1 CPU hotplug Support in Linux(tm) Kernel
5 Rusty Russell <rusty@rustcorp.com.au>
6 Srivatsa Vaddagiri <vatsa@in.ibm.com>
8 Zwane Mwaikambo <zwanem@gmail.com>
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_mask = cpu_present_mask + additional_cpus
52 cede_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
61 ia64 uses the number of disabled local apics in ACPI tables MADT to
62 determine the number of potentially hot-pluggable cpus. The implementation
63 should only rely on this to count the # of cpus, but *MUST* not rely
64 on the apicid values in those tables for disabled apics. In the event
65 BIOS doesn't mark such hot-pluggable cpus as disabled entries, one could
66 use this parameter "additional_cpus=x" to represent those cpus in the
69 possible_cpus=n [s390,x86_64] use this to set hotpluggable cpus.
70 This option sets possible_cpus bits in
71 cpu_possible_mask. Thus keeping the numbers of bits set
72 constant even if the machine gets rebooted.
76 [More on cpumaps and primitive to manipulate, please check
77 include/linux/cpumask.h that has more descriptive text.]
79 cpu_possible_mask: Bitmap of possible CPUs that can ever be available in the
80 system. This is used to allocate some boot time memory for per_cpu variables
81 that aren't designed to grow/shrink as CPUs are made available or removed.
82 Once set during boot time discovery phase, the map is static, i.e no bits
83 are added or removed anytime. Trimming it accurately for your system needs
84 upfront can save some boot time memory. See below for how we use heuristics
85 in x86_64 case to keep this under check.
87 cpu_online_mask: Bitmap of all CPUs currently online. Its set in __cpu_up()
88 after a cpu is available for kernel scheduling and ready to receive
89 interrupts from devices. Its cleared when a cpu is brought down using
90 __cpu_disable(), before which all OS services including interrupts are
91 migrated to another target CPU.
93 cpu_present_mask: Bitmap of CPUs currently present in the system. Not all
94 of them may be online. When physical hotplug is processed by the relevant
95 subsystem (e.g ACPI) can change and new bit either be added or removed
96 from the map depending on the event is hot-add/hot-remove. There are currently
97 no locking rules as of now. Typical usage is to init topology during boot,
98 at which time hotplug is disabled.
100 You really dont need to manipulate any of the system cpu maps. They should
101 be read-only for most use. When setting up per-cpu resources almost always use
102 cpu_possible_mask/for_each_possible_cpu() to iterate.
104 Never 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_mask
109 for_each_online_cpu - Iterate over cpu_online_mask
110 for_each_present_cpu - Iterate over cpu_present_mask
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():
116 The above calls are used to inhibit cpu hotplug operations. While the
117 cpu_hotplug.refcount is non zero, the cpu_online_mask will not change.
118 If you merely need to avoid cpus going away, you could also use
119 preempt_disable() and preempt_enable() for those sections.
120 Just remember the critical section cannot call any
121 function that can sleep or schedule this process away. The preempt_disable()
122 will work as long as stop_machine_run() is used to take a cpu down.
124 CPU Hotplug - Frequently Asked Questions.
126 Q: How to enable my kernel to support CPU hotplug?
127 A: When doing make defconfig, Enable CPU hotplug support
129 "Processor type and Features" -> Support for Hotpluggable CPUs
131 Make sure that you have CONFIG_SMP turned on as well.
133 You would need to enable CONFIG_HOTPLUG_CPU for SMP suspend/resume support
136 Q: What architectures support CPU hotplug?
137 A: As of 2.6.14, the following architectures support CPU hotplug.
139 i386 (Intel), ppc, ppc64, parisc, s390, ia64 and x86_64
141 Q: How to test if hotplug is supported on the newly built kernel?
142 A: You should now notice an entry in sysfs.
144 Check if sysfs is mounted, using the "mount" command. You should notice
145 an entry as shown below in the output.
148 none on /sys type sysfs (rw)
151 If this is not mounted, do the following.
154 #mount -t sysfs sys /sys
156 Now you should see entries for all present cpu, the following is an example
160 #/sys/devices/system/cpu
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
174 Under each directory you would find an "online" file which is the control
175 file to logically online/offline a processor.
