Commit ae039001 authored by Ankur Arora's avatar Ankur Arora Committed by Juergen Gross
Browse files

xen/vcpu: Handle xen_vcpu_setup() failure at boot 



On PVH, PVHVM, at failure in the VCPUOP_register_vcpu_info hypercall
we limit the number of cpus to to MAX_VIRT_CPUS. However, if this
failure had occurred for a cpu beyond MAX_VIRT_CPUS, we continue
to function with > MAX_VIRT_CPUS.

This leads to problems at the next save/restore cycle when there
are > MAX_VIRT_CPUS threads going into stop_machine() but coming
back up there's valid state for only the first MAX_VIRT_CPUS.

This patch pulls the excess CPUs down via cpu_down().

Reviewed-by: default avatarBoris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: default avatarAnkur Arora <ankur.a.arora@oracle.com>
Signed-off-by: default avatarJuergen Gross <jgross@suse.com>
parent c9b5d98b

arch/x86/xen/smp.c

+31 −0
Original line number Diff line number Diff line 
#include <linux/smp.h>

#include <linux/cpu.h>

#include <linux/slab.h>

#include <linux/cpumask.h>

#include <linux/percpu.h>
@@ -114,6 +115,36 @@ int xen_smp_intr_init(unsigned int cpu) 
	return rc;

}



void __init xen_smp_cpus_done(unsigned int max_cpus)

{

	int cpu, rc, count = 0;



	if (xen_hvm_domain())

		native_smp_cpus_done(max_cpus);



	if (xen_have_vcpu_info_placement)

		return;



	for_each_online_cpu(cpu) {

		if (xen_vcpu_nr(cpu) < MAX_VIRT_CPUS)

			continue;



		rc = cpu_down(cpu);



		if (rc == 0) {

			/*

			 * Reset vcpu_info so this cpu cannot be onlined again.

			 */

			xen_vcpu_info_reset(cpu);

			count++;

		} else {

			pr_warn("%s: failed to bring CPU %d down, error %d\n",

				__func__, cpu, rc);

		}

	}

	WARN(count, "%s: brought %d CPUs offline\n", __func__, count);

}



void xen_smp_send_reschedule(int cpu)

{

	xen_send_IPI_one(cpu, XEN_RESCHEDULE_VECTOR);

arch/x86/xen/smp.h

+2 −0
Original line number Diff line number Diff line
@@ -14,6 +14,8 @@ extern void xen_smp_intr_free(unsigned int cpu); 
int xen_smp_intr_init_pv(unsigned int cpu);

void xen_smp_intr_free_pv(unsigned int cpu);



void xen_smp_cpus_done(unsigned int max_cpus);



void xen_smp_send_reschedule(int cpu);

void xen_smp_send_call_function_ipi(const struct cpumask *mask);

void xen_smp_send_call_function_single_ipi(int cpu);

arch/x86/xen/smp_hvm.c

+1 −0
Original line number Diff line number Diff line
@@ -71,4 +71,5 @@ void __init xen_hvm_smp_init(void) 
	smp_ops.send_call_func_ipi = xen_smp_send_call_function_ipi;

	smp_ops.send_call_func_single_ipi = xen_smp_send_call_function_single_ipi;

	smp_ops.smp_prepare_boot_cpu = xen_hvm_smp_prepare_boot_cpu;

	smp_ops.smp_cpus_done = xen_smp_cpus_done;

}

arch/x86/xen/smp_pv.c

+1 −5
Original line number Diff line number Diff line
@@ -371,10 +371,6 @@ static int xen_pv_cpu_up(unsigned int cpu, struct task_struct *idle) 
	return 0;

}



static void xen_pv_smp_cpus_done(unsigned int max_cpus)

{

}



#ifdef CONFIG_HOTPLUG_CPU

static int xen_pv_cpu_disable(void)

{
@@ -469,7 +465,7 @@ static irqreturn_t xen_irq_work_interrupt(int irq, void *dev_id) 
static const struct smp_ops xen_smp_ops __initconst = {

	.smp_prepare_boot_cpu = xen_pv_smp_prepare_boot_cpu,

	.smp_prepare_cpus = xen_pv_smp_prepare_cpus,

	.smp_cpus_done = xen_pv_smp_cpus_done,

	.smp_cpus_done = xen_smp_cpus_done,



	.cpu_up = xen_pv_cpu_up,

	.cpu_die = xen_pv_cpu_die,

CRA Git | Maintained and supported by SUSTech CRA and CCSE