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 (*) Use this to limit hotpluggable cpus. This option sets
50 cpu_possible_map = cpu_present_map + additional_cpus
52 (*) Option valid only for following architectures
55 ia64 and x86_64 use the number of disabled local apics in ACPI tables MADT
56 to determine the number of potentially hot-pluggable cpus. The implementation
57 should only rely on this to count the # of cpus, but *MUST* not rely on the
58 apicid values in those tables for disabled apics. In the event BIOS doesn't
59 mark such hot-pluggable cpus as disabled entries, one could use this
60 parameter "additional_cpus=x" to represent those cpus in the cpu_possible_map.
62 s390 uses the number of cpus it detects at IPL time to also the number of bits
63 in cpu_possible_map. If it is desired to add additional cpus at a later time
64 the number should be specified using this option or the possible_cpus option.
66 possible_cpus=n [s390 only] use this to set hotpluggable cpus.
67 This option sets possible_cpus bits in
68 cpu_possible_map. Thus keeping the numbers of bits set
69 constant even if the machine gets rebooted.
70 This option overrides additional_cpus.
74 [More on cpumaps and primitive to manipulate, please check
75 include/linux/cpumask.h that has more descriptive text.]
77 cpu_possible_map: Bitmap of possible CPUs that can ever be available in the
78 system. This is used to allocate some boot time memory for per_cpu variables
79 that aren't designed to grow/shrink as CPUs are made available or removed.
80 Once set during boot time discovery phase, the map is static, i.e no bits
81 are added or removed anytime. Trimming it accurately for your system needs
82 upfront can save some boot time memory. See below for how we use heuristics
83 in x86_64 case to keep this under check.
85 cpu_online_map: Bitmap of all CPUs currently online. Its set in __cpu_up()
86 after a cpu is available for kernel scheduling and ready to receive
87 interrupts from devices. Its cleared when a cpu is brought down using
88 __cpu_disable(), before which all OS services including interrupts are
89 migrated to another target CPU.
91 cpu_present_map: Bitmap of CPUs currently present in the system. Not all
92 of them may be online. When physical hotplug is processed by the relevant
93 subsystem (e.g ACPI) can change and new bit either be added or removed
94 from the map depending on the event is hot-add/hot-remove. There are currently
95 no locking rules as of now. Typical usage is to init topology during boot,
96 at which time hotplug is disabled.
98 You really dont need to manipulate any of the system cpu maps. They should
99 be read-only for most use. When setting up per-cpu resources almost always use
100 cpu_possible_map/for_each_possible_cpu() to iterate.
102 Never use anything other than cpumask_t to represent bitmap of CPUs.
104 #include <linux/cpumask.h>
106 for_each_possible_cpu - Iterate over cpu_possible_map
107 for_each_online_cpu - Iterate over cpu_online_map
108 for_each_present_cpu - Iterate over cpu_present_map
109 for_each_cpu_mask(x,mask) - Iterate over some random collection of cpu mask.
111 #include <linux/cpu.h>
112 get_online_cpus() and put_online_cpus():
114 The above calls are used to inhibit cpu hotplug operations. While the
115 cpu_hotplug.refcount is non zero, the cpu_online_map will not change.
116 If you merely need to avoid cpus going away, you could also use
117 preempt_disable() and preempt_enable() for those sections.
118 Just remember the critical section cannot call any
119 function that can sleep or schedule this process away. The preempt_disable()
120 will work as long as stop_machine_run() is used to take a cpu down.
122 CPU Hotplug - Frequently Asked Questions.
124 Q: How to enable my kernel to support CPU hotplug?
125 A: When doing make defconfig, Enable CPU hotplug support
127 "Processor type and Features" -> Support for Hotpluggable CPUs
129 Make sure that you have CONFIG_HOTPLUG, and CONFIG_SMP turned on as well.
131 You would need to enable CONFIG_HOTPLUG_CPU for SMP suspend/resume support
134 Q: What architectures support CPU hotplug?
135 A: As of 2.6.14, the following architectures support CPU hotplug.
137 i386 (Intel), ppc, ppc64, parisc, s390, ia64 and x86_64
139 Q: How to test if hotplug is supported on the newly built kernel?
140 A: You should now notice an entry in sysfs.
142 Check if sysfs is mounted, using the "mount" command. You should notice
143 an entry as shown below in the output.
