1        CPU hotplug Support in Linux(tm) Kernel
3        Maintainers:
4        CPU Hotplug Core:
5            Rusty Russell <>
6            Srivatsa Vaddagiri <>
7        i386:
8            Zwane Mwaikambo <>
9        ppc64:
10            Nathan Lynch <>
11            Joel Schopp <>
12        ia64/x86_64:
13            Ashok Raj <>
14        s390:
15            Heiko Carstens <>
17Authors: Ashok Raj <>
18Lots of feedback: Nathan Lynch <>,
19         Joel Schopp <>
23Modern advances in system architectures have introduced advanced error
24reporting and correction capabilities in processors. CPU architectures permit
25partitioning support, where compute resources of a single CPU could be made
26available to virtual machine environments. There are couple OEMS that
27support NUMA hardware which are hot pluggable as well, where physical
28node insertion and removal require support for CPU hotplug.
30Such advances require CPUs available to a kernel to be removed either for
31provisioning reasons, or for RAS purposes to keep an offending CPU off
32system execution path. Hence the need for CPU hotplug support in the
33Linux kernel.
35A more novel use of CPU-hotplug support is its use today in suspend
36resume support for SMP. Dual-core and HT support makes even
37a laptop run SMP kernels which didn't support these methods. SMP support
38for suspend/resume is a work in progress.
40General Stuff about CPU Hotplug
43Command Line Switches
45maxcpus=n Restrict boot time cpus to n. Say if you have 4 cpus, using
46             maxcpus=2 will only boot 2. You can choose to bring the
47             other cpus later online, read FAQ's for more info.
49additional_cpus=n (*) Use this to limit hotpluggable cpus. This option sets
50              cpu_possible_map = cpu_present_map + additional_cpus
52cede_offline={"off","on"} Use this option to disable/enable putting offlined
53                    processors to an extended H_CEDE state on
54                supported pseries platforms.
55                If nothing is specified,
56                cede_offline is set to "on".
58(*) Option valid only for following architectures
59- ia64
61ia64 uses the number of disabled local apics in ACPI tables MADT to
62determine the number of potentially hot-pluggable cpus. The implementation
63should only rely on this to count the # of cpus, but *MUST* not rely
64on the apicid values in those tables for disabled apics. In the event
65BIOS doesn't mark such hot-pluggable cpus as disabled entries, one could
66use this parameter "additional_cpus=x" to represent those cpus in the
69possible_cpus=n [s390,x86_64] use this to set hotpluggable cpus.
70            This option sets possible_cpus bits in
71            cpu_possible_map. Thus keeping the numbers of bits set
72            constant even if the machine gets rebooted.
74CPU maps and such
76[More on cpumaps and primitive to manipulate, please check
77include/linux/cpumask.h that has more descriptive text.]
79cpu_possible_map: Bitmap of possible CPUs that can ever be available in the
80system. This is used to allocate some boot time memory for per_cpu variables
81that aren't designed to grow/shrink as CPUs are made available or removed.
82Once set during boot time discovery phase, the map is static, i.e no bits
83are added or removed anytime. Trimming it accurately for your system needs
84upfront can save some boot time memory. See below for how we use heuristics
85in x86_64 case to keep this under check.
87cpu_online_map: Bitmap of all CPUs currently online. Its set in __cpu_up()
88after a cpu is available for kernel scheduling and ready to receive
89interrupts from devices. Its cleared when a cpu is brought down using
90__cpu_disable(), before which all OS services including interrupts are
91migrated to another target CPU.
93cpu_present_map: Bitmap of CPUs currently present in the system. Not all
94of them may be online. When physical hotplug is processed by the relevant
95subsystem (e.g ACPI) can change and new bit either be added or removed
96from the map depending on the event is hot-add/hot-remove. There are currently
97no locking rules as of now. Typical usage is to init topology during boot,
98at which time hotplug is disabled.
100You really dont need to manipulate any of the system cpu maps. They should
101be read-only for most use. When setting up per-cpu resources almost always use
102cpu_possible_map/for_each_possible_cpu() to iterate.
104Never use anything other than cpumask_t to represent bitmap of CPUs.
106    #include <linux/cpumask.h>
108    for_each_possible_cpu - Iterate over cpu_possible_map
109    for_each_online_cpu - Iterate over cpu_online_map
110    for_each_present_cpu - Iterate over cpu_present_map
111    for_each_cpu_mask(x,mask) - Iterate over some random collection of cpu mask.
113    #include <linux/cpu.h>
114    get_online_cpus() and put_online_cpus():
116The above calls are used to inhibit cpu hotplug operations. While the
117cpu_hotplug.refcount is non zero, the cpu_online_map will not change.
118If you merely need to avoid cpus going away, you could also use
119preempt_disable() and preempt_enable() for those sections.
120Just remember the critical section cannot call any
121function that can sleep or schedule this process away. The preempt_disable()
122will work as long as stop_machine_run() is used to take a cpu down.
