIMER_OOL_WORKAROUND
@@ -446,8 +410,6 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.match_type = ate_match_dt,
 		.id = "fsl,erratum-a008585",
 		.desc = "Freescale erratum a005858",
-		.read_cntp_tval_el0 = fsl_a008585_read_cntp_tval_el0,
-		.read_cntv_tval_el0 = fsl_a008585_read_cntv_tval_el0,
 		.read_cntpct_el0 = fsl_a008585_read_cntpct_el0,
 		.read_cntvct_el0 = fsl_a008585_read_cntvct_el0,
 		.set_next_event_phys = erratum_set_next_event_tval_phys,
@@ -459,8 +421,6 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.match_type = ate_match_dt,
 		.id = "hisilicon,erratum-161010101",
 		.desc = "HiSilicon erratum 161010101",
-		.read_cntp_tval_el0 = hisi_161010101_read_cntp_tval_el0,
-		.read_cntv_tval_el0 = hisi_161010101_read_cntv_tval_el0,
 		.read_cntpct_el0 = hisi_161010101_read_cntpct_el0,
 		.read_cntvct_el0 = hisi_161010101_read_cntvct_el0,
 		.set_next_event_phys = erratum_set_next_event_tval_phys,
@@ -470,8 +430,6 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.match_type = ate_match_acpi_oem_info,
 		.id = hisi_161010101_oem_info,
 		.desc = "HiSilicon erratum 161010101",
-		.read_cntp_tval_el0 = hisi_161010101_read_cntp_tval_el0,
-		.read_cntv_tval_el0 = hisi_161010101_read_cntv_tval_el0,
 		.read_cntpct_el0 = hisi_161010101_read_cntpct_el0,
 		.read_cntvct_el0 = hisi_161010101_read_cntvct_el0,
 		.set_next_event_phys = erratum_set_next_event_tval_phys,
@@ -492,8 +450,6 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.match_type = ate_match_dt,
 		.id = "allwinner,erratum-unknown1",
 		.desc = "Allwinner erratum UNKNOWN1",
-		.read_cntp_tval_el0 = sun50i_a64_read_cntp_tval_el0,
-		.read_cntv_tval_el0 = sun50i_a64_read_cntv_tval_el0,
 		.read_cntpct_el0 = sun50i_a64_read_cntpct_el0,
 		.read_cntvct_el0 = sun50i_a64_read_cntvct_el0,
 		.set_next_event_phys = erratum_set_next_event_tval_phys,
-- 
2.30.2

[PATCH v2 03/16] clocksource/arm_arch_timer: Extend write side of timer register accessors to u64

[Marc Zyngier]

The various accessors for the timer sysreg and MMIO registers are
currently hardwired to 32bit. However, we are about to introduce
the use of the CVAL registers, which require a 64bit access.

Upgrade the write side of the accessors to take a 64bit value
(the read side is left untouched as we don't plan to ever read
back any of these registers).

No functional change expected.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h    | 10 +++++-----
 arch/arm64/include/asm/arch_timer.h  |  2 +-
 drivers/clocksource/arm_arch_timer.c | 10 +++++-----
 3 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index 7d757085c61a..1482e70da7d3 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -24,15 +24,15 @@ int arch_timer_arch_init(void);
  * the code. At least it does so with a recent GCC (4.6.3).
  */
 static __always_inline
-void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
+void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 {
 	if (access == ARCH_TIMER_PHYS_ACCESS) {
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
-			asm volatile("mcr p15, 0, %0, c14, c2, 1" : : "r" (val));
+			asm volatile("mcr p15, 0, %0, c14, c2, 1" : : "r" ((u32)val));
 			break;
 		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mcr p15, 0, %0, c14, c2, 0" : : "r" (val));
+			asm volatile("mcr p15, 0, %0, c14, c2, 0" : : "r" ((u32)val));
 			break;
 		default:
 			BUILD_BUG();
@@ -40,10 +40,10 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
 	} else if (access == ARCH_TIMER_VIRT_ACCESS) {
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
-			asm volatile("mcr p15, 0, %0, c14, c3, 1" : : "r" (val));
+			asm volatile("mcr p15, 0, %0, c14, c3, 1" : : "r" ((u32)val));
 			break;
 		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mcr p15, 0, %0, c14, c3, 0" : : "r" (val));
+			asm volatile("mcr p15, 0, %0, c14, c3, 0" : : "r" ((u32)val));
 			break;
 		default:
 			BUILD_BUG();
diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index 8332fcfb08e8..43f827b680d0 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -89,7 +89,7 @@ static inline notrace u64 arch_timer_read_cntvct_el0(void)
  * the code.
  */
 static __always_inline
-void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u32 val)
+void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 {
 	if (access == ARCH_TIMER_PHYS_ACCESS) {
 		switch (reg) {
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 67bdc7288f59..a49bcefaa370 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -100,17 +100,17 @@ early_param("clocksource.arm_arch_timer.evtstrm", early_evtstrm_cfg);
  */
 
 static __always_inline
-void arch_timer_reg_write(int access, enum arch_timer_reg reg, u32 val,
+void arch_timer_reg_write(int access, enum arch_timer_reg reg, u64 val,
 			  struct clock_event_device *clk)
 {
 	if (access == ARCH_TIMER_MEM_PHYS_ACCESS) {
 		struct arch_timer *timer = to_arch_timer(clk);
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
-			writel_relaxed(val, timer->base + CNTP_CTL);
+			writel_relaxed((u32)val, timer->base + CNTP_CTL);
 			break;
 		case ARCH_TIMER_REG_TVAL:
-			writel_relaxed(val, timer->base + CNTP_TVAL);
+			writel_relaxed((u32)val, timer->base + CNTP_TVAL);
 			break;
 		default:
 			BUILD_BUG();
@@ -119,10 +119,10 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u32 val,
 		struct arch_timer *timer = to_arch_timer(clk);
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
-			writel_relaxed(val, timer->base + CNTV_CTL);
+			writel_relaxed((u32)val, timer->base + CNTV_CTL);
 			break;
 		case ARCH_TIMER_REG_TVAL:
-			writel_relaxed(val, timer->base + CNTV_TVAL);
+			writel_relaxed((u32)val, timer->base + CNTV_TVAL);
 			break;
 		default:
 			BUILD_BUG();
-- 
2.30.2

[PATCH v2 04/16] clocksource/arm_arch_timer: Move system register timer programming over to CVAL

[Marc Zyngier]

In order to cope better with high frequency counters, move the
programming of the timers from the countdown timer (TVAL) over
to the comparator (CVAL).

The programming model is slightly different, as we now need to
read the current counter value to have an absolute deadline
instead of a relative one.