146 none on /sys type sysfs (rw)
149 If this is not mounted, do the following.
152 #mount -t sysfs sys /sys
154 Now you should see entries for all present cpu, the following is an example
158 #/sys/devices/system/cpu
161 drwxr-xr-x 10 root root 0 Sep 19 07:44 .
162 drwxr-xr-x 13 root root 0 Sep 19 07:45 ..
163 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu0
164 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu1
165 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu2
166 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu3
167 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu4
168 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu5
169 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu6
170 drwxr-xr-x 3 root root 0 Sep 19 07:48 cpu7
172 Under each directory you would find an "online" file which is the control
173 file to logically online/offline a processor.
175 Q: Does hot-add/hot-remove refer to physical add/remove of cpus?
176 A: The usage of hot-add/remove may not be very consistently used in the code.
177 CONFIG_HOTPLUG_CPU enables logical online/offline capability in the kernel.
178 To support physical addition/removal, one would need some BIOS hooks and
179 the platform should have something like an attention button in PCI hotplug.
180 CONFIG_ACPI_HOTPLUG_CPU enables ACPI support for physical add/remove of CPUs.
182 Q: How do i logically offline a CPU?
185 #echo 0 > /sys/devices/system/cpu/cpuX/online
187 Once the logical offline is successful, check
189 #cat /proc/interrupts
191 You should now not see the CPU that you removed. Also online file will report
192 the state as 0 when a cpu if offline and 1 when its online.
194 #To display the current cpu state.
195 #cat /sys/devices/system/cpu/cpuX/online
197 Q: Why cant i remove CPU0 on some systems?
198 A: Some architectures may have some special dependency on a certain CPU.
200 For e.g in IA64 platforms we have ability to sent platform interrupts to the
201 OS. a.k.a Corrected Platform Error Interrupts (CPEI). In current ACPI
202 specifications, we didn't have a way to change the target CPU. Hence if the
203 current ACPI version doesn't support such re-direction, we disable that CPU
204 by making it not-removable.
206 In such cases you will also notice that the online file is missing under cpu0.
208 Q: How do i find out if a particular CPU is not removable?
209 A: Depending on the implementation, some architectures may show this by the
210 absence of the "online" file. This is done if it can be determined ahead of
211 time that this CPU cannot be removed.
213 In some situations, this can be a run time check, i.e if you try to remove the
214 last CPU, this will not be permitted. You can find such failures by
215 investigating the return value of the "echo" command.
217 Q: What happens when a CPU is being logically offlined?
218 A: The following happen, listed in no particular order :-)
220 - A notification is sent to in-kernel registered modules by sending an event
221 CPU_DOWN_PREPARE or CPU_DOWN_PREPARE_FROZEN, depending on whether or not the
222 CPU is being offlined while tasks are frozen due to a suspend operation in
224 - All processes are migrated away from this outgoing CPU to new CPUs.
225 The new CPU is chosen from each process' current cpuset, which may be
226 a subset of all online CPUs.
227 - All interrupts targeted to this CPU is migrated to a new CPU
228 - timers/bottom half/task lets are also migrated to a new CPU
229 - Once all services are migrated, kernel calls an arch specific routine
230 __cpu_disable() to perform arch specific cleanup.
231 - Once this is successful, an event for successful cleanup is sent by an event
232 CPU_DEAD (or CPU_DEAD_FROZEN if tasks are frozen due to a suspend while the
233 CPU is being offlined).
235 "It is expected that each service cleans up when the CPU_DOWN_PREPARE
236 notifier is called, when CPU_DEAD is called its expected there is nothing
237 running on behalf of this CPU that was offlined"
239 Q: If i have some kernel code that needs to be aware of CPU arrival and
240 departure, how to i arrange for proper notification?
241 A: This is what you would need in your kernel code to receive notifications.
243 #include <linux/cpu.h>
244 static int __cpuinit foobar_cpu_callback(struct notifier_block *nfb,
245 unsigned long action, void *hcpu)
247 unsigned int cpu = (unsigned long)hcpu;
251 case CPU_ONLINE_FROZEN:
252 foobar_online_action(cpu);
255 case CPU_DEAD_FROZEN:
256 foobar_dead_action(cpu);
262 static struct notifier_block __cpuinitdata foobar_cpu_notifer =
264 .notifier_call = foobar_cpu_callback,
267 You need to call register_cpu_notifier() from your init function.
268 Init functions could be of two types:
269 1. early init (init function called when only the boot processor is online).
