Skip to contentSkip to conversation

kexec: disable cpu hotplug until the rebooting cpu is stable

PATCH Superseded Pingfan Liu 2 participants 7 messages Jan 27, 2022 – Feb 14, 2022

Subsystems (4)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 kexec: disable cpu hotplug until the rebooting cpu is stable

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:
migrate_to_reboot_cpu();

   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu

machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

Note: this patch only affects the kexec-reboot on arches, which rely on cpu hotplug mechanism.

Cc Eric Biederman <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc Vincent Donnefort <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Mark Rutland <[email protected]>
Cc YueHaibing <[email protected]>
Cc Baokun Li <[email protected]>
Cc Randy Dunlap <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc kexec at lists.infradead.org

To: [email protected]
---
v1 -> v2:
improve commit log

 kernel/cpu.c        | 16 ++++++++++------
 kernel/kexec_core.c | 10 ++++------
 2 files changed, 14 insertions(+), 12 deletions(-)
kernel/cpu.c +10 -6
index 9c92147f0812..87bdf21de950 100644
@@ -1240,20 +1240,24 @@ int remove_cpu(unsigned int cpu)
 }
 EXPORT_SYMBOL_GPL(remove_cpu);
 
+/* primary_cpu keeps unchanged after migrate_to_reboot_cpu() */
 void smp_shutdown_nonboot_cpus(unsigned int primary_cpu)
 {
 	unsigned int cpu;
 	int error;
 
+	/*
+	 * Block other cpu hotplug event, so primary_cpu is always online if
+	 * it is not touched by us
+	 */
 	cpu_maps_update_begin();
-
 	/*
-	 * Make certain the cpu I'm about to reboot on is online.
-	 *
-	 * This is inline to what migrate_to_reboot_cpu() already do.
+	 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
+	 * no further code needs to use CPU hotplug (which is true in
+	 * the reboot case). However, the kexec path depends on using
+	 * CPU hotplug again; so re-enable it here.
 	 */
-	if (!cpu_online(primary_cpu))
-		primary_cpu = cpumask_first(cpu_online_mask);
+	__cpu_hotplug_enable();
 
 	for_each_online_cpu(cpu) {
 		if (cpu == primary_cpu)
kernel/kexec_core.c +4 -6
index 68480f731192..db4fa6b174e3 100644
@@ -1168,14 +1168,12 @@ int kernel_kexec(void)
 		kexec_in_progress = true;
 		kernel_restart_prepare("kexec reboot");
 		migrate_to_reboot_cpu();
-
 		/*
-		 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
-		 * no further code needs to use CPU hotplug (which is true in
-		 * the reboot case). However, the kexec path depends on using
-		 * CPU hotplug again; so re-enable it here.
+		 * migrate_to_reboot_cpu() disables CPU hotplug. If an arch
+		 * relies on the cpu teardown to achieve reboot, it needs to
+		 * re-enable CPU hotplug there.
 		 */
-		cpu_hotplug_enable();
+
 		pr_notice("Starting new kernel\n");
 		machine_shutdown();
 	}
Signature · 1 line
2.31.1
#1 lore raw

Hi Pingfan,

Quoted 12 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();
Quoted 6 lines

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

If I didn't go through code wrongly, you may miss the x86 case. Several ARCHes do call smp_shutdown_nonboot_cpus() in machine_shutdown() in kexec reboot code path, while x86 doesn't. If I am right, you may need reconsider if this patch is needed or need be adjustd.

Are you optimizing code path, or you meet a real problem? I haven't checked v1, but I also didn't see it's told in patch log which case it is.

Quoted 87 lines

Note: this patch only affects the kexec-reboot on arches, which rely on cpu hotplug mechanism.

