1 CPU hotplug Support in Linux(tm) Kernel
5 Rusty Russell <rusty@rustycorp.com.au>
6 Srivatsa Vaddagiri <vatsa@in.ibm.com>
8 Zwane Mwaikambo <zwane@arm.linux.org.uk>
10 Nathan Lynch <nathanl@austin.ibm.com>
11 Joel Schopp <jschopp@austin.ibm.com>
13 Ashok Raj <ashok.raj@intel.com>
15 Heiko Carstens <heiko.carstens@de.ibm.com>
17 Authors: Ashok Raj <ashok.raj@intel.com>
18 Lots of feedback: Nathan Lynch <nathanl@austin.ibm.com>,
19 Joel Schopp <jschopp@austin.ibm.com>
23 Modern advances in system architectures have introduced advanced error
24 reporting and correction capabilities in processors. CPU architectures permit
25 partitioning support, where compute resources of a single CPU could be made
26 available to virtual machine environments. There are couple OEMS that
27 support NUMA hardware which are hot pluggable as well, where physical
28 node insertion and removal require support for CPU hotplug.
30 Such advances require CPUs available to a kernel to be removed either for
31 provisioning reasons, or for RAS purposes to keep an offending CPU off
32 system execution path. Hence the need for CPU hotplug support in the
35 A more novel use of CPU-hotplug support is its use today in suspend
36 resume support for SMP. Dual-core and HT support makes even
37 a laptop run SMP kernels which didn't support these methods. SMP support
38 for suspend/resume is a work in progress.
40 General Stuff about CPU Hotplug
41 --------------------------------
45 maxcpus=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.
49 additional_cpus=n [x86_64, s390 only] use this to limit hotpluggable cpus.
51 cpu_possible_map = cpu_present_map + additional_cpus
53 ia64 and x86_64 use the number of disabled local apics in ACPI tables MADT
54 to determine the number of potentially hot-pluggable cpus. The implementation
55 should only rely on this to count the #of cpus, but *MUST* not rely on the
56 apicid values in those tables for disabled apics. In the event BIOS doesnt
57 mark such hot-pluggable cpus as disabled entries, one could use this
58 parameter "additional_cpus=x" to represent those cpus in the cpu_possible_map.
61 possible_cpus=n [s390 only] use this to set hotpluggable cpus.
62 This option sets possible_cpus bits in
63 cpu_possible_map. Thus keeping the numbers of bits set
64 constant even if the machine gets rebooted.
65 This option overrides additional_cpus.
69 [More on cpumaps and primitive to manipulate, please check
70 include/linux/cpumask.h that has more descriptive text.]
72 cpu_possible_map: Bitmap of possible CPUs that can ever be available in the
73 system. This is used to allocate some boot time memory for per_cpu variables
74 that aren't designed to grow/shrink as CPUs are made available or removed.
75 Once set during boot time discovery phase, the map is static, i.e no bits
76 are added or removed anytime. Trimming it accurately for your system needs
77 upfront can save some boot time memory. See below for how we use heuristics
78 in x86_64 case to keep this under check.
80 cpu_online_map: Bitmap of all CPUs currently online. Its set in __cpu_up()
81 after a cpu is available for kernel scheduling and ready to receive
82 interrupts from devices. Its cleared when a cpu is brought down using
83 __cpu_disable(), before which all OS services including interrupts are
84 migrated to another target CPU.
86 cpu_present_map: Bitmap of CPUs currently present in the system. Not all
87 of them may be online. When physical hotplug is processed by the relevant
88 subsystem (e.g ACPI) can change and new bit either be added or removed
89 from the map depending on the event is hot-add/hot-remove. There are currently
90 no locking rules as of now. Typical usage is to init topology during boot,
91 at which time hotplug is disabled.
93 You really dont need to manipulate any of the system cpu maps. They should
94 be read-only for most use. When setting up per-cpu resources almost always use
95 cpu_possible_map/for_each_cpu() to iterate.
97 Never use anything other than cpumask_t to represent bitmap of CPUs.
99 #include <linux/cpumask.h>
101 for_each_cpu - Iterate over cpu_possible_map
102 for_each_online_cpu - Iterate over cpu_online_map
103 for_each_present_cpu - Iterate over cpu_present_map
104 for_each_cpu_mask(x,mask) - Iterate over some random collection of cpu mask.