177 Q: Does hot-add/hot-remove refer to physical add/remove of cpus?
178 A: The usage of hot-add/remove may not be very consistently used in the code.
179 CONFIG_HOTPLUG_CPU enables logical online/offline capability in the kernel.
180 To support physical addition/removal, one would need some BIOS hooks and
181 the platform should have something like an attention button in PCI hotplug.
182 CONFIG_ACPI_HOTPLUG_CPU enables ACPI support for physical add/remove of CPUs.
184 Q: How do i logically offline a CPU?
187 #echo 0 > /sys/devices/system/cpu/cpuX/online
189 Once the logical offline is successful, check
191 #cat /proc/interrupts
193 You should now not see the CPU that you removed. Also online file will report
194 the 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
199 Q: Why can't i remove CPU0 on some systems?
200 A: Some architectures may have some special dependency on a certain CPU.
202 For e.g in IA64 platforms we have ability to sent platform interrupts to the
203 OS. a.k.a Corrected Platform Error Interrupts (CPEI). In current ACPI
204 specifications, we didn't have a way to change the target CPU. Hence if the
205 current ACPI version doesn't support such re-direction, we disable that CPU
206 by making it not-removable.
208 In such cases you will also notice that the online file is missing under cpu0.
210 Q: Is CPU0 removable on X86?
211 A: Yes. If kernel is compiled with CONFIG_BOOTPARAM_HOTPLUG_CPU0=y, CPU0 is
212 removable by default. Otherwise, CPU0 is also removable by kernel option
215 But some features depend on CPU0. Two known dependencies are:
217 1. Resume from hibernate/suspend depends on CPU0. Hibernate/suspend will fail if
218 CPU0 is offline and you need to online CPU0 before hibernate/suspend can
220 2. PIC interrupts also depend on CPU0. CPU0 can't be removed if a PIC interrupt
223 It's said poweroff/reboot may depend on CPU0 on some machines although I haven't
224 seen any poweroff/reboot failure so far after CPU0 is offline on a few tested
227 Please let me know if you know or see any other dependencies of CPU0.
229 If the dependencies are under your control, you can turn on CPU0 hotplug feature
230 either by CONFIG_BOOTPARAM_HOTPLUG_CPU0 or by kernel parameter cpu0_hotplug.
232 --Fenghua Yu <fenghua.yu@intel.com>
234 Q: How do i find out if a particular CPU is not removable?
235 A: Depending on the implementation, some architectures may show this by the
236 absence of the "online" file. This is done if it can be determined ahead of
237 time that this CPU cannot be removed.
239 In some situations, this can be a run time check, i.e if you try to remove the
240 last CPU, this will not be permitted. You can find such failures by
241 investigating the return value of the "echo" command.
243 Q: What happens when a CPU is being logically offlined?
244 A: The following happen, listed in no particular order :-)
246 - A notification is sent to in-kernel registered modules by sending an event
247 CPU_DOWN_PREPARE or CPU_DOWN_PREPARE_FROZEN, depending on whether or not the
248 CPU is being offlined while tasks are frozen due to a suspend operation in
250 - All processes are migrated away from this outgoing CPU to new CPUs.
251 The new CPU is chosen from each process' current cpuset, which may be
252 a subset of all online CPUs.
253 - All interrupts targeted to this CPU is migrated to a new CPU
254 - timers/bottom half/task lets are also migrated to a new CPU
255 - Once all services are migrated, kernel calls an arch specific routine
256 __cpu_disable() to perform arch specific cleanup.
257 - Once this is successful, an event for successful cleanup is sent by an event
258 CPU_DEAD (or CPU_DEAD_FROZEN if tasks are frozen due to a suspend while the
259 CPU is being offlined).
261 "It is expected that each service cleans up when the CPU_DOWN_PREPARE
262 notifier is called, when CPU_DEAD is called its expected there is nothing
263 running on behalf of this CPU that was offlined"
265 Q: If i have some kernel code that needs to be aware of CPU arrival and
266 departure, how to i arrange for proper notification?