124CPU Hotplug - Frequently Asked Questions.
126Q: How to enable my kernel to support CPU hotplug?
127A: When doing make defconfig, Enable CPU hotplug support
129   "Processor type and Features" -> Support for Hotpluggable CPUs
131Make sure that you have CONFIG_HOTPLUG, and CONFIG_SMP turned on as well.
133You would need to enable CONFIG_HOTPLUG_CPU for SMP suspend/resume support
134as well.
136Q: What architectures support CPU hotplug?
137A: As of 2.6.14, the following architectures support CPU hotplug.
139i386 (Intel), ppc, ppc64, parisc, s390, ia64 and x86_64
141Q: How to test if hotplug is supported on the newly built kernel?
142A: You should now notice an entry in sysfs.
144Check if sysfs is mounted, using the "mount" command. You should notice
145an entry as shown below in the output.
147    ....
148    none on /sys type sysfs (rw)
149    ....
151If this is not mounted, do the following.
153     #mkdir /sysfs
154    #mount -t sysfs sys /sys
156Now you should see entries for all present cpu, the following is an example
157in a 8-way system.
159    #pwd
160    #/sys/devices/system/cpu
161    #ls -l
162    total 0
163    drwxr-xr-x 10 root root 0 Sep 19 07:44 .
164    drwxr-xr-x 13 root root 0 Sep 19 07:45 ..
165    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu0
166    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu1
167    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu2
168    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu3
169    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu4
170    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu5
171    drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu6
172    drwxr-xr-x 3 root root 0 Sep 19 07:48 cpu7
174Under each directory you would find an "online" file which is the control
175file to logically online/offline a processor.
177Q: Does hot-add/hot-remove refer to physical add/remove of cpus?
178A: The usage of hot-add/remove may not be very consistently used in the code.
179CONFIG_HOTPLUG_CPU enables logical online/offline capability in the kernel.
180To support physical addition/removal, one would need some BIOS hooks and
181the platform should have something like an attention button in PCI hotplug.
182CONFIG_ACPI_HOTPLUG_CPU enables ACPI support for physical add/remove of CPUs.
184Q: How do i logically offline a CPU?
185A: Do the following.
187    #echo 0 > /sys/devices/system/cpu/cpuX/online
189Once the logical offline is successful, check
191    #cat /proc/interrupts
193You should now not see the CPU that you removed. Also online file will report
194the state as 0 when a cpu if offline and 1 when its online.
196    #To display the current cpu state.
197    #cat /sys/devices/system/cpu/cpuX/online
199Q: Why cant i remove CPU0 on some systems?
200A: Some architectures may have some special dependency on a certain CPU.
202For e.g in IA64 platforms we have ability to sent platform interrupts to the
203OS. a.k.a Corrected Platform Error Interrupts (CPEI). In current ACPI
204specifications, we didn't have a way to change the target CPU. Hence if the
205current ACPI version doesn't support such re-direction, we disable that CPU
206by making it not-removable.
208In such cases you will also notice that the online file is missing under cpu0.
210Q: How do i find out if a particular CPU is not removable?
211A: Depending on the implementation, some architectures may show this by the
212absence of the "online" file. This is done if it can be determined ahead of
213time that this CPU cannot be removed.
215In some situations, this can be a run time check, i.e if you try to remove the
216last CPU, this will not be permitted. You can find such failures by
217investigating the return value of the "echo" command.
219Q: What happens when a CPU is being logically offlined?
220A: The following happen, listed in no particular order :-)
222- A notification is sent to in-kernel registered modules by sending an event
223  CPU_DOWN_PREPARE or CPU_DOWN_PREPARE_FROZEN, depending on whether or not the
224  CPU is being offlined while tasks are frozen due to a suspend operation in
225  progress
226- All processes are migrated away from this outgoing CPU to new CPUs.
227  The new CPU is chosen from each process' current cpuset, which may be
228  a subset of all online CPUs.
229- All interrupts targeted to this CPU is migrated to a new CPU
230- timers/bottom half/task lets are also migrated to a new CPU
231- Once all services are migrated, kernel calls an arch specific routine
232  __cpu_disable() to perform arch specific cleanup.
233- Once this is successful, an event for successful cleanup is sent by an event
234  CPU_DEAD (or CPU_DEAD_FROZEN if tasks are frozen due to a suspend while the
235  CPU is being offlined).
237  "It is expected that each service cleans up when the CPU_DOWN_PREPARE
238  notifier is called, when CPU_DEAD is called its expected there is nothing
239  running on behalf of this CPU that was offlined"
241Q: If i have some kernel code that needs to be aware of CPU arrival and
242   departure, how to i arrange for proper notification?
243A: This is what you would need in your kernel code to receive notifications.