106 #include <linux/cpu.h>
107 lock_cpu_hotplug() and unlock_cpu_hotplug():
109 The above calls are used to inhibit cpu hotplug operations. While holding the
110 cpucontrol mutex, cpu_online_map will not change. If you merely need to avoid
111 cpus going away, you could also use preempt_disable() and preempt_enable()
112 for those sections. Just remember the critical section cannot call any
113 function that can sleep or schedule this process away. The preempt_disable()
114 will work as long as stop_machine_run() is used to take a cpu down.
116 CPU Hotplug - Frequently Asked Questions.
118 Q: How to i enable my kernel to support CPU hotplug?
119 A: When doing make defconfig, Enable CPU hotplug support
121 "Processor type and Features" -> Support for Hotpluggable CPUs
123 Make sure that you have CONFIG_HOTPLUG, and CONFIG_SMP turned on as well.
125 You would need to enable CONFIG_HOTPLUG_CPU for SMP suspend/resume support
128 Q: What architectures support CPU hotplug?
129 A: As of 2.6.14, the following architectures support CPU hotplug.
131 i386 (Intel), ppc, ppc64, parisc, s390, ia64 and x86_64
133 Q: How to test if hotplug is supported on the newly built kernel?
134 A: You should now notice an entry in sysfs.
136 Check if sysfs is mounted, using the "mount" command. You should notice
137 an entry as shown below in the output.
140 none on /sys type sysfs (rw)
143 if this is not mounted, do the following.
146 #mount -t sysfs sys /sys
148 now you should see entries for all present cpu, the following is an example
152 #/sys/devices/system/cpu
155 drwxr-xr-x 10 root root 0 Sep 19 07:44 .
156 drwxr-xr-x 13 root root 0 Sep 19 07:45 ..
157 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu0
158 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu1
159 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu2
160 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu3
161 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu4
162 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu5
163 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu6
164 drwxr-xr-x 3 root root 0 Sep 19 07:48 cpu7
166 Under each directory you would find an "online" file which is the control
167 file to logically online/offline a processor.
169 Q: Does hot-add/hot-remove refer to physical add/remove of cpus?
170 A: The usage of hot-add/remove may not be very consistently used in the code.
171 CONFIG_CPU_HOTPLUG enables logical online/offline capability in the kernel.
172 To support physical addition/removal, one would need some BIOS hooks and
173 the platform should have something like an attention button in PCI hotplug.
174 CONFIG_ACPI_HOTPLUG_CPU enables ACPI support for physical add/remove of CPUs.
176 Q: How do i logically offline a CPU?
179 #echo 0 > /sys/devices/system/cpu/cpuX/online
181 once the logical offline is successful, check
183 #cat /proc/interrupts
185 you should now not see the CPU that you removed. Also online file will report
186 the state as 0 when a cpu if offline and 1 when its online.
188 #To display the current cpu state.
189 #cat /sys/devices/system/cpu/cpuX/online
191 Q: Why cant i remove CPU0 on some systems?
192 A: Some architectures may have some special dependency on a certain CPU.
194 For e.g in IA64 platforms we have ability to sent platform interrupts to the
195 OS. a.k.a Corrected Platform Error Interrupts (CPEI). In current ACPI
196 specifications, we didn't have a way to change the target CPU. Hence if the
197 current ACPI version doesn't support such re-direction, we disable that CPU
198 by making it not-removable.
200 In such cases you will also notice that the online file is missing under cpu0.
202 Q: How do i find out if a particular CPU is not removable?
203 A: Depending on the implementation, some architectures may show this by the
204 absence of the "online" file. This is done if it can be determined ahead of
205 time that this CPU cannot be removed.
207 In some situations, this can be a run time check, i.e if you try to remove the
208 last CPU, this will not be permitted. You can find such failures by
209 investigating the return value of the "echo" command.
211 Q: What happens when a CPU is being logically offlined?
212 A: The following happen, listed in no particular order :-)
214 - A notification is sent to in-kernel registered modules by sending an event
216 - All process is migrated away from this outgoing CPU to a new CPU
217 - All interrupts targeted to this CPU is migrated to a new CPU
218 - timers/bottom half/task lets are also migrated to a new CPU
219 - Once all services are migrated, kernel calls an arch specific routine
220 __cpu_disable() to perform arch specific cleanup.