270 2. late init (init function called _after_ all the CPUs are online).
272 For the first case, you should add the following to your init function
274 register_cpu_notifier(&foobar_cpu_notifier);
276 For the second case, you should add the following to your init function
278 register_hotcpu_notifier(&foobar_cpu_notifier);
280 You can fail PREPARE notifiers if something doesn't work to prepare resources.
281 This will stop the activity and send a following CANCELED event back.
283 CPU_DEAD should not be failed, its just a goodness indication, but bad
284 things will happen if a notifier in path sent a BAD notify code.
286 Q: I don't see my action being called for all CPUs already up and running?
287 A: Yes, CPU notifiers are called only when new CPUs are on-lined or offlined.
288 If you need to perform some action for each cpu already in the system, then
290 for_each_online_cpu(i) {
291 foobar_cpu_callback(&foobar_cpu_notifier, CPU_UP_PREPARE, i);
292 foobar_cpu_callback(&foobar_cpu_notifier, CPU_ONLINE, i);
295 Q: If i would like to develop cpu hotplug support for a new architecture,
296 what do i need at a minimum?
297 A: The following are what is required for CPU hotplug infrastructure to work
300 - Make sure you have an entry in Kconfig to enable CONFIG_HOTPLUG_CPU
301 - __cpu_up() - Arch interface to bring up a CPU
302 - __cpu_disable() - Arch interface to shutdown a CPU, no more interrupts
303 can be handled by the kernel after the routine
304 returns. Including local APIC timers etc are
306 - __cpu_die() - This actually supposed to ensure death of the CPU.
307 Actually look at some example code in other arch
308 that implement CPU hotplug. The processor is taken
309 down from the idle() loop for that specific
310 architecture. __cpu_die() typically waits for some
311 per_cpu state to be set, to ensure the processor
312 dead routine is called to be sure positively.
314 Q: I need to ensure that a particular cpu is not removed when there is some
315 work specific to this cpu is in progress.
316 A: First switch the current thread context to preferred cpu
318 int my_func_on_cpu(int cpu)
320 cpumask_t saved_mask, new_mask = CPU_MASK_NONE;
321 int curr_cpu, err = 0;
323 saved_mask = current->cpus_allowed;
324 cpu_set(cpu, new_mask);
325 err = set_cpus_allowed(current, new_mask);
331 * If we got scheduled out just after the return from
332 * set_cpus_allowed() before running the work, this ensures
335 curr_cpu = get_cpu();
337 if (curr_cpu != cpu) {
342 * Do work : But cant sleep, since get_cpu() disables preempt
347 set_cpus_allowed(current, saved_mask);
352 Q: How do we determine how many CPUs are available for hotplug.
353 A: There is no clear spec defined way from ACPI that can give us that
354 information today. Based on some input from Natalie of Unisys,
355 that the ACPI MADT (Multiple APIC Description Tables) marks those possible
356 CPUs in a system with disabled status.
358 Andi implemented some simple heuristics that count the number of disabled
359 CPUs in MADT as hotpluggable CPUS. In the case there are no disabled CPUS
360 we assume 1/2 the number of CPUs currently present can be hotplugged.
362 Caveat: Today's ACPI MADT can only provide 256 entries since the apicid field
363 in MADT is only 8 bits.
365 User Space Notification
367 Hotplug support for devices is common in Linux today. Its being used today to
368 support automatic configuration of network, usb and pci devices. A hotplug
369 event can be used to invoke an agent script to perform the configuration task.
371 You can add /etc/hotplug/cpu.agent to handle hotplug notification user space
376 # Kernel hotplug params include:
377 #ACTION=%s [online or offline]
381 . ./hotplug.functions
385 echo `date` ":cpu.agent" add cpu >> /tmp/hotplug.txt
388 echo `date` ":cpu.agent" remove cpu >>/tmp/hotplug.txt
391 debug_mesg CPU $ACTION event not supported