blob: 94bbc27ddd4fbf50ff361177450c5f524210f1cb [file] [log] [blame]
Ashok Rajc8094062006-01-08 01:03:17 -08001 CPU hotplug Support in Linux(tm) Kernel
2
3 Maintainers:
4 CPU Hotplug Core:
5 Rusty Russell <rusty@rustycorp.com.au>
6 Srivatsa Vaddagiri <vatsa@in.ibm.com>
7 i386:
8 Zwane Mwaikambo <zwane@arm.linux.org.uk>
9 ppc64:
10 Nathan Lynch <nathanl@austin.ibm.com>
11 Joel Schopp <jschopp@austin.ibm.com>
12 ia64/x86_64:
13 Ashok Raj <ashok.raj@intel.com>
Heiko Carstens255acee2006-02-17 13:52:46 -080014 s390:
15 Heiko Carstens <heiko.carstens@de.ibm.com>
Ashok Rajc8094062006-01-08 01:03:17 -080016
17Authors: Ashok Raj <ashok.raj@intel.com>
18Lots of feedback: Nathan Lynch <nathanl@austin.ibm.com>,
19 Joel Schopp <jschopp@austin.ibm.com>
20
21Introduction
22
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.
29
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.
34
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.
39
40General Stuff about CPU Hotplug
41--------------------------------
42
43Command Line Switches
44---------------------
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.
48
Satoru Takeuchica926e82006-10-19 23:29:06 -070049additional_cpus=n (*) Use this to limit hotpluggable cpus. This option sets
Heiko Carstens255acee2006-02-17 13:52:46 -080050 cpu_possible_map = cpu_present_map + additional_cpus
Ashok Raj8f8b1132006-02-16 14:01:48 -080051
Heiko Carstens6303dbf2006-02-20 18:27:58 -080052(*) Option valid only for following architectures
Heiko Carstens48483b32008-01-26 14:11:05 +010053- x86_64, ia64
Heiko Carstens6303dbf2006-02-20 18:27:58 -080054
Ashok Raj8f8b1132006-02-16 14:01:48 -080055ia64 and x86_64 use the number of disabled local apics in ACPI tables MADT
56to determine the number of potentially hot-pluggable cpus. The implementation
Matt LaPlante5d3f0832006-11-30 05:21:10 +010057should only rely on this to count the # of cpus, but *MUST* not rely on the
58apicid values in those tables for disabled apics. In the event BIOS doesn't
Ashok Raj8f8b1132006-02-16 14:01:48 -080059mark such hot-pluggable cpus as disabled entries, one could use this
60parameter "additional_cpus=x" to represent those cpus in the cpu_possible_map.
61
Heiko Carstens37a33022006-02-17 13:52:47 -080062possible_cpus=n [s390 only] use this to set hotpluggable cpus.
63 This option sets possible_cpus bits in
64 cpu_possible_map. Thus keeping the numbers of bits set
65 constant even if the machine gets rebooted.
Heiko Carstens37a33022006-02-17 13:52:47 -080066
Ashok Rajc8094062006-01-08 01:03:17 -080067CPU maps and such
68-----------------
69[More on cpumaps and primitive to manipulate, please check
70include/linux/cpumask.h that has more descriptive text.]
71
72cpu_possible_map: Bitmap of possible CPUs that can ever be available in the
73system. This is used to allocate some boot time memory for per_cpu variables
74that aren't designed to grow/shrink as CPUs are made available or removed.
75Once set during boot time discovery phase, the map is static, i.e no bits
76are added or removed anytime. Trimming it accurately for your system needs
77upfront can save some boot time memory. See below for how we use heuristics
78in x86_64 case to keep this under check.
79
80cpu_online_map: Bitmap of all CPUs currently online. Its set in __cpu_up()
81after a cpu is available for kernel scheduling and ready to receive
82interrupts from devices. Its cleared when a cpu is brought down using
83__cpu_disable(), before which all OS services including interrupts are
84migrated to another target CPU.
