From: Ashwin Chaugule Date: Wed, 7 May 2014 14:18:36 +0000 (-0400) Subject: ARM: Check if a CPU has gone offline X-Git-Url: http://git.lede-project.org./?a=commitdiff_plain;h=c814ca029e1015bb0ecec312f4bb9751ba1a711a;p=openwrt%2Fstaging%2Fblogic.git ARM: Check if a CPU has gone offline PSCIv0.2 adds a new function called AFFINITY_INFO, which can be used to query if a specified CPU has actually gone offline. Calling this function via cpu_kill ensures that a CPU has quiesced after a call to cpu_die. This helps prevent the CPU from doing arbitrary bad things when data or instructions are clobbered (as happens with kexec) in the window between a CPU announcing that it is dead and said CPU leaving the kernel. Signed-off-by: Ashwin Chaugule Signed-off-by: Mark Rutland Reviewed-by: Rob Herring Acked-by: Catalin Marinas --- diff --git a/arch/arm/kernel/psci_smp.c b/arch/arm/kernel/psci_smp.c index 570a48cc3d64..28a1db4da704 100644 --- a/arch/arm/kernel/psci_smp.c +++ b/arch/arm/kernel/psci_smp.c @@ -16,6 +16,8 @@ #include #include #include +#include +#include #include #include @@ -66,6 +68,36 @@ void __ref psci_cpu_die(unsigned int cpu) /* We should never return */ panic("psci: cpu %d failed to shutdown\n", cpu); } + +int __ref psci_cpu_kill(unsigned int cpu) +{ + int err, i; + + if (!psci_ops.affinity_info) + return 1; + /* + * cpu_kill could race with cpu_die and we can + * potentially end up declaring this cpu undead + * while it is dying. So, try again a few times. + */ + + for (i = 0; i < 10; i++) { + err = psci_ops.affinity_info(cpu_logical_map(cpu), 0); + if (err == PSCI_0_2_AFFINITY_LEVEL_OFF) { + pr_info("CPU%d killed.\n", cpu); + return 1; + } + + msleep(10); + pr_info("Retrying again to check for CPU kill\n"); + } + + pr_warn("CPU%d may not have shut down cleanly (AFFINITY_INFO reports %d)\n", + cpu, err); + /* Make platform_cpu_kill() fail. */ + return 0; +} + #endif bool __init psci_smp_available(void) @@ -78,5 +110,6 @@ struct smp_operations __initdata psci_smp_ops = { .smp_boot_secondary = psci_boot_secondary, #ifdef CONFIG_HOTPLUG_CPU .cpu_die = psci_cpu_die, + .cpu_kill = psci_cpu_kill, #endif }; diff --git a/arch/arm64/include/asm/cpu_ops.h b/arch/arm64/include/asm/cpu_ops.h index 152413076503..d7b4b38a8e86 100644 --- a/arch/arm64/include/asm/cpu_ops.h +++ b/arch/arm64/include/asm/cpu_ops.h @@ -39,6 +39,7 @@ struct device_node; * from the cpu to be killed. * @cpu_die: Makes a cpu leave the kernel. Must not fail. Called from the * cpu being killed. + * @cpu_kill: Ensures a cpu has left the kernel. Called from another cpu. * @cpu_suspend: Suspends a cpu and saves the required context. May fail owing * to wrong parameters or error conditions. Called from the * CPU being suspended. Must be called with IRQs disabled. @@ -52,6 +53,7 @@ struct cpu_operations { #ifdef CONFIG_HOTPLUG_CPU int (*cpu_disable)(unsigned int cpu); void (*cpu_die)(unsigned int cpu); + int (*cpu_kill)(unsigned int cpu); #endif #ifdef CONFIG_ARM64_CPU_SUSPEND int (*cpu_suspend)(unsigned long); diff --git a/arch/arm64/kernel/psci.c b/arch/arm64/kernel/psci.c index 90df6e641227..9e9798f91172 100644 --- a/arch/arm64/kernel/psci.c +++ b/arch/arm64/kernel/psci.c @@ -20,6 +20,7 @@ #include #include #include +#include #include #include @@ -403,6 +404,35 @@ static void cpu_psci_cpu_die(unsigned int cpu) pr_crit("unable to power off CPU%u (%d)\n", cpu, ret); } + +static int cpu_psci_cpu_kill(unsigned int cpu) +{ + int err, i; + + if (!psci_ops.affinity_info) + return 1; + /* + * cpu_kill could race with cpu_die and we can + * potentially end up declaring this cpu undead + * while it is dying. So, try again a few times. + */ + + for (i = 0; i < 10; i++) { + err = psci_ops.affinity_info(cpu_logical_map(cpu), 0); + if (err == PSCI_0_2_AFFINITY_LEVEL_OFF) { + pr_info("CPU%d killed.\n", cpu); + return 1; + } + + msleep(10); + pr_info("Retrying again to check for CPU kill\n"); + } + + pr_warn("CPU%d may not have shut down cleanly (AFFINITY_INFO reports %d)\n", + cpu, err); + /* Make op_cpu_kill() fail. */ + return 0; +} #endif const struct cpu_operations cpu_psci_ops = { @@ -413,6 +443,7 @@ const struct cpu_operations cpu_psci_ops = { #ifdef CONFIG_HOTPLUG_CPU .cpu_disable = cpu_psci_cpu_disable, .cpu_die = cpu_psci_cpu_die, + .cpu_kill = cpu_psci_cpu_kill, #endif }; diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c index f0a141dd5655..c3cb160edc69 100644 --- a/arch/arm64/kernel/smp.c +++ b/arch/arm64/kernel/smp.c @@ -228,6 +228,19 @@ int __cpu_disable(void) return 0; } +static int op_cpu_kill(unsigned int cpu) +{ + /* + * If we have no means of synchronising with the dying CPU, then assume + * that it is really dead. We can only wait for an arbitrary length of + * time and hope that it's dead, so let's skip the wait and just hope. + */ + if (!cpu_ops[cpu]->cpu_kill) + return 1; + + return cpu_ops[cpu]->cpu_kill(cpu); +} + static DECLARE_COMPLETION(cpu_died); /* @@ -241,6 +254,15 @@ void __cpu_die(unsigned int cpu) return; } pr_notice("CPU%u: shutdown\n", cpu); + + /* + * Now that the dying CPU is beyond the point of no return w.r.t. + * in-kernel synchronisation, try to get the firwmare to help us to + * verify that it has really left the kernel before we consider + * clobbering anything it might still be using. + */ + if (!op_cpu_kill(cpu)) + pr_warn("CPU%d may not have shut down cleanly\n", cpu); } /*