[PATCH v3 0/2] irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0

[PATCH v3 0/2] irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0

[Alexandru Elisei]

Trusted Firmware-A's default interrupt routing model is to clear
SCR_EL3.FIQ, which is the only case that GICv3 doesn't support. This series
tries to fix that by detecting it at runtime and using a different priority
value for ICC_PMR_EL1 when masking regular interrupts. As a result, we will
be able to support pseudo-NMIs on all combinations of GIC security states
and firmware configurations.

The first patch was there because when I started working on the PMU NMI
patches I found it confusing that there was no clear message stating that
NMIs were successfully enabled. The second patch is the main patch of the
series, where pseudo-NMIs are enabled even if SCR_EL3.FIQ == 0.

As an aside, the set of priorities that I've added would work in all cases
if there was no need to use GIC_PRIO_PSR_I_SET, but that is a much more
intrusive change and I'm not comfortable attempting it. I'm pretty sure I
will end up breaking things really badly.

I've tested the series using PMU NMIs [1], rebased on top of v5.9-rc4, on
the model and on espressobin-v7. Tests that I've run:

1. On the model:
- Host with SCR_EL3.FIQ == 1 (so using the original priorities), ran
  perf record -a -- iperf3 -c 127.0.0.1 -t 30.
- On a KVM guest (security disabled, so using the original priorities),
  ran the same command.

2. On an espressobin-v7:
- Host with SCR_EL3.FIQ == 0 (using the priority added by the series),
  ran perf record -a -- iperf3 -c 127.0.0.1 -t 60.
- On a KVM guest (security disabled, so using the original priorities),
  ran the same command.
- Stress test for one hour with CONFIG_ARM64_DEBUG_PRIORITY_MASKING set
  for the host and guest. On the host, I ran in parallel
  perf record -a -- iperf3 -c 127.0.0.1 -t 3600  and
  perf record -ae L1-dcache-loads -a -- sleep 3600. On the guest, I ran
  the same iperf3 command as on the host.

Changes since v2:
* Renamed gic_prio_irqoff() -> GIC_PRIO_IRQOFF, thus avoiding any changes
  to daifflags.h and irqflags.h, as per Marc's excellent suggestion.
* Added comment gic_cpu_sys_reg_init() explaining why we cannot get the
  warning on the boot CPU.

Changes since v1:
* Rebased on top of v5.9-rc1
* Changed pmr to u64 in arch_local_irq_{enable,disable} to stop
  clang from complaining and to match local_daif_restore().

[1] https://lkml.org/lkml/2020/8/19/671

Alexandru Elisei (2):
  irqchip/gicv3: Spell out when pseudo-NMIs are enabled
  irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0

 arch/arm64/include/asm/arch_gicv3.h |  8 +++-
 arch/arm64/include/asm/ptrace.h     | 14 ++++++-
 arch/arm64/kernel/image-vars.h      |  2 +
 drivers/irqchip/irq-gic-v3.c        | 60 ++++++++++++++++++++++-------
 4 files changed, 68 insertions(+), 16 deletions(-)

-- 
2.28.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v3 1/2] irqchip/gicv3: Spell out when pseudo-NMIs are enabled

[Alexandru Elisei]

When NMIs cannot be enabled, the driver prints a message stating that
unambiguously. When they are enabled, the only feedback we get is a message
regarding the use of synchronization for ICC_PMR_EL1 writes, which is not
as useful for a user who is not intimately familiar with how NMIs are
implemented.

Let's make it obvious that pseudo-NMIs are enabled. Keep the message about
using a barrier for ICC_PMR_EL1 writes, because it has a non-negligible
impact on performance.

Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
---
 drivers/irqchip/irq-gic-v3.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index 850842f27bee..aa9b43d412b2 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -1564,8 +1564,8 @@ static void gic_enable_nmi_support(void)
 	if (gic_read_ctlr() & ICC_CTLR_EL1_PMHE_MASK)
 		static_branch_enable(&gic_pmr_sync);
 
-	pr_info("%s ICC_PMR_EL1 synchronisation\n",
-		static_branch_unlikely(&gic_pmr_sync) ? "Forcing" : "Relaxing");
+	pr_info("Pseudo-NMIs enabled using %s ICC_PMR_EL1 synchronisation\n",
+		static_branch_unlikely(&gic_pmr_sync) ? "forced" : "relaxed");
 
 	static_branch_enable(&supports_pseudo_nmis);
 
-- 
2.28.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v3 2/2] irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0

[Alexandru Elisei]

The GIC's internal view of the priority mask register and the assigned
interrupt priorities are based on whether GIC security is enabled and
whether firmware routes Group 0 interrupts to EL3. At the moment, we
support priority masking when ICC_PMR_EL1 and interrupt priorities are
either both modified by the GIC, or both left unchanged.

Trusted Firmware-A's default interrupt routing model allows Group 0
interrupts to be delivered to the non-secure world (SCR_EL3.FIQ == 0).
Unfortunately, this is precisely the case that the GIC driver doesn't
support: ICC_PMR_EL1 remains unchanged, but the GIC's view of interrupt
priorities is different from the software programmed values.

Support pseudo-NMIs when SCR_EL3.FIQ == 0 by using a different value to
mask regular interrupts. All the other values remain the same.

Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
---
 arch/arm64/include/asm/arch_gicv3.h |  8 ++++-
 arch/arm64/include/asm/ptrace.h     | 14 +++++++-
 arch/arm64/kernel/image-vars.h      |  2 ++
 drivers/irqchip/irq-gic-v3.c        | 56 ++++++++++++++++++++++-------
 4 files changed, 66 insertions(+), 14 deletions(-)

diff --git a/arch/arm64/include/asm/arch_gicv3.h b/arch/arm64/include/asm/arch_gicv3.h
index 6647ae4f0231..880b9054d75c 100644
--- a/arch/arm64/include/asm/arch_gicv3.h
+++ b/arch/arm64/include/asm/arch_gicv3.h
@@ -153,7 +153,7 @@ static inline bool gic_prio_masking_enabled(void)
 
 static inline void gic_pmr_mask_irqs(void)
 {
-	BUILD_BUG_ON(GICD_INT_DEF_PRI < (GIC_PRIO_IRQOFF |
+	BUILD_BUG_ON(GICD_INT_DEF_PRI < (__GIC_PRIO_IRQOFF |
 					 GIC_PRIO_PSR_I_SET));
 	BUILD_BUG_ON(GICD_INT_DEF_PRI >= GIC_PRIO_IRQON);
 	/*
@@ -162,6 +162,12 @@ static inline void gic_pmr_mask_irqs(void)
 	 * are applied to IRQ priorities
 	 */
 	BUILD_BUG_ON((0x80 | (GICD_INT_DEF_PRI >> 1)) >= GIC_PRIO_IRQON);
+	/*
+	 * Same situation as above, but now we make sure that we can mask
+	 * regular interrupts.
+	 */
+	BUILD_BUG_ON((0x80 | (GICD_INT_DEF_PRI >> 1)) < (__GIC_PRIO_IRQOFF_NS |
+							 GIC_PRIO_PSR_I_SET));
 	gic_write_pmr(GIC_PRIO_IRQOFF);
 }
 
diff --git a/arch/arm64/include/asm/ptrace.h b/arch/arm64/include/asm/ptrace.h
index 966ed30ed5f7..997cf8c8cd52 100644
--- a/arch/arm64/include/asm/ptrace.h
+++ b/arch/arm64/include/asm/ptrace.h
@@ -31,9 +31,21 @@
  * interrupt disabling temporarily does not rely on IRQ priorities.
  */
 #define GIC_PRIO_IRQON			0xe0
-#define GIC_PRIO_IRQOFF			(GIC_PRIO_IRQON & ~0x80)
+#define __GIC_PRIO_IRQOFF		(GIC_PRIO_IRQON & ~0x80)
+#define __GIC_PRIO_IRQOFF_NS		0xa0
 #define GIC_PRIO_PSR_I_SET		(1 << 4)
 