85
86cpu_present_map: Bitmap of CPUs currently present in the system. Not all
87of them may be online. When physical hotplug is processed by the relevant
88subsystem (e.g ACPI) can change and new bit either be added or removed
89from the map depending on the event is hot-add/hot-remove. There are currently
90no locking rules as of now. Typical usage is to init topology during boot,
91at which time hotplug is disabled.
92
93You really dont need to manipulate any of the system cpu maps. They should
94be read-only for most use. When setting up per-cpu resources almost always use
KAMEZAWA Hiroyuki3c30a752006-03-28 01:56:39 -080095cpu_possible_map/for_each_possible_cpu() to iterate.
Ashok Rajc8094062006-01-08 01:03:17 -080096
97Never use anything other than cpumask_t to represent bitmap of CPUs.
98
Satoru Takeuchica926e82006-10-19 23:29:06 -070099 #include <linux/cpumask.h>
Ashok Rajc8094062006-01-08 01:03:17 -0800100
Satoru Takeuchica926e82006-10-19 23:29:06 -0700101 for_each_possible_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.
Ashok Rajc8094062006-01-08 01:03:17 -0800105
Satoru Takeuchica926e82006-10-19 23:29:06 -0700106 #include <linux/cpu.h>
Gautham R Shenoy86ef5c92008-01-25 21:08:02 +0100107 get_online_cpus() and put_online_cpus():
Ashok Rajc8094062006-01-08 01:03:17 -0800108
Gautham R Shenoy86ef5c92008-01-25 21:08:02 +0100109The above calls are used to inhibit cpu hotplug operations. While the
110cpu_hotplug.refcount is non zero, the cpu_online_map will not change.
111If you merely need to avoid cpus going away, you could also use
112preempt_disable() and preempt_enable() for those sections.
113Just remember the critical section cannot call any
Ashok Rajc8094062006-01-08 01:03:17 -0800114function that can sleep or schedule this process away. The preempt_disable()
115will work as long as stop_machine_run() is used to take a cpu down.
116
117CPU Hotplug - Frequently Asked Questions.
118
Satoru Takeuchica926e82006-10-19 23:29:06 -0700119Q: How to enable my kernel to support CPU hotplug?
Ashok Rajc8094062006-01-08 01:03:17 -0800120A: When doing make defconfig, Enable CPU hotplug support
121
122 "Processor type and Features" -> Support for Hotpluggable CPUs
123
124Make sure that you have CONFIG_HOTPLUG, and CONFIG_SMP turned on as well.
125
126You would need to enable CONFIG_HOTPLUG_CPU for SMP suspend/resume support
127as well.
128
129Q: What architectures support CPU hotplug?
130A: As of 2.6.14, the following architectures support CPU hotplug.
131
132i386 (Intel), ppc, ppc64, parisc, s390, ia64 and x86_64
133
134Q: How to test if hotplug is supported on the newly built kernel?
135A: You should now notice an entry in sysfs.
136
137Check if sysfs is mounted, using the "mount" command. You should notice
138an entry as shown below in the output.
139
Satoru Takeuchica926e82006-10-19 23:29:06 -0700140 ....
141 none on /sys type sysfs (rw)
142 ....
Ashok Rajc8094062006-01-08 01:03:17 -0800143
Satoru Takeuchica926e82006-10-19 23:29:06 -0700144If this is not mounted, do the following.
Ashok Rajc8094062006-01-08 01:03:17 -0800145
Satoru Takeuchica926e82006-10-19 23:29:06 -0700146 #mkdir /sysfs
147 #mount -t sysfs sys /sys
Ashok Rajc8094062006-01-08 01:03:17 -0800148
Satoru Takeuchica926e82006-10-19 23:29:06 -0700149Now you should see entries for all present cpu, the following is an example
Ashok Rajc8094062006-01-08 01:03:17 -0800150in a 8-way system.
151
Satoru Takeuchica926e82006-10-19 23:29:06 -0700152 #pwd
153 #/sys/devices/system/cpu
154 #ls -l
155 total 0
156 drwxr-xr-x 10 root root 0 Sep 19 07:44 .
157 drwxr-xr-x 13 root root 0 Sep 19 07:45 ..