There is a small overhead to this change, which we will address
in the following patches.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h    |  8 ++++----
 arch/arm64/include/asm/arch_timer.h  | 10 +++++-----
 drivers/clocksource/arm_arch_timer.c | 26 +++++++++++++++++++++++---
 include/clocksource/arm_arch_timer.h |  1 +
 4 files changed, 33 insertions(+), 12 deletions(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index 1482e70da7d3..a9b2b721c7f9 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -31,8 +31,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mcr p15, 0, %0, c14, c2, 1" : : "r" ((u32)val));
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mcr p15, 0, %0, c14, c2, 0" : : "r" ((u32)val));
+		case ARCH_TIMER_REG_CVAL:
+			asm volatile("mcrr p15, 2, %Q0, %R0, c14" : : "r" (val));
 			break;
 		default:
 			BUILD_BUG();
@@ -42,8 +42,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mcr p15, 0, %0, c14, c3, 1" : : "r" ((u32)val));
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			asm volatile("mcr p15, 0, %0, c14, c3, 0" : : "r" ((u32)val));
+		case ARCH_TIMER_REG_CVAL:
+			asm volatile("mcrr p15, 3, %Q0, %R0, c14" : : "r" (val));
 			break;
 		default:
 			BUILD_BUG();
diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index 43f827b680d0..4f4aa13dd01e 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -96,8 +96,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		case ARCH_TIMER_REG_CTRL:
 			write_sysreg(val, cntp_ctl_el0);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			write_sysreg(val, cntp_tval_el0);
+		case ARCH_TIMER_REG_CVAL:
+			write_sysreg(val, cntp_cval_el0);
 			break;
 		default:
 			BUILD_BUG();
@@ -107,8 +107,8 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		case ARCH_TIMER_REG_CTRL:
 			write_sysreg(val, cntv_ctl_el0);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			write_sysreg(val, cntv_tval_el0);
+		case ARCH_TIMER_REG_CVAL:
+			write_sysreg(val, cntv_cval_el0);
 			break;
 		default:
 			BUILD_BUG();
@@ -121,7 +121,7 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 }
 
 static __always_inline
-u32 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
+u64 arch_timer_reg_read_cp15(int access, enum arch_timer_reg reg)
 {
 	if (access == ARCH_TIMER_PHYS_ACCESS) {
 		switch (reg) {
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index a49bcefaa370..322165468edf 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -691,10 +691,18 @@ static __always_inline void set_next_event(const int access, unsigned long evt,
 					   struct clock_event_device *clk)
 {
 	unsigned long ctrl;
+	u64 cnt;
+
 	ctrl = arch_timer_reg_read(access, ARCH_TIMER_REG_CTRL, clk);
 	ctrl |= ARCH_TIMER_CTRL_ENABLE;
 	ctrl &= ~ARCH_TIMER_CTRL_IT_MASK;
-	arch_timer_reg_write(access, ARCH_TIMER_REG_TVAL, evt, clk);
+
+	if (access == ARCH_TIMER_PHYS_ACCESS)
+		cnt = __arch_counter_get_cntpct();
+	else
+		cnt = __arch_counter_get_cntvct();
+
+	arch_timer_reg_write(access, ARCH_TIMER_REG_CVAL, evt + cnt, clk);
 	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
 }
 
@@ -712,17 +720,29 @@ static int arch_timer_set_next_event_phys(unsigned long evt,
 	return 0;
 }
 
+static __always_inline void set_next_event_mem(const int access, unsigned long evt,
+					   struct clock_event_device *clk)
+{
+	unsigned long ctrl;
+	ctrl = arch_timer_reg_read(access, ARCH_TIMER_REG_CTRL, clk);
+	ctrl |= ARCH_TIMER_CTRL_ENABLE;
+	ctrl &= ~ARCH_TIMER_CTRL_IT_MASK;
+
+	arch_timer_reg_write(access, ARCH_TIMER_REG_TVAL, evt, clk);
+	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
+}
+
 static int arch_timer_set_next_event_virt_mem(unsigned long evt,
 					      struct clock_event_device *clk)
 {
-	set_next_event(ARCH_TIMER_MEM_VIRT_ACCESS, evt, clk);
+	set_next_event_mem(ARCH_TIMER_MEM_VIRT_ACCESS, evt, clk);
 	return 0;
 }
 
 static int arch_timer_set_next_event_phys_mem(unsigned long evt,
 					      struct clock_event_device *clk)
 {
-	set_next_event(ARCH_TIMER_MEM_PHYS_ACCESS, evt, clk);
+	set_next_event_mem(ARCH_TIMER_MEM_PHYS_ACCESS, evt, clk);
 	return 0;
 }
 
diff --git a/include/clocksource/arm_arch_timer.h b/include/clocksource/arm_arch_timer.h
index 73c7139c866f..d59537afb29d 100644
--- a/include/clocksource/arm_arch_timer.h
+++ b/include/clocksource/arm_arch_timer.h
@@ -25,6 +25,7 @@
 enum arch_timer_reg {
 	ARCH_TIMER_REG_CTRL,
 	ARCH_TIMER_REG_TVAL,
+	ARCH_TIMER_REG_CVAL,
 };
 
 enum arch_timer_ppi_nr {
-- 
2.30.2

[PATCH v2 05/16] clocksource/arm_arch_timer: Move drop _tval from erratum function names

[Marc Zyngier]

The '_tval' name in the erratum handling function names doesn't
make much sense anymore (and they were using CVAL the first place).

Drop the _tval tag.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 drivers/clocksource/arm_arch_timer.c | 26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 322165468edf..8afe8c814eba 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -369,7 +369,7 @@ EXPORT_SYMBOL_GPL(timer_unstable_counter_workaround);
 
 static atomic_t timer_unstable_counter_workaround_in_use = ATOMIC_INIT(0);
 
-static void erratum_set_next_event_tval_generic(const int access, unsigned long evt,
+static void erratum_set_next_event_generic(const int access, unsigned long evt,
 						struct clock_event_device *clk)
 {
 	unsigned long ctrl;
@@ -390,17 +390,17 @@ static void erratum_set_next_event_tval_generic(const int access, unsigned long
 	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
 }
 
-static __maybe_unused int erratum_set_next_event_tval_virt(unsigned long evt,
+static __maybe_unused int erratum_set_next_event_virt(unsigned long evt,
 					    struct clock_event_device *clk)
 {
-	erratum_set_next_event_tval_generic(ARCH_TIMER_VIRT_ACCESS, evt, clk);
+	erratum_set_next_event_generic(ARCH_TIMER_VIRT_ACCESS, evt, clk);
 	return 0;
 }
 
-static __maybe_unused int erratum_set_next_event_tval_phys(unsigned long evt,
+static __maybe_unused int erratum_set_next_event_phys(unsigned long evt,
 					    struct clock_event_device *clk)
 {
-	erratum_set_next_event_tval_generic(ARCH_TIMER_PHYS_ACCESS, evt, clk);
+	erratum_set_next_event_generic(ARCH_TIMER_PHYS_ACCESS, evt, clk);
 	return 0;
 }
 
