cpumask: Fix crash on updating CPU enabled mask

The CPU enabled mask instead of the CPU possible mask should be used
by set_cpu_enabled(). Otherwise, we run into crash due to write to
the read-only CPU possible mask when vCPU is hot added on ARM64.

  (qemu) device_add host-arm-cpu,id=cpu1,socket-id=1
  Unable to handle kernel write to read-only memory at virtual address ffff800080fa7190
    :
  Call trace:
    register_cpu+0x1a4/0x2e8
    arch_register_cpu+0x84/0xd8
    acpi_processor_add+0x480/0x5b0
    acpi_bus_attach+0x1c4/0x300
    acpi_dev_for_one_check+0x3c/0x50
    device_for_each_child+0x68/0xc8
    acpi_dev_for_each_child+0x48/0x80
    acpi_bus_attach+0x84/0x300
    acpi_bus_scan+0x74/0x220
    acpi_scan_rescan_bus+0x54/0x88
    acpi_device_hotplug+0x208/0x478
    acpi_hotplug_work_fn+0x2c/0x50
    process_one_work+0x15c/0x3c0
    worker_thread+0x2ec/0x400
    kthread+0x120/0x130
    ret_from_fork+0x10/0x20

Fix it by passing the CPU enabled mask instead of the CPU possible
mask to set_cpu_enabled().

Fixes: 51c4767 ("Merge tag 'bitmap-6.11-rc1' of https://github.com:/norov/linux")
Signed-off-by: Gavin Shan <[email protected]>
Reviewed-by: Jonathan Cameron <[email protected]>
Signed-off-by: Yury Norov <[email protected]>

Author: Gavin Shan

include/linux/cpumask.h

@@ -1037,7 +1037,7 @@ void init_cpu_online(const struct cpumask *src);
 	assign_bit(cpumask_check(cpu), cpumask_bits(mask), (val))
 
 #define set_cpu_possible(cpu, possible)	assign_cpu((cpu), &__cpu_possible_mask, (possible))
-#define set_cpu_enabled(cpu, enabled)	assign_cpu((cpu), &__cpu_possible_mask, (enabled))
+#define set_cpu_enabled(cpu, enabled)	assign_cpu((cpu), &__cpu_enabled_mask, (enabled))
 #define set_cpu_present(cpu, present)	assign_cpu((cpu), &__cpu_present_mask, (present))
 #define set_cpu_active(cpu, active)	assign_cpu((cpu), &__cpu_active_mask, (active))
 #define set_cpu_dying(cpu, dying)	assign_cpu((cpu), &__cpu_dying_mask, (dying))