245    #include <linux/cpu.h>
246    static int __cpuinit foobar_cpu_callback(struct notifier_block *nfb,
247                        unsigned long action, void *hcpu)
248    {
249        unsigned int cpu = (unsigned long)hcpu;
251        switch (action) {
252        case CPU_ONLINE:
253        case CPU_ONLINE_FROZEN:
254            foobar_online_action(cpu);
255            break;
256        case CPU_DEAD:
257        case CPU_DEAD_FROZEN:
258            foobar_dead_action(cpu);
259            break;
260        }
261        return NOTIFY_OK;
262    }
264    static struct notifier_block __cpuinitdata foobar_cpu_notifer =
265    {
266       .notifier_call = foobar_cpu_callback,
267    };
269You need to call register_cpu_notifier() from your init function.
270Init functions could be of two types:
2711. early init (init function called when only the boot processor is online).
2722. late init (init function called _after_ all the CPUs are online).
274For the first case, you should add the following to your init function
276    register_cpu_notifier(&foobar_cpu_notifier);
278For the second case, you should add the following to your init function
280    register_hotcpu_notifier(&foobar_cpu_notifier);
282You can fail PREPARE notifiers if something doesn't work to prepare resources.
283This will stop the activity and send a following CANCELED event back.
285CPU_DEAD should not be failed, its just a goodness indication, but bad
286things will happen if a notifier in path sent a BAD notify code.
288Q: I don't see my action being called for all CPUs already up and running?
289A: Yes, CPU notifiers are called only when new CPUs are on-lined or offlined.
290   If you need to perform some action for each cpu already in the system, then
292    for_each_online_cpu(i) {
293        foobar_cpu_callback(&foobar_cpu_notifier, CPU_UP_PREPARE, i);
294        foobar_cpu_callback(&foobar_cpu_notifier, CPU_ONLINE, i);
295    }
297Q: If i would like to develop cpu hotplug support for a new architecture,
298   what do i need at a minimum?
299A: The following are what is required for CPU hotplug infrastructure to work
300   correctly.
302    - Make sure you have an entry in Kconfig to enable CONFIG_HOTPLUG_CPU
303    - __cpu_up() - Arch interface to bring up a CPU
304    - __cpu_disable() - Arch interface to shutdown a CPU, no more interrupts
305                          can be handled by the kernel after the routine
306                          returns. Including local APIC timers etc are
307                          shutdown.
308     - __cpu_die() - This actually supposed to ensure death of the CPU.
309                          Actually look at some example code in other arch
310                          that implement CPU hotplug. The processor is taken
311                          down from the idle() loop for that specific
312                          architecture. __cpu_die() typically waits for some
313                          per_cpu state to be set, to ensure the processor
314                          dead routine is called to be sure positively.
316Q: I need to ensure that a particular cpu is not removed when there is some
317   work specific to this cpu is in progress.
318A: There are two ways. If your code can be run in interrupt context, use
319   smp_call_function_single(), otherwise use work_on_cpu(). Note that
320   work_on_cpu() is slow, and can fail due to out of memory:
322    int my_func_on_cpu(int cpu)
323    {
324        int err;
325        get_online_cpus();
326        if (!cpu_online(cpu))
327            err = -EINVAL;
328        else
330            err = work_on_cpu(cpu, __my_func_on_cpu, NULL);
332            smp_call_function_single(cpu, __my_func_on_cpu, &err,
333                         true);
335        put_online_cpus();
336        return err;
337    }
339Q: How do we determine how many CPUs are available for hotplug.
340A: There is no clear spec defined way from ACPI that can give us that
341   information today. Based on some input from Natalie of Unisys,
342   that the ACPI MADT (Multiple APIC Description Tables) marks those possible
343   CPUs in a system with disabled status.
345   Andi implemented some simple heuristics that count the number of disabled
346   CPUs in MADT as hotpluggable CPUS. In the case there are no disabled CPUS
347   we assume 1/2 the number of CPUs currently present can be hotplugged.
349   Caveat: Today's ACPI MADT can only provide 256 entries since the apicid field
350   in MADT is only 8 bits.
352User Space Notification
354Hotplug support for devices is common in Linux today. Its being used today to
355support automatic configuration of network, usb and pci devices. A hotplug
356event can be used to invoke an agent script to perform the configuration task.
358You can add /etc/hotplug/cpu.agent to handle hotplug notification user space
361    #!/bin/bash
362    # $Id: cpu.agent
363    # Kernel hotplug params include:
364    #ACTION=%s [online or offline]
365    #DEVPATH=%s
366    #
367    cd /etc/hotplug
368    . ./hotplug.functions
370    case $ACTION in
371        online)
372            echo `date` ":cpu.agent" add cpu >> /tmp/hotplug.txt
373            ;;
374        offline)
375            echo `date` ":cpu.agent" remove cpu >>/tmp/hotplug.txt
376            ;;
377        *)
378            debug_mesg CPU $ACTION event not supported
379        exit 1
380        ;;
381    esac

Archive Download this file