221 - Once this is successful, an event for successful cleanup is sent by an event
224 "It is expected that each service cleans up when the CPU_DOWN_PREPARE
225 notifier is called, when CPU_DEAD is called its expected there is nothing
226 running on behalf of this CPU that was offlined"
228 Q: If i have some kernel code that needs to be aware of CPU arrival and
229 departure, how to i arrange for proper notification?
230 A: This is what you would need in your kernel code to receive notifications.
232 #include <linux/cpu.h>
233 static int __cpuinit foobar_cpu_callback(struct notifier_block *nfb,
234 unsigned long action, void *hcpu)
236 unsigned int cpu = (unsigned long)hcpu;
240 foobar_online_action(cpu);
243 foobar_dead_action(cpu);
249 static struct notifier_block foobar_cpu_notifer =
251 .notifier_call = foobar_cpu_callback,
255 In your init function,
257 register_cpu_notifier(&foobar_cpu_notifier);
259 You can fail PREPARE notifiers if something doesn't work to prepare resources.
260 This will stop the activity and send a following CANCELED event back.
262 CPU_DEAD should not be failed, its just a goodness indication, but bad
263 things will happen if a notifier in path sent a BAD notify code.
265 Q: I don't see my action being called for all CPUs already up and running?
266 A: Yes, CPU notifiers are called only when new CPUs are on-lined or offlined.
267 If you need to perform some action for each cpu already in the system, then
269 for_each_online_cpu(i) {
270 foobar_cpu_callback(&foobar_cpu_notifier, CPU_UP_PREPARE, i);
271 foobar_cpu_callback(&foobar-cpu_notifier, CPU_ONLINE, i);
274 Q: If i would like to develop cpu hotplug support for a new architecture,
275 what do i need at a minimum?
276 A: The following are what is required for CPU hotplug infrastructure to work
279 - Make sure you have an entry in Kconfig to enable CONFIG_HOTPLUG_CPU
280 - __cpu_up() - Arch interface to bring up a CPU
281 - __cpu_disable() - Arch interface to shutdown a CPU, no more interrupts
282 can be handled by the kernel after the routine
283 returns. Including local APIC timers etc are
285 - __cpu_die() - This actually supposed to ensure death of the CPU.
286 Actually look at some example code in other arch
287 that implement CPU hotplug. The processor is taken
288 down from the idle() loop for that specific
289 architecture. __cpu_die() typically waits for some
290 per_cpu state to be set, to ensure the processor
291 dead routine is called to be sure positively.
293 Q: I need to ensure that a particular cpu is not removed when there is some
294 work specific to this cpu is in progress.
295 A: First switch the current thread context to preferred cpu
297 int my_func_on_cpu(int cpu)
299 cpumask_t saved_mask, new_mask = CPU_MASK_NONE;
300 int curr_cpu, err = 0;
302 saved_mask = current->cpus_allowed;
303 cpu_set(cpu, new_mask);
304 err = set_cpus_allowed(current, new_mask);
310 * If we got scheduled out just after the return from
311 * set_cpus_allowed() before running the work, this ensures
314 curr_cpu = get_cpu();
316 if (curr_cpu != cpu) {
321 * Do work : But cant sleep, since get_cpu() disables preempt
326 set_cpus_allowed(current, saved_mask);
331 Q: How do we determine how many CPUs are available for hotplug.
332 A: There is no clear spec defined way from ACPI that can give us that
333 information today. Based on some input from Natalie of Unisys,
334 that the ACPI MADT (Multiple APIC Description Tables) marks those possible
335 CPUs in a system with disabled status.
337 Andi implemented some simple heuristics that count the number of disabled
338 CPUs in MADT as hotpluggable CPUS. In the case there are no disabled CPUS
339 we assume 1/2 the number of CPUs currently present can be hotplugged.
341 Caveat: Today's ACPI MADT can only provide 256 entries since the apicid field
342 in MADT is only 8 bits.
344 User Space Notification
346 Hotplug support for devices is common in Linux today. Its being used today to
347 support automatic configuration of network, usb and pci devices. A hotplug
348 event can be used to invoke an agent script to perform the configuration task.
350 You can add /etc/hotplug/cpu.agent to handle hotplug notification user space
355 # Kernel hotplug params include:
356 #ACTION=%s [online or offline]
360 . ./hotplug.functions
364 echo `date` ":cpu.agent" add cpu >> /tmp/hotplug.txt
367 echo `date` ":cpu.agent" remove cpu >>/tmp/hotplug.txt
370 debug_mesg CPU $ACTION event not supported