All of lore.kernel.org
 help / color / mirror / Atom feed
From: Thomas Gleixner <tglx@linutronix.de>
To: LKML <linux-kernel@vger.kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>, Peter Anvin <hpa@zytor.com>,
	Marc Zyngier <marc.zyngier@arm.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Borislav Petkov <bp@alien8.de>, Chen Yu <yu.c.chen@intel.com>,
	Rui Zhang <rui.zhang@intel.com>,
	"Rafael J. Wysocki" <rjw@rjwysocki.net>,
	Len Brown <lenb@kernel.org>,
	Dan Williams <dan.j.williams@intel.com>,
	Christoph Hellwig <hch@lst.de>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Joerg Roedel <joro@8bytes.org>,
	Boris Ostrovsky <boris.ostrovsky@oracle.com>,
	Juergen Gross <jgross@suse.com>, Tony Luck <tony.luck@intel.com>,
	"K. Y. Srinivasan" <kys@microsoft.com>,
	Alok Kataria <akataria@vmware.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Arjan van de Ven <arjan@linux.intel.com>
Subject: [patch 22/52] x86/apic/x2apic: Simplify cluster management
Date: Wed, 13 Sep 2017 23:29:24 +0200	[thread overview]
Message-ID: <20170913213153.992629420@linutronix.de> (raw)
In-Reply-To: 20170913212902.530704676@linutronix.de

[-- Attachment #1: x86-apic-x2apic--Simplify-cluster-management.patch --]
[-- Type: text/plain, Size: 8306 bytes --]

The cluster management code creates a cluster mask per cpu, which requires
that on cpu on/offline all cluster masks have to be iterated and
updated. Other information about the cluster is in different per cpu
variables.

Create a data structure which holds all information about a cluster and
fill it in when the first CPU of a cluster comes online. If another CPU of
a cluster comes online it just finds the pointer to the existing cluster
structure and reuses it.

That simplifies all usage sites and gets rid of quite some pointless
iterations over the online cpus to find the cpus which belong to the
cluster.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/kernel/apic/x2apic_cluster.c |  154 ++++++++++++++++------------------
 1 file changed, 76 insertions(+), 78 deletions(-)

--- a/arch/x86/kernel/apic/x2apic_cluster.c
+++ b/arch/x86/kernel/apic/x2apic_cluster.c
@@ -10,20 +10,22 @@
 #include <asm/smp.h>
 #include "x2apic.h"
 
+struct cluster_mask {
+	unsigned int	clusterid;
+	int		node;
+	struct cpumask	mask;
+};
+
 static DEFINE_PER_CPU(u32, x86_cpu_to_logical_apicid);
-static DEFINE_PER_CPU(cpumask_var_t, cpus_in_cluster);
 static DEFINE_PER_CPU(cpumask_var_t, ipi_mask);
+static DEFINE_PER_CPU(struct cluster_mask *, cluster_masks);
+static struct cluster_mask *cluster_hotplug_mask;
 
 static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 {
 	return x2apic_enabled();
 }
 
-static inline u32 x2apic_cluster(int cpu)
-{
-	return per_cpu(x86_cpu_to_logical_apicid, cpu) >> 16;
-}
-
 static void x2apic_send_IPI(int cpu, int vector)
 {
 	u32 dest = per_cpu(x86_cpu_to_logical_apicid, cpu);
@@ -35,49 +37,34 @@ static void x2apic_send_IPI(int cpu, int
 static void
 __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 {
-	struct cpumask *cpus_in_cluster_ptr;
-	struct cpumask *ipi_mask_ptr;
-	unsigned int cpu, this_cpu;
+	unsigned int cpu, clustercpu;
+	struct cpumask *tmpmsk;
 	unsigned long flags;
 	u32 dest;
 
 	x2apic_wrmsr_fence();
-
 	local_irq_save(flags);
 
-	this_cpu = smp_processor_id();
+	tmpmsk = this_cpu_cpumask_var_ptr(ipi_mask);
+	cpumask_copy(tmpmsk, mask);
+	/* If IPI should not be sent to self, clear current CPU */
+	if (apic_dest != APIC_DEST_ALLINC)
+		cpumask_clear_cpu(smp_processor_id(), tmpmsk);
+
+	/* Collapse cpus in a cluster so a single IPI per cluster is sent */
+	for_each_cpu(cpu, tmpmsk) {
+		struct cluster_mask *cmsk = per_cpu(cluster_masks, cpu);
 
