All of lore.kernel.org
 help / color / mirror / Atom feed
From: Christoffer Dall <christoffer.dall@linaro.org>
To: kvmarm@lists.cs.columbia.edu, linux-arm-kernel@lists.infradead.org
Cc: Christoffer Dall <cdall@linaro.org>,
	kvm@vger.kernel.org, Marc Zyngier <marc.zyngier@arm.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will.deacon@arm.com>,
	Shih-Wei Li <shihwei@cs.columbia.edu>
Subject: [PATCH v4 03/20] arm64: Use physical counter for in-kernel reads when booted in EL2
Date: Fri, 20 Oct 2017 13:49:22 +0200	[thread overview]
Message-ID: <20171020114939.12554-4-christoffer.dall@linaro.org> (raw)
In-Reply-To: <20171020114939.12554-1-christoffer.dall@linaro.org>

From: Christoffer Dall <cdall@linaro.org>

Using the physical counter allows KVM to retain the offset between the
virtual and physical counter as long as it is actively running a VCPU.

As soon as a VCPU is released, another thread is scheduled or we start
running userspace applications, we reset the offset to 0, so that
userspace accessing the virtual timer can still read the virtual counter
and get the same view of time as the kernel.

This opens up potential improvements for KVM performance, but we have to
make a few adjustments to preserve system consistency.

Currently get_cycles() is hardwired to arch_counter_get_cntvct() on
arm64, but as we move to using the physical timer for the in-kernel
time-keeping on systems that boot in EL2, we should use the same counter
for get_cycles() as for other in-kernel timekeeping operations.

Similarly, implementations of arch_timer_set_next_event_phys() is
modified to use the counter specific to the timer being programmed.

VHE kernels or kernels continuing to use the virtual timer are
unaffected.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
---

Notes:
    Changes since v3:
     - Consolidated the change of the arch_timer_read_counter function
       pointer and the change to get_cycles().  A new patch was added prior
       to this one to support stable reads of the physical counter including
       errata workaround.
    
     - Added !is_hyp_mode_available() clause to ARM64 systems when selecting
       the in-kernel counter, so that arm64 kernels not booting in EL2 remain
       unmodified.
    
     - Added EXPORT_SYMBOL_GPL(arch_timer_read_counter) to support building
       modules that reference get_cycles() and thereby
       arch_timer_read_counter.

 arch/arm64/include/asm/timex.h       |  2 +-
 drivers/clocksource/arm_arch_timer.c | 12 ++++++++----
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/include/asm/timex.h b/arch/arm64/include/asm/timex.h
index 81a076eb37fa..9ad60bae5c8d 100644
--- a/arch/arm64/include/asm/timex.h
+++ b/arch/arm64/include/asm/timex.h
@@ -22,7 +22,7 @@
  * Use the current timer as a cycle counter since this is what we use for
  * the delay loop.
  */
-#define get_cycles()	arch_counter_get_cntvct()
+#define get_cycles()	arch_timer_read_counter()
 
 #include <asm-generic/timex.h>
 
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index ff8f8a177156..061476e92db7 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -158,6 +158,7 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
  * if we don't have the cp15 accessors we won't have a problem.
  */
 u64 (*arch_timer_read_counter)(void) = arch_counter_get_cntvct;
+EXPORT_SYMBOL_GPL(arch_timer_read_counter);
 
 static u64 arch_counter_read(struct clocksource *cs)
 {
@@ -329,16 +330,19 @@ static void erratum_set_next_event_tval_generic(const int access, unsigned long
 						struct clock_event_device *clk)
 {
 	unsigned long ctrl;
-	u64 cval = evt + arch_counter_get_cntvct();
+	u64 cval;
 
 	ctrl = arch_timer_reg_read(access, ARCH_TIMER_REG_CTRL, clk);
 	ctrl |= ARCH_TIMER_CTRL_ENABLE;
 	ctrl &= ~ARCH_TIMER_CTRL_IT_MASK;
 
-	if (access == ARCH_TIMER_PHYS_ACCESS)
+	if (access == ARCH_TIMER_PHYS_ACCESS) {
+		cval = evt + arch_counter_get_cntpct();
 		write_sysreg(cval, cntp_cval_el0);
-	else
+	} else {
+		cval = evt + arch_counter_get_cntvct();
 		write_sysreg(cval, cntv_cval_el0);
+	}
 
 	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
 }
@@ -913,7 +917,7 @@ static void __init arch_counter_register(unsigned type)
 
