[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v1 3/4] arm/sysctl: Implement cpu hotplug ops



Hi Mykyta,

On 23/09/2025 14:37, Mykyta Poturai wrote:
On 18.09.25 16:35, Julien Grall wrote:
Hi Mykyta,

On 18/09/2025 13:16, Mykyta Poturai wrote:
Implement XEN_SYSCTL_CPU_HOTPLUG_* calls to allow for enabling/disabling
CPU cores in runtime.

Signed-off-by: Mykyta Poturai <mykyta_poturai@xxxxxxxx>
---
   xen/arch/arm/sysctl.c | 67 +++++++++++++++++++++++++++++++++++++++++++
   1 file changed, 67 insertions(+)

diff --git a/xen/arch/arm/sysctl.c b/xen/arch/arm/sysctl.c
index 32cab4feff..ca8fb550fd 100644
--- a/xen/arch/arm/sysctl.c
+++ b/xen/arch/arm/sysctl.c
@@ -12,6 +12,7 @@
   #include <xen/dt-overlay.h>
   #include <xen/errno.h>
   #include <xen/hypercall.h>
+#include <xen/cpu.h>
   #include <asm/arm64/sve.h>
   #include <public/sysctl.h>
@@ -23,6 +24,68 @@ void arch_do_physinfo(struct xen_sysctl_physinfo *pi)
XEN_SYSCTL_PHYSCAP_ARM_SVE_MASK);
   }
+static long cpu_up_helper(void *data)
+{
+    unsigned long cpu = (unsigned long) data;
+    return cpu_up(cpu);
+}
+
+static long cpu_down_helper(void *data)
+{
+    unsigned long cpu = (unsigned long) data;
+    return cpu_down(cpu);
+}
+
+static long smt_up_down_helper(void *data)

Looking at the code, you will effectively disable all the CPUs but CPU0.
But I don't understand why. From the name is goal seems to be disable
SMT threading.


Sorry I have slightly misunderstood the x86 implementation/reasoning of
this ops. I will drop them in V2.

+{
+    bool up = (bool) data;
+    unsigned int cpu;
+    int ret;
+
+    for_each_present_cpu ( cpu )
+    {
+        if ( cpu == 0 )
+            continue;
+
+        if ( up )
+            ret = cpu_up(cpu);
+        else
+            ret = cpu_down(cpu);
+

Regardless what I wrote above, you likely want to handle preemption.

+        if ( ret )
+            return ret;
  > +    }
+
+    return 0;
+}
+
+static long cpu_hotplug_sysctl(struct xen_sysctl_cpu_hotplug *hotplug)
+{
+    bool up;
+
+    switch (hotplug->op) {
+        case XEN_SYSCTL_CPU_HOTPLUG_ONLINE:
+            if ( hotplug->cpu == 0 )

I can't find a similar check on x86. Do you have any pointer?

Jan correctly mentioned that CPU0 can't be disabled so this is a short
circuit for clarity.

I have replied to Jan. In short, the clarify you are referring is what would make more difficult to support offlining CPU0. So I would rather prefer if they are not present.

+            return continue_hypercall_on_cpu(0, cpu_up_helper,
_p(hotplug->cpu));
+
+        case XEN_SYSCTL_CPU_HOTPLUG_OFFLINE:
+            if ( hotplug->cpu == 0 )
+                return -EINVAL;
+            return continue_hypercall_on_cpu(0, cpu_down_helper,
_p(hotplug->cpu));
+
+        case XEN_SYSCTL_CPU_HOTPLUG_SMT_ENABLE:
+        case XEN_SYSCTL_CPU_HOTPLUG_SMT_DISABLE:

Why are we implementing those helpers on Arm?

+            if ( CONFIG_NR_CPUS <= 1 )
+                return 0;
+            up = hotplug->op == XEN_SYSCTL_CPU_HOTPLUG_SMT_ENABLE;
+            return continue_hypercall_on_cpu(0, smt_up_down_helper,
_p(up));
+
+        default:
+            return -EINVAL;
+    }
+}
+
   long arch_do_sysctl(struct xen_sysctl *sysctl,
                       XEN_GUEST_HANDLE_PARAM(xen_sysctl_t) u_sysctl)
   {
@@ -34,6 +97,10 @@ long arch_do_sysctl(struct xen_sysctl *sysctl,
           ret = dt_overlay_sysctl(&sysctl->u.dt_overlay);
           break;
+    case XEN_SYSCTL_cpu_hotplug:

This will also enable CPU hotplug on 32-bit Arm. Is this what you
intended? (I see patch #4 only mention 64-bit Arm).

It wasn't intended. I will additionally check if it works on arm32 end
explicitly specify it.

It will not work properly on arm32 because of the page table code. We have per-CPU pagetables (see init_domheap_mappings()) and they will need to be freed.

Note this is not a request to add support for arm32 CPU offlining.

Cheers,

--
Julien Grall




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.