-	/*
-	 * We are to modify mask, so we need an own copy
-	 * and be sure it's manipulated with irq off.
-	 */
-	ipi_mask_ptr = this_cpu_cpumask_var_ptr(ipi_mask);
-	cpumask_copy(ipi_mask_ptr, mask);
-
-	/*
-	 * The idea is to send one IPI per cluster.
-	 */
-	for_each_cpu(cpu, ipi_mask_ptr) {
-		unsigned long i;
-
-		cpus_in_cluster_ptr = per_cpu(cpus_in_cluster, cpu);
 		dest = 0;
-
-		/* Collect cpus in cluster. */
-		for_each_cpu_and(i, ipi_mask_ptr, cpus_in_cluster_ptr) {
-			if (apic_dest == APIC_DEST_ALLINC || i != this_cpu)
-				dest |= per_cpu(x86_cpu_to_logical_apicid, i);
-		}
+		for_each_cpu_and(clustercpu, tmpmsk, &cmsk->mask)
+			dest |= per_cpu(x86_cpu_to_logical_apicid, clustercpu);
 
 		if (!dest)
 			continue;
 
 		__x2apic_send_IPI_dest(dest, vector, apic->dest_logical);
-		/*
-		 * Cluster sibling cpus should be discared now so
-		 * we would not send IPI them second time.
-		 */
-		cpumask_andnot(ipi_mask_ptr, ipi_mask_ptr, cpus_in_cluster_ptr);
+		/* Remove cluster CPUs from tmpmask */
+		cpumask_andnot(tmpmsk, tmpmsk, &cmsk->mask);
 	}
 
 	local_irq_restore(flags);
@@ -109,91 +96,100 @@ x2apic_cpu_mask_to_apicid(const struct c
 			  unsigned int *apicid)
 {
 	struct cpumask *effmsk = irq_data_get_effective_affinity_mask(irqdata);
+	struct cluster_mask *cmsk;
 	unsigned int cpu;
 	u32 dest = 0;
-	u16 cluster;
 
 	cpu = cpumask_first(mask);
 	if (cpu >= nr_cpu_ids)
 		return -EINVAL;
 
-	dest = per_cpu(x86_cpu_to_logical_apicid, cpu);
-	cluster = x2apic_cluster(cpu);
-
+	cmsk = per_cpu(cluster_masks, cpu);
 	cpumask_clear(effmsk);
-	for_each_cpu(cpu, mask) {
-		if (cluster != x2apic_cluster(cpu))
-			continue;
+	for_each_cpu_and(cpu, &cmsk->mask, mask) {
 		dest |= per_cpu(x86_cpu_to_logical_apicid, cpu);
 		cpumask_set_cpu(cpu, effmsk);
 	}
-
 	*apicid = dest;
 	return 0;
 }
 
 static void init_x2apic_ldr(void)
 {
-	unsigned int this_cpu = smp_processor_id();
+	struct cluster_mask *cmsk = this_cpu_read(cluster_masks);
+	u32 cluster, apicid = apic_read(APIC_LDR);
 	unsigned int cpu;
 
-	per_cpu(x86_cpu_to_logical_apicid, this_cpu) = apic_read(APIC_LDR);
+	this_cpu_write(x86_cpu_to_logical_apicid, apicid);
 
-	cpumask_set_cpu(this_cpu, per_cpu(cpus_in_cluster, this_cpu));
+	if (cmsk)
+		goto update;
+
+	cluster = apicid >> 16;
 	for_each_online_cpu(cpu) {
-		if (x2apic_cluster(this_cpu) != x2apic_cluster(cpu))
-			continue;
-		cpumask_set_cpu(this_cpu, per_cpu(cpus_in_cluster, cpu));
-		cpumask_set_cpu(cpu, per_cpu(cpus_in_cluster, this_cpu));
+		cmsk = per_cpu(cluster_masks, cpu);
+		/* Matching cluster found. Link and update it. */
+		if (cmsk && cmsk->clusterid == cluster)
+			goto update;
 	}
+	cmsk = cluster_hotplug_mask;
+	cluster_hotplug_mask = NULL;
+update:
+	this_cpu_write(cluster_masks, cmsk);
+	cpumask_set_cpu(smp_processor_id(), &cmsk->mask);
 }
 