 	/* Register the CP15 based counter if we have one */
 	if (type & ARCH_TIMER_TYPE_CP15) {
-		if (IS_ENABLED(CONFIG_ARM64) ||
+		if ((IS_ENABLED(CONFIG_ARM64) && !is_hyp_mode_available()) ||
 		    arch_timer_uses_ppi == ARCH_TIMER_VIRT_PPI)
 			arch_timer_read_counter = arch_counter_get_cntvct;
 		else
-- 
2.14.2

WARNING: multiple messages have this Message-ID (diff)
From: christoffer.dall@linaro.org (Christoffer Dall)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH v4 03/20] arm64: Use physical counter for in-kernel reads when booted in EL2
Date: Fri, 20 Oct 2017 13:49:22 +0200	[thread overview]
Message-ID: <20171020114939.12554-4-christoffer.dall@linaro.org> (raw)
In-Reply-To: <20171020114939.12554-1-christoffer.dall@linaro.org>

From: Christoffer Dall <cdall@linaro.org>

Using the physical counter allows KVM to retain the offset between the
virtual and physical counter as long as it is actively running a VCPU.

As soon as a VCPU is released, another thread is scheduled or we start
running userspace applications, we reset the offset to 0, so that
userspace accessing the virtual timer can still read the virtual counter
and get the same view of time as the kernel.

This opens up potential improvements for KVM performance, but we have to
make a few adjustments to preserve system consistency.

Currently get_cycles() is hardwired to arch_counter_get_cntvct() on
arm64, but as we move to using the physical timer for the in-kernel
time-keeping on systems that boot in EL2, we should use the same counter
for get_cycles() as for other in-kernel timekeeping operations.

Similarly, implementations of arch_timer_set_next_event_phys() is
modified to use the counter specific to the timer being programmed.

VHE kernels or kernels continuing to use the virtual timer are
unaffected.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@linaro.org>
---

Notes:
    Changes since v3:
     - Consolidated the change of the arch_timer_read_counter function
       pointer and the change to get_cycles().  A new patch was added prior
       to this one to support stable reads of the physical counter including
       errata workaround.
    
     - Added !is_hyp_mode_available() clause to ARM64 systems when selecting
       the in-kernel counter, so that arm64 kernels not booting in EL2 remain
       unmodified.
    
     - Added EXPORT_SYMBOL_GPL(arch_timer_read_counter) to support building
       modules that reference get_cycles() and thereby
       arch_timer_read_counter.

 arch/arm64/include/asm/timex.h       |  2 +-
 drivers/clocksource/arm_arch_timer.c | 12 ++++++++----
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/include/asm/timex.h b/arch/arm64/include/asm/timex.h
index 81a076eb37fa..9ad60bae5c8d 100644
--- a/arch/arm64/include/asm/timex.h
+++ b/arch/arm64/include/asm/timex.h
@@ -22,7 +22,7 @@
  * Use the current timer as a cycle counter since this is what we use for
  * the delay loop.
  */
-#define get_cycles()	arch_counter_get_cntvct()
+#define get_cycles()	arch_timer_read_counter()
 
 #include <asm-generic/timex.h>
 
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index ff8f8a177156..061476e92db7 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -158,6 +158,7 @@ u32 arch_timer_reg_read(int access, enum arch_timer_reg reg,
  * if we don't have the cp15 accessors we won't have a problem.
  */
 u64 (*arch_timer_read_counter)(void) = arch_counter_get_cntvct;
+EXPORT_SYMBOL_GPL(arch_timer_read_counter);
 
 static u64 arch_counter_read(struct clocksource *cs)
 {
@@ -329,16 +330,19 @@ static void erratum_set_next_event_tval_generic(const int access, unsigned long
 						struct clock_event_device *clk)
 {
 	unsigned long ctrl;
-	u64 cval = evt + arch_counter_get_cntvct();
+	u64 cval;
 
 	ctrl = arch_timer_reg_read(access, ARCH_TIMER_REG_CTRL, clk);
 	ctrl |= ARCH_TIMER_CTRL_ENABLE;
 	ctrl &= ~ARCH_TIMER_CTRL_IT_MASK;
 
-	if (access == ARCH_TIMER_PHYS_ACCESS)
+	if (access == ARCH_TIMER_PHYS_ACCESS) {
+		cval = evt + arch_counter_get_cntpct();
 		write_sysreg(cval, cntp_cval_el0);
-	else
+	} else {
+		cval = evt + arch_counter_get_cntvct();
 		write_sysreg(cval, cntv_cval_el0);
+	}
 
 	arch_timer_reg_write(access, ARCH_TIMER_REG_CTRL, ctrl, clk);
 }
@@ -913,7 +917,7 @@ static void __init arch_counter_register(unsigned type)
 