+#define GIC_PRIO_IRQOFF							\
+	({								\
+		extern struct static_key_false gic_nonsecure_priorities;\
+		u8 __prio = __GIC_PRIO_IRQOFF;				\
+									\
+		if (static_branch_unlikely(&gic_nonsecure_priorities))	\
+			__prio = __GIC_PRIO_IRQOFF_NS;			\
+									\
+		__prio;							\
+	})
+
 /* Additional SPSR bits not exposed in the UABI */
 #define PSR_MODE_THREAD_BIT	(1 << 0)
 #define PSR_IL_BIT		(1 << 20)
diff --git a/arch/arm64/kernel/image-vars.h b/arch/arm64/kernel/image-vars.h
index 8982b68289b7..98a5215c1598 100644
--- a/arch/arm64/kernel/image-vars.h
+++ b/arch/arm64/kernel/image-vars.h
@@ -101,6 +101,8 @@ KVM_NVHE_ALIAS(vgic_v3_cpuif_trap);
 /* Static key checked in pmr_sync(). */
 #ifdef CONFIG_ARM64_PSEUDO_NMI
 KVM_NVHE_ALIAS(gic_pmr_sync);
+/* Static key checked in GIC_PRIO_IRQOFF. */
+KVM_NVHE_ALIAS(gic_nonsecure_priorities);
 #endif
 
 /* EL2 exception handling */
diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index aa9b43d412b2..bb55d89d7d49 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -75,16 +75,14 @@ static DEFINE_STATIC_KEY_TRUE(supports_deactivate_key);
  *
  * If SCR_EL3.FIQ == 1, the values writen to/read from PMR and RPR at non-secure
  * EL1 are subject to a similar operation thus matching the priorities presented
- * from the (re)distributor when security is enabled.
+ * from the (re)distributor when security is enabled. When SCR_EL3.FIQ == 0,
+ * these values are unchanched by the GIC.
  *
  * see GICv3/GICv4 Architecture Specification (IHI0069D):
  * - section 4.8.1 Non-secure accesses to register fields for Secure interrupt
  *   priorities.
  * - Figure 4-7 Secure read of the priority field for a Non-secure Group 1
  *   interrupt.
- *
- * For now, we only support pseudo-NMIs if we have non-secure view of
- * priorities.
  */
 static DEFINE_STATIC_KEY_FALSE(supports_pseudo_nmis);
 
@@ -97,6 +95,9 @@ static DEFINE_STATIC_KEY_FALSE(supports_pseudo_nmis);
 DEFINE_STATIC_KEY_FALSE(gic_pmr_sync);
 EXPORT_SYMBOL(gic_pmr_sync);
 
+DEFINE_STATIC_KEY_FALSE(gic_nonsecure_priorities);
+EXPORT_SYMBOL(gic_nonsecure_priorities);
+
 /* ppi_nmi_refs[n] == number of cpus having ppi[n + 16] set as NMI */
 static refcount_t *ppi_nmi_refs;
 
@@ -932,14 +933,20 @@ static void gic_cpu_sys_reg_init(void)
 	/* Set priority mask register */
 	if (!gic_prio_masking_enabled()) {
 		write_gicreg(DEFAULT_PMR_VALUE, ICC_PMR_EL1);
-	} else {
+	} else if (gic_supports_nmi()) {
 		/*
 		 * Mismatch configuration with boot CPU, the system is likely
 		 * to die as interrupt masking will not work properly on all
 		 * CPUs
+		 *
+		 * The boot CPU calls this function before enabling NMI support,
+		 * and as a result we'll never see this warning in the boot path
+		 * for that CPU.
 		 */
-		WARN_ON(gic_supports_nmi() && group0 &&
-			!gic_dist_security_disabled());
+		if (static_branch_unlikely(&gic_nonsecure_priorities))
+			WARN_ON(!group0 || gic_dist_security_disabled());
+		else
+			WARN_ON(group0 && !gic_dist_security_disabled());
 	}
 