158 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu0
159 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu1
160 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu2
161 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu3
162 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu4
163 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu5
164 drwxr-xr-x 3 root root 0 Sep 19 07:44 cpu6
165 drwxr-xr-x 3 root root 0 Sep 19 07:48 cpu7
Ashok Rajc8094062006-01-08 01:03:17 -0800166
167Under each directory you would find an "online" file which is the control
168file to logically online/offline a processor.
169
170Q: Does hot-add/hot-remove refer to physical add/remove of cpus?
171A: The usage of hot-add/remove may not be very consistently used in the code.
Satoru Takeuchica926e82006-10-19 23:29:06 -0700172CONFIG_HOTPLUG_CPU enables logical online/offline capability in the kernel.
Ashok Rajc8094062006-01-08 01:03:17 -0800173To support physical addition/removal, one would need some BIOS hooks and
174the platform should have something like an attention button in PCI hotplug.
175CONFIG_ACPI_HOTPLUG_CPU enables ACPI support for physical add/remove of CPUs.
176
177Q: How do i logically offline a CPU?
178A: Do the following.
179
Satoru Takeuchica926e82006-10-19 23:29:06 -0700180 #echo 0 > /sys/devices/system/cpu/cpuX/online
Ashok Rajc8094062006-01-08 01:03:17 -0800181
Satoru Takeuchica926e82006-10-19 23:29:06 -0700182Once the logical offline is successful, check
Ashok Rajc8094062006-01-08 01:03:17 -0800183
Satoru Takeuchica926e82006-10-19 23:29:06 -0700184 #cat /proc/interrupts
Ashok Rajc8094062006-01-08 01:03:17 -0800185
Satoru Takeuchica926e82006-10-19 23:29:06 -0700186You should now not see the CPU that you removed. Also online file will report
Ashok Rajc8094062006-01-08 01:03:17 -0800187the state as 0 when a cpu if offline and 1 when its online.
188
Satoru Takeuchica926e82006-10-19 23:29:06 -0700189 #To display the current cpu state.
190 #cat /sys/devices/system/cpu/cpuX/online
Ashok Rajc8094062006-01-08 01:03:17 -0800191
192Q: Why cant i remove CPU0 on some systems?
193A: Some architectures may have some special dependency on a certain CPU.
194
195For e.g in IA64 platforms we have ability to sent platform interrupts to the
196OS. a.k.a Corrected Platform Error Interrupts (CPEI). In current ACPI
197specifications, we didn't have a way to change the target CPU. Hence if the
198current ACPI version doesn't support such re-direction, we disable that CPU
199by making it not-removable.
200
201In such cases you will also notice that the online file is missing under cpu0.
202
203Q: How do i find out if a particular CPU is not removable?
204A: Depending on the implementation, some architectures may show this by the
205absence of the "online" file. This is done if it can be determined ahead of
206time that this CPU cannot be removed.
207
208In some situations, this can be a run time check, i.e if you try to remove the
209last CPU, this will not be permitted. You can find such failures by
210investigating the return value of the "echo" command.
211
212Q: What happens when a CPU is being logically offlined?
213A: The following happen, listed in no particular order :-)
214
215- A notification is sent to in-kernel registered modules by sending an event
Rafael J. Wysocki8bb78442007-05-09 02:35:10 -0700216 CPU_DOWN_PREPARE or CPU_DOWN_PREPARE_FROZEN, depending on whether or not the
217 CPU is being offlined while tasks are frozen due to a suspend operation in
218 progress
Cliff Wickman470fd642007-10-18 23:40:46 -0700219- All processes are migrated away from this outgoing CPU to new CPUs.
220 The new CPU is chosen from each process' current cpuset, which may be
221 a subset of all online CPUs.
Ashok Rajc8094062006-01-08 01:03:17 -0800222- All interrupts targeted to this CPU is migrated to a new CPU
223- timers/bottom half/task lets are also migrated to a new CPU
224- Once all services are migrated, kernel calls an arch specific routine
225 __cpu_disable() to perform arch specific cleanup.