@@ -412,8 +412,8 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.desc = "Freescale erratum a005858",
 		.read_cntpct_el0 = fsl_a008585_read_cntpct_el0,
 		.read_cntvct_el0 = fsl_a008585_read_cntvct_el0,
-		.set_next_event_phys = erratum_set_next_event_tval_phys,
-		.set_next_event_virt = erratum_set_next_event_tval_virt,
+		.set_next_event_phys = erratum_set_next_event_phys,
+		.set_next_event_virt = erratum_set_next_event_virt,
 	},
 #endif
 #ifdef CONFIG_HISILICON_ERRATUM_161010101
@@ -423,8 +423,8 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.desc = "HiSilicon erratum 161010101",
 		.read_cntpct_el0 = hisi_161010101_read_cntpct_el0,
 		.read_cntvct_el0 = hisi_161010101_read_cntvct_el0,
-		.set_next_event_phys = erratum_set_next_event_tval_phys,
-		.set_next_event_virt = erratum_set_next_event_tval_virt,
+		.set_next_event_phys = erratum_set_next_event_phys,
+		.set_next_event_virt = erratum_set_next_event_virt,
 	},
 	{
 		.match_type = ate_match_acpi_oem_info,
@@ -432,8 +432,8 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.desc = "HiSilicon erratum 161010101",
 		.read_cntpct_el0 = hisi_161010101_read_cntpct_el0,
 		.read_cntvct_el0 = hisi_161010101_read_cntvct_el0,
-		.set_next_event_phys = erratum_set_next_event_tval_phys,
-		.set_next_event_virt = erratum_set_next_event_tval_virt,
+		.set_next_event_phys = erratum_set_next_event_phys,
+		.set_next_event_virt = erratum_set_next_event_virt,
 	},
 #endif
 #ifdef CONFIG_ARM64_ERRATUM_858921
@@ -452,8 +452,8 @@ static const struct arch_timer_erratum_workaround ool_workarounds[] = {
 		.desc = "Allwinner erratum UNKNOWN1",
 		.read_cntpct_el0 = sun50i_a64_read_cntpct_el0,
 		.read_cntvct_el0 = sun50i_a64_read_cntvct_el0,
-		.set_next_event_phys = erratum_set_next_event_tval_phys,
-		.set_next_event_virt = erratum_set_next_event_tval_virt,
+		.set_next_event_phys = erratum_set_next_event_phys,
+		.set_next_event_virt = erratum_set_next_event_virt,
 	},
 #endif
 #ifdef CONFIG_ARM64_ERRATUM_1418040
-- 
2.30.2

[PATCH v2 06/16] clocksource/arm_arch_timer: Fix MMIO base address vs callback ordering issue

[Marc Zyngier]

The MMIO timer base address gets published after we have registered
the callbacks and the interrupt handler, which is... a bit dangerous.

Fix this by moving the base address publication to the point where
we register the timer, and expose a pointer to the timer structure
itself rather than a naked value.

Reviewed-by: Oliver Upton <oupton@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 drivers/clocksource/arm_arch_timer.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 8afe8c814eba..bede10f67f9a 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -54,13 +54,13 @@
 
 static unsigned arch_timers_present __initdata;
 
-static void __iomem *arch_counter_base __ro_after_init;
-
 struct arch_timer {
 	void __iomem *base;
 	struct clock_event_device evt;
 };
 
+static struct arch_timer *arch_timer_mem __ro_after_init;
+
 #define to_arch_timer(e) container_of(e, struct arch_timer, evt)
 
 static u32 arch_timer_rate __ro_after_init;
@@ -973,9 +973,9 @@ static u64 arch_counter_get_cntvct_mem(void)
 	u32 vct_lo, vct_hi, tmp_hi;
 
 	do {
-		vct_hi = readl_relaxed(arch_counter_base + CNTVCT_HI);
-		vct_lo = readl_relaxed(arch_counter_base + CNTVCT_LO);
-		tmp_hi = readl_relaxed(arch_counter_base + CNTVCT_HI);
+		vct_hi = readl_relaxed(arch_timer_mem->base + CNTVCT_HI);
+		vct_lo = readl_relaxed(arch_timer_mem->base + CNTVCT_LO);
+		tmp_hi = readl_relaxed(arch_timer_mem->base + CNTVCT_HI);
 	} while (vct_hi != tmp_hi);
 
 	return ((u64) vct_hi << 32) | vct_lo;
@@ -1166,25 +1166,25 @@ static int __init arch_timer_mem_register(void __iomem *base, unsigned int irq)
 {
 	int ret;
 	irq_handler_t func;
-	struct arch_timer *t;
 
-	t = kzalloc(sizeof(*t), GFP_KERNEL);
-	if (!t)
+	arch_timer_mem = kzalloc(sizeof(*arch_timer_mem), GFP_KERNEL);
+	if (!arch_timer_mem)
 		return -ENOMEM;
 
-	t->base = base;
-	t->evt.irq = irq;
-	__arch_timer_setup(ARCH_TIMER_TYPE_MEM, &t->evt);
+	arch_timer_mem->base = base;
+	arch_timer_mem->evt.irq = irq;
+	__arch_timer_setup(ARCH_TIMER_TYPE_MEM, &arch_timer_mem->evt);
 
 	if (arch_timer_mem_use_virtual)
 		func = arch_timer_handler_virt_mem;
 	else
 		func = arch_timer_handler_phys_mem;
 
-	ret = request_irq(irq, func, IRQF_TIMER, "arch_mem_timer", &t->evt);
+	ret = request_irq(irq, func, IRQF_TIMER, "arch_mem_timer", &arch_timer_mem->evt);
 	if (ret) {
 		pr_err("Failed to request mem timer irq\n");
-		kfree(t);
+		kfree(arch_timer_mem);
+		arch_timer_mem = NULL;
 	}
 
 	return ret;
@@ -1442,7 +1442,6 @@ arch_timer_mem_frame_register(struct arch_timer_mem_frame *frame)
 		return ret;
 	}
 
-	arch_counter_base = base;
 	arch_timers_present |= ARCH_TIMER_TYPE_MEM;
 
 	return 0;
-- 
2.30.2

[PATCH v2 07/16] clocksource/arm_arch_timer: Move MMIO timer programming over to CVAL

[Marc Zyngier]

Similarily to the sysreg-based timer, move the MMIO over to using
the CVAL registers instead of TVAL. Note that there is no warranty
that the 64bit MMIO access will be atomic, but the timer is always
disabled at the point where we program CVAL.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h    |  1 +
 drivers/clocksource/arm_arch_timer.c | 50 +++++++++++++++++++++-------
 2 files changed, 39 insertions(+), 12 deletions(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index a9b2b721c7f9..9f4b895b78f7 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -7,6 +7,7 @@
 #include <asm/hwcap.h>
 #include <linux/clocksource.h>
 #include <linux/init.h>
+#include <linux/io-64-nonatomic-lo-hi.h>
 #include <linux/types.h>
 
 #include <clocksource/arm_arch_timer.h>
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index bede10f67f9a..f4db3a65bc79 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -44,11 +44,13 @@
 #define CNTACR_RWVT	BIT(4)
 #define CNTACR_RWPT	BIT(5)
 