-/*
- * At CPU state changes, update the x2apic cluster sibling info.
- */
-static int x2apic_prepare_cpu(unsigned int cpu)
+static int alloc_clustermask(unsigned int cpu, int node)
 {
-	if (!zalloc_cpumask_var(&per_cpu(cpus_in_cluster, cpu), GFP_KERNEL))
-		return -ENOMEM;
+	if (per_cpu(cluster_masks, cpu))
+		return 0;
+	/*
+	 * If a hotplug spare mask exists, check whether it's on the right
+	 * node. If not, free it and allocate a new one.
+	 */
+	if (cluster_hotplug_mask) {
+		if (cluster_hotplug_mask->node == node)
+			return 0;
+		kfree(cluster_hotplug_mask);
+	}
 
-	if (!zalloc_cpumask_var(&per_cpu(ipi_mask, cpu), GFP_KERNEL)) {
-		free_cpumask_var(per_cpu(cpus_in_cluster, cpu));
+	cluster_hotplug_mask = kzalloc_node(sizeof(*cluster_hotplug_mask),
+					    GFP_KERNEL, node);
+	if (!cluster_hotplug_mask)
 		return -ENOMEM;
-	}
+	cluster_hotplug_mask->node = node;
+	return 0;
+}
 
+static int x2apic_prepare_cpu(unsigned int cpu)
+{
+	if (alloc_clustermask(cpu, cpu_to_node(cpu)) < 0)
+		return -ENOMEM;
+	if (!zalloc_cpumask_var(&per_cpu(ipi_mask, cpu), GFP_KERNEL))
+		return -ENOMEM;
 	return 0;
 }
 
-static int x2apic_dead_cpu(unsigned int this_cpu)
+static int x2apic_dead_cpu(unsigned int dead_cpu)
 {
-	int cpu;
+	struct cluster_mask *cmsk = per_cpu(cluster_masks, dead_cpu);
 
-	for_each_online_cpu(cpu) {
-		if (x2apic_cluster(this_cpu) != x2apic_cluster(cpu))
-			continue;
-		cpumask_clear_cpu(this_cpu, per_cpu(cpus_in_cluster, cpu));
-		cpumask_clear_cpu(cpu, per_cpu(cpus_in_cluster, this_cpu));
-	}
-	free_cpumask_var(per_cpu(cpus_in_cluster, this_cpu));
-	free_cpumask_var(per_cpu(ipi_mask, this_cpu));
+	cpumask_clear_cpu(smp_processor_id(), &cmsk->mask);
+	free_cpumask_var(per_cpu(ipi_mask, dead_cpu));
 	return 0;
 }
 
 static int x2apic_cluster_probe(void)
 {
-	int cpu = smp_processor_id();
-	int ret;
-
 	if (!x2apic_mode)
 		return 0;
 
-	ret = cpuhp_setup_state(CPUHP_X2APIC_PREPARE, "x86/x2apic:prepare",
-				x2apic_prepare_cpu, x2apic_dead_cpu);
-	if (ret < 0) {
+	if (cpuhp_setup_state(CPUHP_X2APIC_PREPARE, "x86/x2apic:prepare",
+			      x2apic_prepare_cpu, x2apic_dead_cpu) < 0) {
 		pr_err("Failed to register X2APIC_PREPARE\n");
 		return 0;
 	}
-	cpumask_set_cpu(cpu, per_cpu(cpus_in_cluster, cpu));
+	init_x2apic_ldr();
 	return 1;
 }
 