226- Once this is successful, an event for successful cleanup is sent by an event
Rafael J. Wysocki8bb78442007-05-09 02:35:10 -0700227 CPU_DEAD (or CPU_DEAD_FROZEN if tasks are frozen due to a suspend while the
228 CPU is being offlined).
Ashok Rajc8094062006-01-08 01:03:17 -0800229
230 "It is expected that each service cleans up when the CPU_DOWN_PREPARE
231 notifier is called, when CPU_DEAD is called its expected there is nothing
232 running on behalf of this CPU that was offlined"
233
234Q: If i have some kernel code that needs to be aware of CPU arrival and
235 departure, how to i arrange for proper notification?
236A: This is what you would need in your kernel code to receive notifications.
237
Satoru Takeuchica926e82006-10-19 23:29:06 -0700238 #include <linux/cpu.h>
239 static int __cpuinit foobar_cpu_callback(struct notifier_block *nfb,
Ashok Rajc8094062006-01-08 01:03:17 -0800240 unsigned long action, void *hcpu)
241 {
242 unsigned int cpu = (unsigned long)hcpu;
243
244 switch (action) {
245 case CPU_ONLINE:
Rafael J. Wysocki8bb78442007-05-09 02:35:10 -0700246 case CPU_ONLINE_FROZEN:
Ashok Rajc8094062006-01-08 01:03:17 -0800247 foobar_online_action(cpu);
248 break;
249 case CPU_DEAD:
Rafael J. Wysocki8bb78442007-05-09 02:35:10 -0700250 case CPU_DEAD_FROZEN:
Ashok Rajc8094062006-01-08 01:03:17 -0800251 foobar_dead_action(cpu);
252 break;
253 }
254 return NOTIFY_OK;
255 }
256
Chandra Seetharaman7c7165c2006-07-30 03:03:36 -0700257 static struct notifier_block __cpuinitdata foobar_cpu_notifer =
Ashok Rajc8094062006-01-08 01:03:17 -0800258 {
259 .notifier_call = foobar_cpu_callback,
260 };
261
Chandra Seetharaman7c7165c2006-07-30 03:03:36 -0700262You need to call register_cpu_notifier() from your init function.
263Init functions could be of two types:
2641. early init (init function called when only the boot processor is online).
2652. late init (init function called _after_ all the CPUs are online).
Ashok Rajc8094062006-01-08 01:03:17 -0800266
Chandra Seetharaman7c7165c2006-07-30 03:03:36 -0700267For the first case, you should add the following to your init function
Ashok Rajc8094062006-01-08 01:03:17 -0800268
269 register_cpu_notifier(&foobar_cpu_notifier);
270
Chandra Seetharaman7c7165c2006-07-30 03:03:36 -0700271For the second case, you should add the following to your init function
272
273 register_hotcpu_notifier(&foobar_cpu_notifier);
274
Ashok Rajc8094062006-01-08 01:03:17 -0800275You can fail PREPARE notifiers if something doesn't work to prepare resources.
276This will stop the activity and send a following CANCELED event back.
277
278CPU_DEAD should not be failed, its just a goodness indication, but bad
279things will happen if a notifier in path sent a BAD notify code.
280
281Q: I don't see my action being called for all CPUs already up and running?
282A: Yes, CPU notifiers are called only when new CPUs are on-lined or offlined.
283 If you need to perform some action for each cpu already in the system, then
284
Satoru Takeuchica926e82006-10-19 23:29:06 -0700285 for_each_online_cpu(i) {
Ashok Rajc8094062006-01-08 01:03:17 -0800286 foobar_cpu_callback(&foobar_cpu_notifier, CPU_UP_PREPARE, i);
Satoru Takeuchica926e82006-10-19 23:29:06 -0700287 foobar_cpu_callback(&foobar_cpu_notifier, CPU_ONLINE, i);
288 }
Ashok Rajc8094062006-01-08 01:03:17 -0800289
290Q: If i would like to develop cpu hotplug support for a new architecture,
291 what do i need at a minimum?
292A: The following are what is required for CPU hotplug infrastructure to work
293 correctly.