-#define CNTVCT_LO	0x08
-#define CNTVCT_HI	0x0c
+#define CNTVCT_LO	0x00
+#define CNTPCT_LO	0x08
 #define CNTFRQ		0x10
+#define CNTP_CVAL_LO	0x20
 #define CNTP_TVAL	0x28
 #define CNTP_CTL	0x2c
+#define CNTV_CVAL_LO	0x30
 #define CNTV_TVAL	0x38
 #define CNTV_CTL	0x3c
 
@@ -112,6 +114,13 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u64 val,
 		case ARCH_TIMER_REG_TVAL:
 			writel_relaxed((u32)val, timer->base + CNTP_TVAL);
 			break;
+		case ARCH_TIMER_REG_CVAL:
+			/*
+			 * Not guaranteed to be atomic, so the timer
+			 * must be disabled at this point.
+			 */
+			writeq_relaxed(val, timer->base + CNTP_CVAL_LO);
+			break;
 		default:
 			BUILD_BUG();
 		}
@@ -124,6 +133,10 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u64 val,
 		case ARCH_TIMER_REG_TVAL:
 			writel_relaxed((u32)val, timer->base + CNTV_TVAL);
 			break;
+		case ARCH_TIMER_REG_CVAL:
+			/* Same restriction as above */
+			writeq_relaxed(val, timer->base + CNTV_CVAL_LO);
+			break;
 		default:
 			BUILD_BUG();
 		}
@@ -720,15 +733,36 @@ static int arch_timer_set_next_event_phys(unsigned long evt,
 	return 0;
 }
 
+static u64 arch_counter_get_cnt_mem(struct arch_timer *t, int offset_lo)
+{
+	u32 cnt_lo, cnt_hi, tmp_hi;
+
+	do {
+		cnt_hi = readl_relaxed(t->base + offset_lo + 4);
+		cnt_lo = readl_relaxed(t->base + offset_lo);
+		tmp_hi = readl_relaxed(t->base + offset_lo + 4);
+	} while (cnt_hi != tmp_hi);
+
+	return ((u64) cnt_hi << 32) | cnt_lo;
+}
+
 static __always_inline void set_next_event_mem(const int access, unsigned long evt,
 					   struct clock_event_device *clk)
 {
+	struct arch_timer *timer = to_arch_timer(clk);
 	unsigned long ctrl;
+	u64 cnt;
+
 	ctrl = arch_timer_reg_read(access, ARCH_TIMER_REG_CTRL, clk);
 	ctrl |= ARCH_TIMER_CTRL_ENABLE;
 	ctrl &= ~ARCH_TIMER_CTRL_IT_MASK;
 
-	arch_timer_reg_write(access, ARCH_TIMER_REG_TVAL, evt, clk);
+	if (access ==  ARCH_TIMER_MEM_VIRT_ACCESS)
+		cnt = arch_counter_get_cnt_mem(timer, CNTVCT_LO);
+	else
+		cnt = arch_counter_get_cnt_mem(timer, CNTPCT_LO);
+
+	arch_timer_reg_write(access, ARCH_TIMER_REG_CVAL, evt + cnt, clk);
 	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
 }
 
@@ -970,15 +1004,7 @@ bool arch_timer_evtstrm_available(void)
 
 static u64 arch_counter_get_cntvct_mem(void)
 {
-	u32 vct_lo, vct_hi, tmp_hi;
-
-	do {
-		vct_hi = readl_relaxed(arch_timer_mem->base + CNTVCT_HI);
-		vct_lo = readl_relaxed(arch_timer_mem->base + CNTVCT_LO);
-		tmp_hi = readl_relaxed(arch_timer_mem->base + CNTVCT_HI);
-	} while (vct_hi != tmp_hi);
-
-	return ((u64) vct_hi << 32) | vct_lo;
+	return arch_counter_get_cnt_mem(arch_timer_mem, CNTVCT_LO);
 }
 
 static struct arch_timer_kvm_info arch_timer_kvm_info;
-- 
2.30.2

[PATCH v2 08/16] clocksource/arm_arch_timer: Advertise 56bit timer to the core code

[Marc Zyngier]

Proudly tell the code code that we have a timer able to handle
56 bits deltas.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 drivers/clocksource/arm_arch_timer.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index f4db3a65bc79..36e091412151 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -834,7 +834,7 @@ static void __arch_timer_setup(unsigned type,
 
 	clk->set_state_shutdown(clk);
 
-	clockevents_config_and_register(clk, arch_timer_rate, 0xf, 0x7fffffff);
+	clockevents_config_and_register(clk, arch_timer_rate, 0xf, CLOCKSOURCE_MASK(56));
 }
 
 static void arch_timer_evtstrm_enable(int divider)
-- 
2.30.2

[PATCH v2 09/16] clocksource/arm_arch_timer: Work around broken CVAL implementations

[Marc Zyngier]

The Applied Micro XGene-1 SoC has a busted implementation of the
CVAL register: it looks like it is based on TVAL instead of the
other way around. The net effect of this implementation blunder
is that the maximum deadline you can program in the timer is
32bit wide.

Use a MIDR check to notice the broken CPU, and reduce the width
of the timer to 32bit.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 drivers/clocksource/arm_arch_timer.c | 28 +++++++++++++++++++++++++++-
 1 file changed, 27 insertions(+), 1 deletion(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 36e091412151..ef3f83865dcd 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -780,9 +780,32 @@ static int arch_timer_set_next_event_phys_mem(unsigned long evt,
 	return 0;
 }
 
+static u64 __arch_timer_check_delta(void)
+{
+#ifdef CONFIG_ARM64
+	const struct midr_range broken_cval_midrs[] = {
+		/*
+		 * XGene-1 implements CVAL in terms of TVAL, meaning
+		 * that the maximum timer range is 32bit. Shame on them.
+		 */
+		MIDR_ALL_VERSIONS(MIDR_CPU_MODEL(ARM_CPU_IMP_APM,
+						 APM_CPU_PART_POTENZA)),
+		{},
+	};
+
+	if (is_midr_in_range_list(read_cpuid_id(), broken_cval_midrs)) {
+		pr_warn_once("Broken CNTx_CVAL_EL1, limiting width to 32bits");
+		return CLOCKSOURCE_MASK(32);
+	}
+#endif
+	return CLOCKSOURCE_MASK(56);
+}
+
 static void __arch_timer_setup(unsigned type,
 			       struct clock_event_device *clk)
 {
+	u64 max_delta;
+
 	clk->features = CLOCK_EVT_FEAT_ONESHOT;
 
 	if (type == ARCH_TIMER_TYPE_CP15) {
@@ -814,6 +837,7 @@ static void __arch_timer_setup(unsigned type,
 		}
 
 		clk->set_next_event = sne;
+		max_delta = __arch_timer_check_delta();
 	} else {
 		clk->features |= CLOCK_EVT_FEAT_DYNIRQ;
 		clk->name = "arch_mem_timer";
@@ -830,11 +854,13 @@ static void __arch_timer_setup(unsigned type,
 			clk->set_next_event =
 				arch_timer_set_next_event_phys_mem;
 		}
+
+		max_delta = CLOCKSOURCE_MASK(56);
 	}
 
 	clk->set_state_shutdown(clk);
 