@@ -208,6 +204,8 @@ static const struct cpumask *x2apic_clus
 static void cluster_vector_allocation_domain(int cpu, struct cpumask *retmask,
 					     const struct cpumask *mask)
 {
+	struct cluster_mask *cmsk = per_cpu(cluster_masks, cpu);
+
 	/*
 	 * To minimize vector pressure, default case of boot, device bringup
 	 * etc will use a single cpu for the interrupt destination.
@@ -220,7 +218,7 @@ static void cluster_vector_allocation_do
 	if (mask == x2apic_cluster_target_cpus())
 		cpumask_copy(retmask, cpumask_of(cpu));
 	else
-		cpumask_and(retmask, mask, per_cpu(cpus_in_cluster, cpu));
+		cpumask_and(retmask, mask, &cmsk->mask);
 }
 
 static struct apic apic_x2apic_cluster __ro_after_init = {

  parent reply	other threads:[~2017-09-13 21:43 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-09-13 21:29 [patch 00/52] x86: Rework the vector management Thomas Gleixner
2017-09-13 21:29 ` [patch 01/52] genirq: Fix cpumask check in __irq_startup_managed() Thomas Gleixner
2017-09-16 18:24   ` [tip:irq/urgent] " tip-bot for Thomas Gleixner
2017-09-13 21:29 ` [patch 02/52] genirq/debugfs: Show debug information for all irq descriptors Thomas Gleixner
2017-09-13 21:29 ` [patch 03/52] genirq/msi: Capture device name for debugfs Thomas Gleixner
2017-09-13 21:29 ` [patch 04/52] irqdomain/debugfs: Provide domain specific debug callback Thomas Gleixner
2017-09-13 21:29 ` [patch 05/52] genirq: Make state consistent for !IRQ_DOMAIN_HIERARCHY Thomas Gleixner
2017-09-13 21:29 ` [patch 06/52] genirq: Set managed shut down flag at init Thomas Gleixner
2017-09-13 21:29 ` [patch 07/52] genirq: Separate activation and startup Thomas Gleixner
2017-09-13 21:29 ` [patch 08/52] genirq/irqdomain: Update irq_domain_ops.activate() signature Thomas Gleixner
2017-09-13 21:29 ` [patch 09/52] genirq/irqdomain: Allow irq_domain_activate_irq() to fail Thomas Gleixner
2017-09-13 21:29 ` [patch 10/52] genirq/irqdomain: Propagate early activation Thomas Gleixner
2017-09-13 21:29 ` [patch 11/52] genirq/irqdomain: Add force reactivation flag to irq domains Thomas Gleixner
2017-09-13 21:29 ` [patch 12/52] genirq: Implement bitmap matrix allocator Thomas Gleixner
2017-09-13 21:29 ` [patch 13/52] genirq/matrix: Add tracepoints Thomas Gleixner
2017-09-13 21:29 ` [patch 14/52] x86/apic: Deinline x2apic functions Thomas Gleixner
2017-09-13 21:29 ` [patch 15/52] x86/apic: Sanitize return value of apic.set_apic_id() Thomas Gleixner
2017-09-13 21:29 ` [patch 16/52] x86/apic: Sanitize return value of check_apicid_used() Thomas Gleixner
2017-09-13 21:29 ` [patch 17/52] x86/apic: Move probe32 specific APIC functions Thomas Gleixner
2017-09-13 21:29 ` [patch 18/52] x86/apic: Move APIC noop specific functions Thomas Gleixner
2017-09-13 21:29 ` [patch 19/52] x86/apic: Sanitize 32/64bit APIC callbacks Thomas Gleixner
2017-09-13 21:29 ` [patch 20/52] x86/apic: Move common " Thomas Gleixner
2017-09-13 21:29 ` [patch 21/52] x86/apic: Reorganize struct apic Thomas Gleixner
2017-09-13 21:29 ` Thomas Gleixner [this message]
2017-09-13 21:29 ` [patch 23/52] x86/apic: Get rid of apic->target_cpus Thomas Gleixner
2017-09-13 21:29 ` [patch 24/52] x86/vector: Rename used_vectors to system_vectors Thomas Gleixner
2017-09-13 21:29 ` [patch 25/52] x86/apic: Get rid of multi CPU affinity Thomas Gleixner
2017-09-13 21:29 ` [patch 26/52] x86/ioapic: Remove obsolete post hotplug update Thomas Gleixner