294
295 - Make sure you have an entry in Kconfig to enable CONFIG_HOTPLUG_CPU
296 - __cpu_up() - Arch interface to bring up a CPU
297 - __cpu_disable() - Arch interface to shutdown a CPU, no more interrupts
298 can be handled by the kernel after the routine
299 returns. Including local APIC timers etc are
300 shutdown.
301 - __cpu_die() - This actually supposed to ensure death of the CPU.
302 Actually look at some example code in other arch
303 that implement CPU hotplug. The processor is taken
304 down from the idle() loop for that specific
305 architecture. __cpu_die() typically waits for some
306 per_cpu state to be set, to ensure the processor
307 dead routine is called to be sure positively.
308
309Q: I need to ensure that a particular cpu is not removed when there is some
310 work specific to this cpu is in progress.
311A: First switch the current thread context to preferred cpu
312
Satoru Takeuchica926e82006-10-19 23:29:06 -0700313 int my_func_on_cpu(int cpu)
314 {
315 cpumask_t saved_mask, new_mask = CPU_MASK_NONE;
316 int curr_cpu, err = 0;
Ashok Rajc8094062006-01-08 01:03:17 -0800317
Satoru Takeuchica926e82006-10-19 23:29:06 -0700318 saved_mask = current->cpus_allowed;
319 cpu_set(cpu, new_mask);
320 err = set_cpus_allowed(current, new_mask);
Ashok Rajc8094062006-01-08 01:03:17 -0800321
Satoru Takeuchica926e82006-10-19 23:29:06 -0700322 if (err)
323 return err;
Ashok Rajc8094062006-01-08 01:03:17 -0800324
Satoru Takeuchica926e82006-10-19 23:29:06 -0700325 /*
326 * If we got scheduled out just after the return from
327 * set_cpus_allowed() before running the work, this ensures
328 * we stay locked.
329 */
330 curr_cpu = get_cpu();
Ashok Rajc8094062006-01-08 01:03:17 -0800331
Satoru Takeuchica926e82006-10-19 23:29:06 -0700332 if (curr_cpu != cpu) {
333 err = -EAGAIN;
334 goto ret;
335 } else {
336 /*
337 * Do work : But cant sleep, since get_cpu() disables preempt
338 */
339 }
340 ret:
341 put_cpu();
342 set_cpus_allowed(current, saved_mask);
343 return err;
344 }
Ashok Rajc8094062006-01-08 01:03:17 -0800345
346
347Q: How do we determine how many CPUs are available for hotplug.
348A: There is no clear spec defined way from ACPI that can give us that
349 information today. Based on some input from Natalie of Unisys,
350 that the ACPI MADT (Multiple APIC Description Tables) marks those possible
351 CPUs in a system with disabled status.
352
353 Andi implemented some simple heuristics that count the number of disabled
354 CPUs in MADT as hotpluggable CPUS. In the case there are no disabled CPUS
355 we assume 1/2 the number of CPUs currently present can be hotplugged.
356
357 Caveat: Today's ACPI MADT can only provide 256 entries since the apicid field
358 in MADT is only 8 bits.
359
360User Space Notification
361
362Hotplug support for devices is common in Linux today. Its being used today to
363support automatic configuration of network, usb and pci devices. A hotplug
364event can be used to invoke an agent script to perform the configuration task.
365
366You can add /etc/hotplug/cpu.agent to handle hotplug notification user space
367scripts.
368
369 #!/bin/bash
370 # $Id: cpu.agent
371 # Kernel hotplug params include:
372 #ACTION=%s [online or offline]
373 #DEVPATH=%s
374 #
375 cd /etc/hotplug
376 . ./hotplug.functions
377
378 case $ACTION in
379 online)
380 echo `date` ":cpu.agent" add cpu >> /tmp/hotplug.txt
381 ;;
382 offline)
383 echo `date` ":cpu.agent" remove cpu >>/tmp/hotplug.txt
384 ;;
385 *)
386 debug_mesg CPU $ACTION event not supported
387 exit 1
388 ;;
389 esac