 	/* Register the CP15 based counter if we have one */
 	if (type & ARCH_TIMER_TYPE_CP15) {
-		if (IS_ENABLED(CONFIG_ARM64) ||
+		if ((IS_ENABLED(CONFIG_ARM64) && !is_hyp_mode_available()) ||
 		    arch_timer_uses_ppi == ARCH_TIMER_VIRT_PPI)
 			arch_timer_read_counter = arch_counter_get_cntvct;
 		else
-- 
2.14.2

  parent reply	other threads:[~2017-10-20 11:49 UTC|newest]

Thread overview: 74+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-10-20 11:49 [PATCH v4 00/20] KVM: arm/arm64: Optimize arch timer register handling Christoffer Dall
2017-10-20 11:49 ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 01/20] irqchip/gic: Deal with broken firmware exposing only 4kB of GICv2 CPU interface Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 02/20] arm64: Implement arch_counter_get_cntpct to read the physical counter Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 13:41   ` Marc Zyngier
2017-10-25 13:41     ` Marc Zyngier
2017-10-20 11:49 ` Christoffer Dall [this message]
2017-10-20 11:49   ` [PATCH v4 03/20] arm64: Use physical counter for in-kernel reads when booted in EL2 Christoffer Dall
2017-10-25 13:43   ` Marc Zyngier
2017-10-25 13:43     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 04/20] KVM: arm/arm64: Guard kvm_vgic_map_is_active against !vgic_initialized Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 05/20] KVM: arm/arm64: Support calling vgic_update_irq_pending from irq context Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 13:45   ` Marc Zyngier
2017-10-25 13:45     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 06/20] KVM: arm/arm64: Check that system supports split eoi/deactivate Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 13:47   ` Marc Zyngier
2017-10-25 13:47     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 07/20] KVM: arm/arm64: Make timer_arm and timer_disarm helpers more generic Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 13:49   ` Marc Zyngier
2017-10-25 13:49     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 08/20] KVM: arm/arm64: Rename soft timer to bg_timer Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 09/20] KVM: arm/arm64: Move timer/vgic flush/sync under disabled irq Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 10/20] KVM: arm/arm64: Use separate timer for phys timer emulation Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 13:59   ` Marc Zyngier
2017-10-25 13:59     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 11/20] KVM: arm/arm64: Move timer save/restore out of the hyp code Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 12/20] genirq: Document vcpu_info usage for percpu_devid interrupts Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:56   ` Thomas Gleixner
2017-10-20 11:56     ` Thomas Gleixner
2017-10-21 14:29     ` Christoffer Dall
2017-10-21 14:29       ` Christoffer Dall
2017-10-21 14:55       ` Thomas Gleixner
2017-10-21 14:55         ` Thomas Gleixner
2017-10-21 15:30         ` Christoffer Dall
2017-10-21 15:30           ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 13/20] KVM: arm/arm64: Set VCPU affinity for virt timer irq Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 14/20] KVM: arm/arm64: Avoid timer save/restore in vcpu entry/exit Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 14:36   ` Marc Zyngier
2017-10-25 14:36     ` Marc Zyngier
2017-11-16 20:30   ` Jintack Lim
2017-11-16 20:30     ` Jintack Lim
2017-11-16 20:32     ` Jintack Lim
2017-11-16 20:32       ` Jintack Lim
2017-11-20 11:15     ` Christoffer Dall
2017-11-20 11:15       ` Christoffer Dall
2017-11-20 16:32       ` Jintack Lim
2017-11-20 16:32         ` Jintack Lim
2017-10-20 11:49 ` [PATCH v4 15/20] KVM: arm/arm64: Support EL1 phys timer register access in set/get reg Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-25 14:38   ` Marc Zyngier
2017-10-25 14:38     ` Marc Zyngier
2017-10-20 11:49 ` [PATCH v4 16/20] KVM: arm/arm64: Use kvm_arm_timer_set/get_reg for guest register traps Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 17/20] KVM: arm/arm64: Move phys_timer_emulate function Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 18/20] KVM: arm/arm64: Avoid phys timer emulation in vcpu entry/exit Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 19/20] KVM: arm/arm64: Get rid of kvm_timer_flush_hwstate Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall
2017-10-20 11:49 ` [PATCH v4 20/20] KVM: arm/arm64: Rework kvm_timer_should_fire Christoffer Dall
2017-10-20 11:49   ` Christoffer Dall

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20171020114939.12554-4-christoffer.dall@linaro.org \
    --to=christoffer.dall@linaro.org \
    --cc=catalin.marinas@arm.com \
    --cc=cdall@linaro.org \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=marc.zyngier@arm.com \
    --cc=shihwei@cs.columbia.edu \
    --cc=will.deacon@arm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.