 	/*
@@ -1544,11 +1551,6 @@ static void gic_enable_nmi_support(void)
 	if (!gic_prio_masking_enabled())
 		return;
 
-	if (gic_has_group0() && !gic_dist_security_disabled()) {
-		pr_warn("SCR_EL3.FIQ is cleared, cannot enable use of pseudo-NMIs\n");
-		return;
-	}
-
 	ppi_nmi_refs = kcalloc(gic_data.ppi_nr, sizeof(*ppi_nmi_refs), GFP_KERNEL);
 	if (!ppi_nmi_refs)
 		return;
@@ -1567,6 +1569,36 @@ static void gic_enable_nmi_support(void)
 	pr_info("Pseudo-NMIs enabled using %s ICC_PMR_EL1 synchronisation\n",
 		static_branch_unlikely(&gic_pmr_sync) ? "forced" : "relaxed");
 
+	/*
+	 * How priority values are used by the GIC depends on two things:
+	 * the security state of the GIC (controlled by the GICD_CTRL.DS bit)
+	 * and if Group 0 interrupts can be delivered to Linux in the non-secure
+	 * world as FIQs (controlled by the SCR_EL3.FIQ bit). These affect the
+	 * the ICC_PMR_EL1 register and the priority that software assigns to
+	 * interrupts:
+	 *
+	 * GICD_CTRL.DS | SCR_EL3.FIQ | ICC_PMR_EL1 | Group 1 priority
+	 * -----------------------------------------------------------
+	 *      1       |      -      |  unchanged  |    unchanged
+	 * -----------------------------------------------------------
+	 *      0       |      1      |  non-secure |    non-secure
+	 * -----------------------------------------------------------
+	 *      0       |      0      |  unchanged  |    non-secure
+	 *
+	 * where non-secure means that the value is right-shifted by one and the
+	 * MSB bit set, to make it fit in the non-secure priority range.
+	 *
+	 * In the first two cases, where ICC_PMR_EL1 and the interrupt priority
+	 * are both either modified or unchanged, we can use the same set of
+	 * priorities.
+	 *
+	 * In the last case, where only the interrupt priorities are modified to
+	 * be in the non-secure range, we use a different PMR value to mask IRQs
+	 * and the rest of the values that we use remain unchanged.
+	 */
+	if (gic_has_group0() && !gic_dist_security_disabled())
+		static_branch_enable(&gic_nonsecure_priorities);
+
 	static_branch_enable(&supports_pseudo_nmis);
 
 	if (static_branch_likely(&supports_deactivate_key))
-- 
2.28.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v3 0/2] irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0

[Marc Zyngier]

On Sat, 12 Sep 2020 16:37:05 +0100, Alexandru Elisei wrote:
> Trusted Firmware-A's default interrupt routing model is to clear
> SCR_EL3.FIQ, which is the only case that GICv3 doesn't support. This series
> tries to fix that by detecting it at runtime and using a different priority
> value for ICC_PMR_EL1 when masking regular interrupts. As a result, we will
> be able to support pseudo-NMIs on all combinations of GIC security states
> and firmware configurations.
> 
> [...]

Applied to irq/irqchip-next, thanks!

[1/2] irqchip/gic-v3: Spell out when pseudo-NMIs are enabled
      commit: 4e594ad1068ea1db359d6161f580f03edecf6cb0
[2/2] irqchip/gic-v3: Support pseudo-NMIs when SCR_EL3.FIQ == 0
      commit: 336780590990efa69596884114cad3f517b6333b

Cheers,

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



_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel