@@ -252,5 +252,10 @@ static inline void gic_arch_enable_irqs(void)
WARN_ON_ONCE(true);
}
+static inline bool gic_has_relaxed_pmr_sync(void)
+{
+ return false;
+}
+
#endif /* !__ASSEMBLY__ */
#endif /* !__ASM_ARCH_GICV3_H */
@@ -190,5 +190,10 @@ static inline void gic_arch_enable_irqs(void)
asm volatile ("msr daifclr, #3" : : : "memory");
}
+static inline bool gic_has_relaxed_pmr_sync(void)
+{
+ return cpus_have_cap(ARM64_HAS_GIC_PRIO_RELAXED_SYNC);
+}
+
#endif /* __ASSEMBLY__ */
#endif /* __ASM_ARCH_GICV3_H */
@@ -11,6 +11,8 @@
#include <linux/kasan-checks.h>
+#include <asm/alternative-macros.h>
+
#define __nops(n) ".rept " #n "\nnop\n.endr\n"
#define nops(n) asm volatile(__nops(n))
@@ -41,10 +43,11 @@
#ifdef CONFIG_ARM64_PSEUDO_NMI
#define pmr_sync() \
do { \
- extern struct static_key_false gic_pmr_sync; \
- \
- if (static_branch_unlikely(&gic_pmr_sync)) \
- dsb(sy); \
+ asm volatile( \
+ ALTERNATIVE_CB("dsb sy", \
+ ARM64_HAS_GIC_PRIO_RELAXED_SYNC, \
+ alt_cb_patch_nops) \
+ ); \
} while(0)
#else
#define pmr_sync() do {} while (0)
@@ -2056,6 +2056,34 @@ static bool can_use_gic_priorities(const struct arm64_cpu_capabilities *entry,
return enable_pseudo_nmi;
}
+
+static bool has_gic_prio_relaxed_sync(const struct arm64_cpu_capabilities *entry,
+ int scope)
+{
+ /*
+ * If we're not using priority masking then we won't be poking PMR_EL1,
+ * and there's no need to relax synchronization of writes to it, and
+ * ICC_CTLR_EL1 might not be accessible and we must avoid reads from
+ * that.
+ *
+ * ARM64_HAS_GIC_PRIO_MASKING has a lower index, and is a boot CPU
+ * feature, so will be detected earlier.
+ */
+ BUILD_BUG_ON(ARM64_HAS_GIC_PRIO_RELAXED_SYNC <= ARM64_HAS_GIC_PRIO_MASKING);
+ if (!cpus_have_cap(ARM64_HAS_GIC_PRIO_MASKING))
+ return false;
+
+ /*
+ * When Priority Mask Hint Enable (PMHE) == 0b0, PMR is not used as a
+ * hint for interrupt distribution, a DSB is not necessary when
+ * unmasking IRQs via PMR, and we can relax the barrier to a NOP.
+ *
+ * Linux itself doesn't use 1:N distribution, so has no need to
+ * set PMHE. The only reason to have it set is if EL3 requires it
+ * (and we can't change it).
+ */
+ return (gic_read_ctlr() & ICC_CTLR_EL1_PMHE_MASK) == 0;
+}
#endif
#ifdef CONFIG_ARM64_BTI
@@ -2546,6 +2574,14 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
.type = ARM64_CPUCAP_STRICT_BOOT_CPU_FEATURE,
.matches = can_use_gic_priorities,
},
+ {
+ /*
+ * Depends on ARM64_HAS_GIC_PRIO_MASKING
+ */
+ .capability = ARM64_HAS_GIC_PRIO_RELAXED_SYNC,
+ .type = ARM64_CPUCAP_STRICT_BOOT_CPU_FEATURE,
+ .matches = has_gic_prio_relaxed_sync,
+ },
#endif
#ifdef CONFIG_ARM64_E0PD
{
@@ -311,13 +311,16 @@ alternative_else_nop_endif
.endif
#ifdef CONFIG_ARM64_PSEUDO_NMI
- /* Save pmr */
-alternative_if ARM64_HAS_GIC_PRIO_MASKING
+alternative_if_not ARM64_HAS_GIC_PRIO_MASKING
+ b .Lskip_pmr_save\@
+alternative_else_nop_endif
+
mrs_s x20, SYS_ICC_PMR_EL1
str x20, [sp, #S_PMR_SAVE]
mov x20, #GIC_PRIO_IRQON | GIC_PRIO_PSR_I_SET
msr_s SYS_ICC_PMR_EL1, x20
-alternative_else_nop_endif
+
+.Lskip_pmr_save\@:
#endif
/*
@@ -336,15 +339,19 @@ alternative_else_nop_endif
.endif
#ifdef CONFIG_ARM64_PSEUDO_NMI
- /* Restore pmr */
-alternative_if ARM64_HAS_GIC_PRIO_MASKING
+alternative_if_not ARM64_HAS_GIC_PRIO_MASKING
+ b .Lskip_pmr_restore\@
+alternative_else_nop_endif
+
ldr x20, [sp, #S_PMR_SAVE]
msr_s SYS_ICC_PMR_EL1, x20
- mrs_s x21, SYS_ICC_CTLR_EL1
- tbz x21, #6, .L__skip_pmr_sync\@ // Check for ICC_CTLR_EL1.PMHE
- dsb sy // Ensure priority change is seen by redistributor
-.L__skip_pmr_sync\@:
+
+ /* Ensure priority change is seen by redistributor */
+alternative_if_not ARM64_HAS_GIC_PRIO_RELAXED_SYNC
+ dsb sy
alternative_else_nop_endif
+
+.Lskip_pmr_restore\@:
#endif
ldp x21, x22, [sp, #S_PC] // load ELR, SPSR
@@ -67,9 +67,7 @@ KVM_NVHE_ALIAS(__hyp_stub_vectors);
KVM_NVHE_ALIAS(vgic_v2_cpuif_trap);
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
@@ -30,6 +30,7 @@ HAS_GENERIC_AUTH_ARCH_QARMA5
HAS_GENERIC_AUTH_IMP_DEF
HAS_GIC_CPUIF_SYSREGS
HAS_GIC_PRIO_MASKING
+HAS_GIC_PRIO_RELAXED_SYNC
HAS_LDAPR
HAS_LSE_ATOMICS
HAS_NO_FPSIMD
@@ -89,15 +89,6 @@ static DEFINE_STATIC_KEY_TRUE(supports_deactivate_key);
*/
static DEFINE_STATIC_KEY_FALSE(supports_pseudo_nmis);
-/*
- * Global static key controlling whether an update to PMR allowing more
- * interrupts requires to be propagated to the redistributor (DSB SY).
- * And this needs to be exported for modules to be able to enable
- * interrupts...
- */
-DEFINE_STATIC_KEY_FALSE(gic_pmr_sync);
-EXPORT_SYMBOL(gic_pmr_sync);
-
DEFINE_STATIC_KEY_FALSE(gic_nonsecure_priorities);
EXPORT_SYMBOL(gic_nonsecure_priorities);
@@ -1768,16 +1759,8 @@ static void gic_enable_nmi_support(void)
for (i = 0; i < gic_data.ppi_nr; i++)
refcount_set(&ppi_nmi_refs[i], 0);
- /*
- * Linux itself doesn't use 1:N distribution, so has no need to
- * set PMHE. The only reason to have it set is if EL3 requires it
- * (and we can't change it).
- */
- if (gic_read_ctlr() & ICC_CTLR_EL1_PMHE_MASK)
- static_branch_enable(&gic_pmr_sync);
-
pr_info("Pseudo-NMIs enabled using %s ICC_PMR_EL1 synchronisation\n",
- static_branch_unlikely(&gic_pmr_sync) ? "forced" : "relaxed");
+ gic_has_relaxed_pmr_sync() ? "relaxed" : "forced");
/*
* How priority values are used by the GIC depends on two things:
When Priority Mask Hint Enable (PMHE) == 0b1, the GIC may use the PMR value to determine whether to signal an IRQ to a PE, and consequently after a change to the PMR value, a DSB SY may be required to ensure that interrupts are signalled to a CPU in finite time. When PMHE == 0b0, interrupts are always signalled to the relevant PE, and all masking occurs locally, without requiring a DSB SY. Since commit: f226650494c6aa87 ("arm64: Relax ICC_PMR_EL1 accesses when ICC_CTLR_EL1.PMHE is clear") ... we handle this dynamically: in most cases a static key is used to determine whether to issue a DSB SY, but the entry code must read from ICC_CTLR_EL1 as static keys aren't accessible from plain assembly. It would be much nicer to use an alternative instruction sequence for the DSB, as this would avoid the need to read from ICC_CTLR_EL1 in the entry code, and for most other code this will result in simpler code generation with fewer instructions and fewer branches. This patch adds a new ARM64_HAS_GIC_PRIO_RELAXED_SYNC cpucap which is only set when ICC_CTLR_EL1.PMHE == 0b0 (and GIC priority masking is in use). This allows us to replace the existing users of the `gic_pmr_sync` static key with alternative sequences which default to a DSB SY and are relaxed to a NOP when PMHE is not in use. The entry assembly management of the PMR is slightly restructured to use a branch (rather than multiple NOPs) when priority masking is not in use. This is more in keeping with other alternatives in the entry assembly, and permits the use of a separate alternatives for the PMHE-dependent DSB SY (and removal of the conditional branch this currently requires). For consistency I've adjusted both the save and restore paths. According to bloat-o-meter, when building defconfig + CONFIG_ARM64_PSEUDO_NMI=y this shrinks the kernel text by ~4KiB: | add/remove: 4/2 grow/shrink: 42/310 up/down: 332/-5032 (-4700) The resulting vmlinux is ~66KiB smaller, though the resulting Image size is unchanged due to padding and alignment: | [mark@lakrids:~/src/linux]% ls -al vmlinux-* | -rwxr-xr-x 1 mark mark 137508344 Jan 17 14:11 vmlinux-after | -rwxr-xr-x 1 mark mark 137575440 Jan 17 13:49 vmlinux-before | [mark@lakrids:~/src/linux]% ls -al Image-* | -rw-r--r-- 1 mark mark 38777344 Jan 17 14:11 Image-after | -rw-r--r-- 1 mark mark 38777344 Jan 17 13:49 Image-before Prior to this patch we did not verify the state of ICC_CTLR_EL1.PMHE on secondary CPUs. As of this patch this is verified by the cpufeature code when using GIC priority masking (i.e. when using pseudo-NMIs). Note that since commit: 7e3a57fa6ca831fa ("arm64: Document ICC_CTLR_EL3.PMHE setting requirements") ... Documentation/arm64/booting.rst specifies: | - ICC_CTLR_EL3.PMHE (bit 6) must be set to the same value across | all CPUs the kernel is executing on, and must stay constant | for the lifetime of the kernel. ... so that should not adversely affect any compliant systems, and as we'll only check for the absense of PMHE when using pseudo-NMIs, this will only fire when such mismatch will adversely affect the system. Signed-off-by: Mark Rutland <mark.rutland@arm.com> Reviewed-by: Marc Zyngier <maz@kernel.org> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Mark Brown <broonie@kernel.org> Cc: Will Deacon <will@kernel.org> --- arch/arm/include/asm/arch_gicv3.h | 5 ++++ arch/arm64/include/asm/arch_gicv3.h | 5 ++++ arch/arm64/include/asm/barrier.h | 11 +++++---- arch/arm64/kernel/cpufeature.c | 36 +++++++++++++++++++++++++++++ arch/arm64/kernel/entry.S | 25 ++++++++++++-------- arch/arm64/kernel/image-vars.h | 2 -- arch/arm64/tools/cpucaps | 1 + drivers/irqchip/irq-gic-v3.c | 19 +-------------- 8 files changed, 71 insertions(+), 33 deletions(-)