-	clockevents_config_and_register(clk, arch_timer_rate, 0xf, CLOCKSOURCE_MASK(56));
+	clockevents_config_and_register(clk, arch_timer_rate, 0xf, max_delta);
 }
 
 static void arch_timer_evtstrm_enable(int divider)
-- 
2.30.2

[PATCH v2 10/16] clocksource/arm_arch_timer: Remove any trace of the TVAL programming interface

[Marc Zyngier]

TVAL usage is now long gone, get rid of the leftovers.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 drivers/clocksource/arm_arch_timer.c | 8 --------
 include/clocksource/arm_arch_timer.h | 1 -
 2 files changed, 9 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index ef3f83865dcd..6e20bc12dc35 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -48,10 +48,8 @@
 #define CNTPCT_LO	0x08
 #define CNTFRQ		0x10
 #define CNTP_CVAL_LO	0x20
-#define CNTP_TVAL	0x28
 #define CNTP_CTL	0x2c
 #define CNTV_CVAL_LO	0x30
-#define CNTV_TVAL	0x38
 #define CNTV_CTL	0x3c
 
 static unsigned arch_timers_present __initdata;
@@ -111,9 +109,6 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u64 val,
 		case ARCH_TIMER_REG_CTRL:
 			writel_relaxed((u32)val, timer->base + CNTP_CTL);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			writel_relaxed((u32)val, timer->base + CNTP_TVAL);
-			break;
 		case ARCH_TIMER_REG_CVAL:
 			/*
 			 * Not guaranteed to be atomic, so the timer
@@ -130,9 +125,6 @@ void arch_timer_reg_write(int access, enum arch_timer_reg reg, u64 val,
 		case ARCH_TIMER_REG_CTRL:
 			writel_relaxed((u32)val, timer->base + CNTV_CTL);
 			break;
-		case ARCH_TIMER_REG_TVAL:
-			writel_relaxed((u32)val, timer->base + CNTV_TVAL);
-			break;
 		case ARCH_TIMER_REG_CVAL:
 			/* Same restriction as above */
 			writeq_relaxed(val, timer->base + CNTV_CVAL_LO);
diff --git a/include/clocksource/arm_arch_timer.h b/include/clocksource/arm_arch_timer.h
index d59537afb29d..e715bdb720d5 100644
--- a/include/clocksource/arm_arch_timer.h
+++ b/include/clocksource/arm_arch_timer.h
@@ -24,7 +24,6 @@
 
 enum arch_timer_reg {
 	ARCH_TIMER_REG_CTRL,
-	ARCH_TIMER_REG_TVAL,
 	ARCH_TIMER_REG_CVAL,
 };
 
-- 
2.30.2

[PATCH v2 11/16] clocksource/arm_arch_timer: Drop unnecessary ISB on CVAL programming

[Marc Zyngier]

Switching from TVAL to CVAL has a small drawback: we need an ISB
before reading the counter. We cannot get rid of it, but we can
instead remove the one that comes just after writing to CVAL.

This reduces the number of ISBs from 3 to 2 when programming
the timer.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm/include/asm/arch_timer.h   | 4 ++--
 arch/arm64/include/asm/arch_timer.h | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/arm/include/asm/arch_timer.h b/arch/arm/include/asm/arch_timer.h
index 9f4b895b78f7..bb129b6d2366 100644
--- a/arch/arm/include/asm/arch_timer.h
+++ b/arch/arm/include/asm/arch_timer.h
@@ -31,6 +31,7 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mcr p15, 0, %0, c14, c2, 1" : : "r" ((u32)val));
+			isb();
 			break;
 		case ARCH_TIMER_REG_CVAL:
 			asm volatile("mcrr p15, 2, %Q0, %R0, c14" : : "r" (val));
@@ -42,6 +43,7 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			asm volatile("mcr p15, 0, %0, c14, c3, 1" : : "r" ((u32)val));
+			isb();
 			break;
 		case ARCH_TIMER_REG_CVAL:
 			asm volatile("mcrr p15, 3, %Q0, %R0, c14" : : "r" (val));
@@ -52,8 +54,6 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 	} else {
 		BUILD_BUG();
 	}
-
-	isb();
 }
 
 static __always_inline
diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index 4f4aa13dd01e..b8000ef71a2c 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -95,6 +95,7 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			write_sysreg(val, cntp_ctl_el0);
+			isb();
 			break;
 		case ARCH_TIMER_REG_CVAL:
 			write_sysreg(val, cntp_cval_el0);
@@ -106,6 +107,7 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 		switch (reg) {
 		case ARCH_TIMER_REG_CTRL:
 			write_sysreg(val, cntv_ctl_el0);
+			isb();
 			break;
 		case ARCH_TIMER_REG_CVAL:
 			write_sysreg(val, cntv_cval_el0);
@@ -116,8 +118,6 @@ void arch_timer_reg_write_cp15(int access, enum arch_timer_reg reg, u64 val)
 	} else {
 		BUILD_BUG();
 	}
-
-	isb();
 }
 
 static __always_inline
-- 
2.30.2

[PATCH v2 12/16] clocksource/arm_arch_timer: Fix masking for high freq counters

[Marc Zyngier]

From: Oliver Upton <oupton@google.com>

Unfortunately, the architecture provides no means to determine the bit
width of the system counter. However, we do know the following from the
specification:

 - the system counter is at least 56 bits wide
 - Roll-over time of not less than 40 years

To date, the arch timer driver has depended on the first property,
assuming any system counter to be 56 bits wide and masking off the rest.
However, combining a narrow clocksource mask with a high frequency
counter could result in prematurely wrapping the system counter by a
significant margin. For example, a 56 bit wide, 1GHz system counter
would wrap in a mere 2.28 years!

This is a problem for two reasons: v8.6+ implementations are required to
provide a 64 bit, 1GHz system counter. Furthermore, before v8.6,
implementers may select a counter frequency of their choosing.

Fix the issue by deriving a valid clock mask based on the second
property from above. Set the floor at 56 bits, since we know no system
counter is narrower than that.