267 A: This is what you would need in your kernel code to receive notifications.
269 #include <linux/cpu.h>
270 static int foobar_cpu_callback(struct notifier_block *nfb,
271 unsigned long action, void *hcpu)
273 unsigned int cpu = (unsigned long)hcpu;
277 case CPU_ONLINE_FROZEN:
278 foobar_online_action(cpu);
281 case CPU_DEAD_FROZEN:
282 foobar_dead_action(cpu);
288 static struct notifier_block foobar_cpu_notifier =
290 .notifier_call = foobar_cpu_callback,
293 You need to call register_cpu_notifier() from your init function.
294 Init functions could be of two types:
295 1. early init (init function called when only the boot processor is online).
296 2. late init (init function called _after_ all the CPUs are online).
298 For the first case, you should add the following to your init function
300 register_cpu_notifier(&foobar_cpu_notifier);
302 For the second case, you should add the following to your init function
304 register_hotcpu_notifier(&foobar_cpu_notifier);
306 You can fail PREPARE notifiers if something doesn't work to prepare resources.
307 This will stop the activity and send a following CANCELED event back.
309 CPU_DEAD should not be failed, its just a goodness indication, but bad
310 things will happen if a notifier in path sent a BAD notify code.
312 Q: I don't see my action being called for all CPUs already up and running?
313 A: Yes, CPU notifiers are called only when new CPUs are on-lined or offlined.
314 If you need to perform some action for each cpu already in the system, then
316 for_each_online_cpu(i) {
317 foobar_cpu_callback(&foobar_cpu_notifier, CPU_UP_PREPARE, i);
318 foobar_cpu_callback(&foobar_cpu_notifier, CPU_ONLINE, i);
321 Q: If i would like to develop cpu hotplug support for a new architecture,
322 what do i need at a minimum?
323 A: The following are what is required for CPU hotplug infrastructure to work
326 - Make sure you have an entry in Kconfig to enable CONFIG_HOTPLUG_CPU
327 - __cpu_up() - Arch interface to bring up a CPU
328 - __cpu_disable() - Arch interface to shutdown a CPU, no more interrupts
329 can be handled by the kernel after the routine
330 returns. Including local APIC timers etc are
332 - __cpu_die() - This actually supposed to ensure death of the CPU.
333 Actually look at some example code in other arch
334 that implement CPU hotplug. The processor is taken
335 down from the idle() loop for that specific
336 architecture. __cpu_die() typically waits for some
337 per_cpu state to be set, to ensure the processor
338 dead routine is called to be sure positively.
340 Q: I need to ensure that a particular cpu is not removed when there is some
341 work specific to this cpu is in progress.
342 A: There are two ways. If your code can be run in interrupt context, use
343 smp_call_function_single(), otherwise use work_on_cpu(). Note that
344 work_on_cpu() is slow, and can fail due to out of memory:
346 int my_func_on_cpu(int cpu)
350 if (!cpu_online(cpu))
354 err = work_on_cpu(cpu, __my_func_on_cpu, NULL);
356 smp_call_function_single(cpu, __my_func_on_cpu, &err,
363 Q: How do we determine how many CPUs are available for hotplug.
364 A: There is no clear spec defined way from ACPI that can give us that
365 information today. Based on some input from Natalie of Unisys,
366 that the ACPI MADT (Multiple APIC Description Tables) marks those possible
367 CPUs in a system with disabled status.
369 Andi implemented some simple heuristics that count the number of disabled
370 CPUs in MADT as hotpluggable CPUS. In the case there are no disabled CPUS
371 we assume 1/2 the number of CPUs currently present can be hotplugged.
373 Caveat: ACPI MADT can only provide 256 entries in systems with only ACPI 2.0c
374 or earlier ACPI version supported, because the apicid field in MADT is only
375 8 bits. From ACPI 3.0, this limitation was removed since the apicid field
376 was extended to 32 bits with x2APIC introduced.
378 User Space Notification
380 Hotplug support for devices is common in Linux today. Its being used today to
381 support automatic configuration of network, usb and pci devices. A hotplug
382 event can be used to invoke an agent script to perform the configuration task.
384 You can add /etc/hotplug/cpu.agent to handle hotplug notification user space
389 # Kernel hotplug params include:
390 #ACTION=%s [online or offline]
394 . ./hotplug.functions
398 echo `date` ":cpu.agent" add cpu >> /tmp/hotplug.txt
401 echo `date` ":cpu.agent" remove cpu >>/tmp/hotplug.txt
404 debug_mesg CPU $ACTION event not supported