2017-09-13 21:29 ` [patch 27/52] x86/vector: Simplify the CPU hotplug vector update Thomas Gleixner
2017-09-13 21:29 ` [patch 28/52] x86/vector: Cleanup variable names Thomas Gleixner
2017-09-13 21:29 ` [patch 29/52] x86/vector: Store the single CPU targets in apic data Thomas Gleixner
2017-09-13 21:29 ` [patch 30/52] x86/vector: Simplify vector move cleanup Thomas Gleixner
2017-09-13 21:29 ` [patch 31/52] x86/ioapic: Mark legacy vectors at reallocation time Thomas Gleixner
2017-09-13 21:29 ` [patch 32/52] x86/apic: Get rid of the legacy irq data storage Thomas Gleixner
2017-09-13 21:29 ` [patch 33/52] x86/vector: Remove pointless pointer checks Thomas Gleixner
2017-09-13 21:29 ` [patch 34/52] x86/vector: Move helper functions around Thomas Gleixner
2017-09-13 21:29 ` [patch 35/52] x86/apic: Add replacement for cpu_mask_to_apicid() Thomas Gleixner
2017-09-13 21:29 ` [patch 36/52] x86/irq/vector: Initialize matrix allocator Thomas Gleixner
2017-09-13 21:29 ` [patch 37/52] x86/vector: Add vector domain debugfs support Thomas Gleixner
2017-09-13 21:29 ` [patch 38/52] x86/smpboot: Set online before setting up vectors Thomas Gleixner
2017-09-13 21:29 ` [patch 39/52] x86/vector: Add tracepoints for vector management Thomas Gleixner
2017-09-13 21:29 ` [patch 40/52] x86/vector: Use matrix allocator for vector assignment Thomas Gleixner
2017-09-13 21:29 ` [patch 41/52] x86/apic: Remove unused callbacks Thomas Gleixner
2017-09-13 21:29 ` [patch 42/52] x86/vector: Compile SMP only code conditionally Thomas Gleixner
2017-09-13 21:29 ` [patch 43/52] x86/vector: Untangle internal state from irq_cfg Thomas Gleixner
2017-09-13 21:29 ` [patch 44/52] x86/apic/msi: Force reactivation of interrupts at startup time Thomas Gleixner
2017-09-13 21:29 ` [patch 45/52] iommu/vt-d: Reevaluate vector configuration on activate() Thomas Gleixner
2017-09-13 21:29   ` Thomas Gleixner
2017-09-13 21:29 ` [patch 46/52] iommu/amd: " Thomas Gleixner
2017-09-13 21:29   ` Thomas Gleixner
2017-09-13 21:29 ` [patch 47/52] x86/io_apic: " Thomas Gleixner
2017-09-13 21:29 ` [patch 48/52] x86/vector: Handle managed interrupts proper Thomas Gleixner
2017-09-13 21:29 ` [patch 49/52] x86/vector/msi: Switch to global reservation mode Thomas Gleixner
2017-09-13 21:29 ` [patch 50/52] x86/vector: Switch IOAPIC " Thomas Gleixner
2017-09-13 21:29 ` [patch 51/52] x86/irq: Simplify hotplug vector accounting Thomas Gleixner
2017-09-13 21:29 ` [patch 52/52] x86/vector: Respect affinity mask in irq descriptor Thomas Gleixner
2017-09-14 11:21 ` [patch 00/52] x86: Rework the vector management Juergen Gross
2017-09-20 10:21   ` Paolo Bonzini
2017-09-19  9:12 ` Yu Chen

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=20170913213153.992629420@linutronix.de \
    --to=tglx@linutronix.de \
    --cc=akataria@vmware.com \
    --cc=arjan@linux.intel.com \
    --cc=boris.ostrovsky@oracle.com \
    --cc=bp@alien8.de \
    --cc=dan.j.williams@intel.com \
    --cc=hch@lst.de \
    --cc=hpa@zytor.com \
    --cc=jgross@suse.com \
    --cc=joro@8bytes.org \
    --cc=kys@microsoft.com \
    --cc=lenb@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=marc.zyngier@arm.com \
    --cc=mingo@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=peterz@infradead.org \
    --cc=rjw@rjwysocki.net \
    --cc=rostedt@goodmis.org \
    --cc=rui.zhang@intel.com \
    --cc=tony.luck@intel.com \
    --cc=yu.c.chen@intel.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.