Suggested-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Oliver Upton <oupton@google.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
[maz: fixed width computation not to lose the last bit, added
      max delta generation for the timer]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210807191428.3488948-1-oupton@google.com
---
 drivers/clocksource/arm_arch_timer.c | 34 ++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 5 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 6e20bc12dc35..9a04eacc4412 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -52,6 +52,12 @@
 #define CNTV_CVAL_LO	0x30
 #define CNTV_CTL	0x3c
 
+/*
+ * The minimum amount of time a generic counter is guaranteed to not roll over
+ * (40 years)
+ */
+#define MIN_ROLLOVER_SECS	(40ULL * 365 * 24 * 3600)
+
 static unsigned arch_timers_present __initdata;
 
 struct arch_timer {
@@ -95,6 +101,22 @@ static int __init early_evtstrm_cfg(char *buf)
 }
 early_param("clocksource.arm_arch_timer.evtstrm", early_evtstrm_cfg);
 
+/*
+ * Makes an educated guess at a valid counter width based on the Generic Timer
+ * specification. Of note:
+ *   1) the system counter is at least 56 bits wide
+ *   2) a roll-over time of not less than 40 years
+ *
+ * See 'ARM DDI 0487G.a D11.1.2 ("The system counter")' for more details.
+ */
+static int arch_counter_get_width(void)
+{
+	u64 min_cycles = MIN_ROLLOVER_SECS * arch_timer_rate;
+
+	/* guarantee the returned width is within the valid range */
+	return clamp_val(ilog2(min_cycles - 1) + 1, 56, 64);
+}
+
 /*
  * Architected system timer support.
  */
@@ -212,13 +234,11 @@ static struct clocksource clocksource_counter = {
 	.id	= CSID_ARM_ARCH_COUNTER,
 	.rating	= 400,
 	.read	= arch_counter_read,
-	.mask	= CLOCKSOURCE_MASK(56),
 	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
 };
 
 static struct cyclecounter cyclecounter __ro_after_init = {
 	.read	= arch_counter_read_cc,
-	.mask	= CLOCKSOURCE_MASK(56),
 };
 
 struct ate_acpi_oem_info {
@@ -790,7 +810,7 @@ static u64 __arch_timer_check_delta(void)
 		return CLOCKSOURCE_MASK(32);
 	}
 #endif
-	return CLOCKSOURCE_MASK(56);
+	return CLOCKSOURCE_MASK(arch_counter_get_width());
 }
 
 static void __arch_timer_setup(unsigned type,
@@ -1035,6 +1055,7 @@ struct arch_timer_kvm_info *arch_timer_get_kvm_info(void)
 static void __init arch_counter_register(unsigned type)
 {
 	u64 start_count;
+	int width;
 
 	/* Register the CP15 based counter if we have one */
 	if (type & ARCH_TIMER_TYPE_CP15) {
@@ -1059,6 +1080,10 @@ static void __init arch_counter_register(unsigned type)
 		arch_timer_read_counter = arch_counter_get_cntvct_mem;
 	}
 
+	width = arch_counter_get_width();
+	clocksource_counter.mask = CLOCKSOURCE_MASK(width);
+	cyclecounter.mask = CLOCKSOURCE_MASK(width);
+
 	if (!arch_counter_suspend_stop)
 		clocksource_counter.flags |= CLOCK_SOURCE_SUSPEND_NONSTOP;
 	start_count = arch_timer_read_counter();
@@ -1068,8 +1093,7 @@ static void __init arch_counter_register(unsigned type)
 	timecounter_init(&arch_timer_kvm_info.timecounter,
 			 &cyclecounter, start_count);
 
-	/* 56 bits minimum, so we assume worst case rollover */
-	sched_clock_register(arch_timer_read_counter, 56, arch_timer_rate);
+	sched_clock_register(arch_timer_read_counter, width, arch_timer_rate);
 }
 
 static void arch_timer_stop(struct clock_event_device *clk)
-- 
2.30.2

[PATCH v2 13/16] clocksource/arch_arm_timer: Move workaround synchronisation around

[Marc Zyngier]

We currently handle synchronisation when workarounds are enabled
by having an ISB in the __arch_counter_get_cnt?ct_stable() helpers.

WHile this works, this prevents us from relaxing this synchronisation.

Instead, move it closer to the point where the synchronisation is
actually needed. Further patches will subsequently relax this.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm64/include/asm/arch_timer.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index b8000ef71a2c..519ac1f7f859 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -32,7 +32,7 @@
 	({								\
 		const struct arch_timer_erratum_workaround *__wa;	\
 		__wa = __this_cpu_read(timer_unstable_counter_workaround); \
-		(__wa && __wa->h) ? __wa->h : arch_timer_##h;		\
+		(__wa && __wa->h) ? ({ isb(); __wa->h;}) : arch_timer_##h; \
 	})
 
 #else
@@ -64,11 +64,13 @@ DECLARE_PER_CPU(const struct arch_timer_erratum_workaround *,
 
 static inline notrace u64 arch_timer_read_cntpct_el0(void)
 {
+	isb();
 	return read_sysreg(cntpct_el0);
 }
 
 static inline notrace u64 arch_timer_read_cntvct_el0(void)
 {
+	isb();
 	return read_sysreg(cntvct_el0);
 }
 
@@ -163,7 +165,6 @@ static __always_inline u64 __arch_counter_get_cntpct_stable(void)
 {
 	u64 cnt;
 
-	isb();
 	cnt = arch_timer_reg_read_stable(cntpct_el0);
 	arch_counter_enforce_ordering(cnt);
 	return cnt;
@@ -183,7 +184,6 @@ static __always_inline u64 __arch_counter_get_cntvct_stable(void)
 {
 	u64 cnt;
 
-	isb();
 	cnt = arch_timer_reg_read_stable(cntvct_el0);
 	arch_counter_enforce_ordering(cnt);
 	return cnt;
-- 
2.30.2

[PATCH v2 14/16] arm64: Add a capability for FEAT_ECV

[Marc Zyngier]

Add a new capability to detect the Enhanced Counter Virtualization
feature (FEAT_ECV).

Reviewed-by: Oliver Upton <oupton@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm64/kernel/cpufeature.c | 10 ++++++++++
 arch/arm64/tools/cpucaps       |  1 +
 2 files changed, 11 insertions(+)

diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index f8a3067d10c6..26b11ce8fff6 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1926,6 +1926,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
 		.sign = FTR_UNSIGNED,
 		.min_field_value = 1,
 	},
+	{
+		.desc = "Enhanced Counter Virtualization",
+		.capability = ARM64_HAS_ECV,
+		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
+		.matches = has_cpuid_feature,
+		.sys_reg = SYS_ID_AA64MMFR0_EL1,
+		.field_pos = ID_AA64MMFR0_ECV_SHIFT,
+		.sign = FTR_UNSIGNED,
+		.min_field_value = 1,
+	},
 #ifdef CONFIG_ARM64_PAN
 	{
 		.desc = "Privileged Access Never",
diff --git a/arch/arm64/tools/cpucaps b/arch/arm64/tools/cpucaps
index 49305c2e6dfd..7a7c58acd8f0 100644
--- a/arch/arm64/tools/cpucaps
+++ b/arch/arm64/tools/cpucaps
@@ -18,6 +18,7 @@ HAS_CRC32
 HAS_DCPODP
 HAS_DCPOP
 HAS_E0PD
+HAS_ECV
 HAS_EPAN
 HAS_GENERIC_AUTH
 HAS_GENERIC_AUTH_ARCH
-- 
2.30.2

[PATCH v2 15/16] arm64: Add CNT{P,V}CTSS_EL0 alternatives to cnt{p,v}ct_el0

[Marc Zyngier]

CNTPCTSS_EL0 and CNTVCTSS_EL0 are alternatives to the usual
CNTPCT_EL0 and CNTVCT_EL0 that do not require a previous ISB
to be synchronised (SS stands for Self-Synchronising).

Use the ARM64_HAS_ECV capability to control alternative sequences
that switch to these low(er)-cost primitives. Note that the
counter access in the VDSO is for now left alone until we decide
whether we want to allow this.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm64/include/asm/arch_timer.h | 32 +++++++++++++++++++++--------
 arch/arm64/include/asm/sysreg.h     |  3 +++
 2 files changed, 27 insertions(+), 8 deletions(-)

diff --git a/arch/arm64/include/asm/arch_timer.h b/arch/arm64/include/asm/arch_timer.h
index 519ac1f7f859..33a08fff0f06 100644
--- a/arch/arm64/include/asm/arch_timer.h
+++ b/arch/arm64/include/asm/arch_timer.h
@@ -64,14 +64,26 @@ DECLARE_PER_CPU(const struct arch_timer_erratum_workaround *,
 
 static inline notrace u64 arch_timer_read_cntpct_el0(void)
 {
-	isb();
-	return read_sysreg(cntpct_el0);
+	u64 cnt;
+
+	asm volatile(ALTERNATIVE("isb\n mrs %x0, cntpct_el0",
+				 "nop\n" __mrs_s("%x0", SYS_CNTPCTSS_EL0),
+				 ARM64_HAS_ECV)
+		     : "=r" (cnt));
+
+	return cnt;
 }
 
 static inline notrace u64 arch_timer_read_cntvct_el0(void)
 {
-	isb();
-	return read_sysreg(cntvct_el0);
+	u64 cnt;
+
+	asm volatile(ALTERNATIVE("isb\n mrs %x0, cntvct_el0",
+				 "nop\n" __mrs_s("%x0", SYS_CNTVCTSS_EL0),
+				 ARM64_HAS_ECV)
+		     : "=r" (cnt));
+
+	return cnt;
 }
 
 #define arch_timer_reg_read_stable(reg)					\
@@ -174,8 +186,10 @@ static __always_inline u64 __arch_counter_get_cntpct(void)
 {
 	u64 cnt;
 
-	isb();
-	cnt = read_sysreg(cntpct_el0);
+	asm volatile(ALTERNATIVE("isb\n mrs %x0, cntpct_el0",
+				 "nop\n" __mrs_s("%x0", SYS_CNTPCTSS_EL0),
+				 ARM64_HAS_ECV)
+		     : "=r" (cnt));
 	arch_counter_enforce_ordering(cnt);
 	return cnt;
 }
@@ -193,8 +207,10 @@ static __always_inline u64 __arch_counter_get_cntvct(void)
 {
 	u64 cnt;
 
-	isb();
-	cnt = read_sysreg(cntvct_el0);
+	asm volatile(ALTERNATIVE("isb\n mrs %x0, cntvct_el0",
+				 "nop\n" __mrs_s("%x0", SYS_CNTVCTSS_EL0),
+				 ARM64_HAS_ECV)
+		     : "=r" (cnt));
 	arch_counter_enforce_ordering(cnt);
 	return cnt;
 }
diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h
index b268082d67ed..5ce70c034d37 100644
--- a/arch/arm64/include/asm/sysreg.h
+++ b/arch/arm64/include/asm/sysreg.h
@@ -507,6 +507,9 @@
 
 #define SYS_CNTFRQ_EL0			sys_reg(3, 3, 14, 0, 0)
 
+#define SYS_CNTPCTSS_EL0		sys_reg(3, 3, 14, 0, 5)
+#define SYS_CNTVCTSS_EL0		sys_reg(3, 3, 14, 0, 6)
+
 #define SYS_CNTP_TVAL_EL0		sys_reg(3, 3, 14, 2, 0)
 #define SYS_CNTP_CTL_EL0		sys_reg(3, 3, 14, 2, 1)
 #define SYS_CNTP_CVAL_EL0		sys_reg(3, 3, 14, 2, 2)
-- 
2.30.2

[PATCH v2 16/16] arm64: Add handling of CNTVCTSS traps

[Marc Zyngier]

Since CNTVCTSS obey the same control bits as CNTVCT, add the necessary
decoding to the hook table. Note that there is no known user of
this at the moment.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm64/include/asm/esr.h |  6 ++++++
 arch/arm64/kernel/traps.c    | 11 +++++++++++
 2 files changed, 17 insertions(+)

diff --git a/arch/arm64/include/asm/esr.h b/arch/arm64/include/asm/esr.h
index 29f97eb3dad4..a305ce256090 100644
--- a/arch/arm64/include/asm/esr.h
+++ b/arch/arm64/include/asm/esr.h
@@ -227,6 +227,9 @@
 #define ESR_ELx_SYS64_ISS_SYS_CNTVCT	(ESR_ELx_SYS64_ISS_SYS_VAL(3, 3, 2, 14, 0) | \
 					 ESR_ELx_SYS64_ISS_DIR_READ)
 
+#define ESR_ELx_SYS64_ISS_SYS_CNTVCTSS	(ESR_ELx_SYS64_ISS_SYS_VAL(3, 3, 6, 14, 0) | \
+					 ESR_ELx_SYS64_ISS_DIR_READ)
+
 #define ESR_ELx_SYS64_ISS_SYS_CNTFRQ	(ESR_ELx_SYS64_ISS_SYS_VAL(3, 3, 0, 14, 0) | \
 					 ESR_ELx_SYS64_ISS_DIR_READ)
 
@@ -317,6 +320,9 @@
 #define ESR_ELx_CP15_64_ISS_SYS_CNTVCT	(ESR_ELx_CP15_64_ISS_SYS_VAL(1, 14) | \
 					 ESR_ELx_CP15_64_ISS_DIR_READ)
 
+#define ESR_ELx_CP15_64_ISS_SYS_CNTVCTSS (ESR_ELx_CP15_64_ISS_SYS_VAL(9, 14) | \
+					 ESR_ELx_CP15_64_ISS_DIR_READ)
+
 #define ESR_ELx_CP15_32_ISS_SYS_CNTFRQ	(ESR_ELx_CP15_32_ISS_SYS_VAL(0, 0, 14, 0) |\
 					 ESR_ELx_CP15_32_ISS_DIR_READ)
 
diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index b03e383d944a..16710ca55fbb 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c
@@ -653,6 +653,12 @@ static const struct sys64_hook sys64_hooks[] = {
 		.esr_val = ESR_ELx_SYS64_ISS_SYS_CNTVCT,
 		.handler = cntvct_read_handler,
 	},
+	{
+		/* Trap read access to CNTVCTSS_EL0 */
+		.esr_mask = ESR_ELx_SYS64_ISS_SYS_OP_MASK,
+		.esr_val = ESR_ELx_SYS64_ISS_SYS_CNTVCTSS,
+		.handler = cntvct_read_handler,
+	},
 	{
 		/* Trap read access to CNTFRQ_EL0 */
 		.esr_mask = ESR_ELx_SYS64_ISS_SYS_OP_MASK,
@@ -729,6 +735,11 @@ static const struct sys64_hook cp15_64_hooks[] = {
 		.esr_val = ESR_ELx_CP15_64_ISS_SYS_CNTVCT,
 		.handler = compat_cntvct_read_handler,
 	},
+	{
+		.esr_mask = ESR_ELx_CP15_64_ISS_SYS_MASK,
+		.esr_val = ESR_ELx_CP15_64_ISS_SYS_CNTVCTSS,
+		.handler = compat_cntvct_read_handler,
+	},
 	{},
 };
 
-- 
2.30.2

Re: [PATCH v2 14/16] arm64: Add a capability for FEAT_ECV

[Will Deacon]

On Wed, Sep 22, 2021 at 10:19:39PM +0100, Marc Zyngier wrote:
> Add a new capability to detect the Enhanced Counter Virtualization
> feature (FEAT_ECV).
> 
> Reviewed-by: Oliver Upton <oupton@google.com>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> ---
>  arch/arm64/kernel/cpufeature.c | 10 ++++++++++
>  arch/arm64/tools/cpucaps       |  1 +
>  2 files changed, 11 insertions(+)
> 
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index f8a3067d10c6..26b11ce8fff6 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1926,6 +1926,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
>  		.sign = FTR_UNSIGNED,
>  		.min_field_value = 1,
>  	},
> +	{
> +		.desc = "Enhanced Counter Virtualization",
> +		.capability = ARM64_HAS_ECV,
> +		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
> +		.matches = has_cpuid_feature,
> +		.sys_reg = SYS_ID_AA64MMFR0_EL1,
> +		.field_pos = ID_AA64MMFR0_ECV_SHIFT,
> +		.sign = FTR_UNSIGNED,
> +		.min_field_value = 1,
> +	},

Could we add a HWCAP for this and change the field to FTR_VISIBLE, please? I
know most users of the counter are indirected via the vDSO, but there are
some users out there using the counter directly and it would save them
having to probe via SIGILL if there was a hwcap available.

Cheers,

Will

Re: [PATCH v2 14/16] arm64: Add a capability for FEAT_ECV

[Marc Zyngier]

On Wed, 29 Sep 2021 17:03:30 +0100,
Will Deacon <will@kernel.org> wrote:
> 
> On Wed, Sep 22, 2021 at 10:19:39PM +0100, Marc Zyngier wrote:
> > Add a new capability to detect the Enhanced Counter Virtualization
> > feature (FEAT_ECV).
> > 
> > Reviewed-by: Oliver Upton <oupton@google.com>
> > Signed-off-by: Marc Zyngier <maz@kernel.org>
> > ---
> >  arch/arm64/kernel/cpufeature.c | 10 ++++++++++
> >  arch/arm64/tools/cpucaps       |  1 +
> >  2 files changed, 11 insertions(+)
> > 
> > diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> > index f8a3067d10c6..26b11ce8fff6 100644
> > --- a/arch/arm64/kernel/cpufeature.c
> > +++ b/arch/arm64/kernel/cpufeature.c
> > @@ -1926,6 +1926,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
> >  		.sign = FTR_UNSIGNED,
> >  		.min_field_value = 1,
> >  	},
> > +	{
> > +		.desc = "Enhanced Counter Virtualization",
> > +		.capability = ARM64_HAS_ECV,
> > +		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
> > +		.matches = has_cpuid_feature,
> > +		.sys_reg = SYS_ID_AA64MMFR0_EL1,
> > +		.field_pos = ID_AA64MMFR0_ECV_SHIFT,
> > +		.sign = FTR_UNSIGNED,
> > +		.min_field_value = 1,
> > +	},
> 
> Could we add a HWCAP for this and change the field to FTR_VISIBLE, please? I
> know most users of the counter are indirected via the vDSO, but there are
> some users out there using the counter directly and it would save them
> having to probe via SIGILL if there was a hwcap available.

Fair enough, I'll add that.

The problem of the vdso remains though, and is by far the most common
user of the feature. Any idea on how to handle it? Patching the vdso
is ugly, and I'd rather avoid it.

I briefly looked at using ifunc, but it is likely that the indirection
would add an extra cost. Are we OK with that?

	M.

-- 
Without deviation from the norm, progress is not possible.

Re: [PATCH v2 14/16] arm64: Add a capability for FEAT_ECV

[Will Deacon]

On Thu, Sep 30, 2021 at 08:42:56AM +0100, Marc Zyngier wrote:
> On Wed, 29 Sep 2021 17:03:30 +0100,
> Will Deacon <will@kernel.org> wrote:
> > 
> > On Wed, Sep 22, 2021 at 10:19:39PM +0100, Marc Zyngier wrote:
> > > Add a new capability to detect the Enhanced Counter Virtualization
> > > feature (FEAT_ECV).
> > > 
> > > Reviewed-by: Oliver Upton <oupton@google.com>
> > > Signed-off-by: Marc Zyngier <maz@kernel.org>
> > > ---
> > >  arch/arm64/kernel/cpufeature.c | 10 ++++++++++
> > >  arch/arm64/tools/cpucaps       |  1 +
> > >  2 files changed, 11 insertions(+)
> > > 
> > > diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> > > index f8a3067d10c6..26b11ce8fff6 100644
> > > --- a/arch/arm64/kernel/cpufeature.c
> > > +++ b/arch/arm64/kernel/cpufeature.c
> > > @@ -1926,6 +1926,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
> > >  		.sign = FTR_UNSIGNED,
> > >  		.min_field_value = 1,
> > >  	},
> > > +	{
> > > +		.desc = "Enhanced Counter Virtualization",
> > > +		.capability = ARM64_HAS_ECV,
> > > +		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
> > > +		.matches = has_cpuid_feature,
> > > +		.sys_reg = SYS_ID_AA64MMFR0_EL1,
> > > +		.field_pos = ID_AA64MMFR0_ECV_SHIFT,
> > > +		.sign = FTR_UNSIGNED,
> > > +		.min_field_value = 1,
> > > +	},
> > 
> > Could we add a HWCAP for this and change the field to FTR_VISIBLE, please? I
> > know most users of the counter are indirected via the vDSO, but there are
> > some users out there using the counter directly and it would save them
> > having to probe via SIGILL if there was a hwcap available.
> 
> Fair enough, I'll add that.

Thanks!

> The problem of the vdso remains though, and is by far the most common
> user of the feature. Any idea on how to handle it? Patching the vdso
> is ugly, and I'd rather avoid it.
> 
> I briefly looked at using ifunc, but it is likely that the indirection
> would add an extra cost. Are we OK with that?

The vDSO is still miles faster than a system call, so I'd be inclined to
leave it as-is for the time being. I suspect that use-cases which can't
stomach the cost of the ISB can't stomach the cost of the vDSO at all.

Will