Cc Eric Biederman <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc Vincent Donnefort <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Mark Rutland <[email protected]>
Cc YueHaibing <[email protected]>
Cc Baokun Li <[email protected]>
Cc Randy Dunlap <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc kexec at lists.infradead.org

To: linux-kernel at vger.kernel.org
---
v1 -> v2:
improve commit log

 kernel/cpu.c        | 16 ++++++++++------
 kernel/kexec_core.c | 10 ++++------
 2 files changed, 14 insertions(+), 12 deletions(-)
kernel/cpu.c +10 -6
index 9c92147f0812..87bdf21de950 100644
@@ -1240,20 +1240,24 @@ int remove_cpu(unsigned int cpu)
 }
 EXPORT_SYMBOL_GPL(remove_cpu);
 
+/* primary_cpu keeps unchanged after migrate_to_reboot_cpu() */
 void smp_shutdown_nonboot_cpus(unsigned int primary_cpu)
 {
 	unsigned int cpu;
 	int error;
 
+	/*
+	 * Block other cpu hotplug event, so primary_cpu is always online if
+	 * it is not touched by us
+	 */
 	cpu_maps_update_begin();
-
 	/*
-	 * Make certain the cpu I'm about to reboot on is online.
-	 *
-	 * This is inline to what migrate_to_reboot_cpu() already do.
+	 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
+	 * no further code needs to use CPU hotplug (which is true in
+	 * the reboot case). However, the kexec path depends on using
+	 * CPU hotplug again; so re-enable it here.
 	 */
-	if (!cpu_online(primary_cpu))
-		primary_cpu = cpumask_first(cpu_online_mask);
+	__cpu_hotplug_enable();
 
 	for_each_online_cpu(cpu) {
 		if (cpu == primary_cpu)
kernel/kexec_core.c +4 -6
index 68480f731192..db4fa6b174e3 100644
@@ -1168,14 +1168,12 @@ int kernel_kexec(void)
 		kexec_in_progress = true;
 		kernel_restart_prepare("kexec reboot");
 		migrate_to_reboot_cpu();
-
 		/*
-		 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
-		 * no further code needs to use CPU hotplug (which is true in
-		 * the reboot case). However, the kexec path depends on using
-		 * CPU hotplug again; so re-enable it here.
+		 * migrate_to_reboot_cpu() disables CPU hotplug. If an arch
+		 * relies on the cpu teardown to achieve reboot, it needs to
+		 * re-enable CPU hotplug there.
 		 */
-		cpu_hotplug_enable();
+
 		pr_notice("Starting new kernel\n");
 		machine_shutdown();
 	}
Signature · 1 line
2.31.1
#2 lore raw

On Thu, Jan 27, 2022 at 05:41:44PM +0800, Baoquan He wrote:
Hi Baoquan,

Thanks for reviewing, please see comment inlined

Quoted 18 lines

Hi Pingfan,

Quoted 15 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Quoted 9 lines
Quoted 3 lines

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

If I didn't go through code wrongly, you may miss the x86 case. Several ARCHes do call smp_shutdown_nonboot_cpus() in machine_shutdown() in kexec reboot code path, while x86 doesn't. If I am right, you may need reconsider if this patch is needed or need be adjustd.

Citing the code piece in kernel_kexec()

                migrate_to_reboot_cpu();

                /*
                 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
                 * no further code needs to use CPU hotplug (which is true in
                 * the reboot case). However, the kexec path depends on using
                 * CPU hotplug again; so re-enable it here.
                 */
                cpu_hotplug_enable();
                pr_notice("Starting new kernel\n");
                machine_shutdown();

So maybe it can be considered in such way: "cpu_hotplug_enable()" is not needed by x86 and ppc, so this patch removes it, while re-displace it in a more appropriate place for arm64/riscv ...

Quoted 4 lines

Are you optimizing code path, or you meet a real problem? I haven't checked v1, but I also didn't see it's told in patch log which case it is.

Simplify the code path and make the logic look straight forward.

And sorry for bad expression. I had thought I expressed it by (citing git log)

|| The following identical code piece appears in both

                 ^^^^^^^^
|| migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():
|| 
|| 	if (!cpu_online(primary_cpu))
|| 		primary_cpu = cpumask_first(cpu_online_mask);
|| 
|| This is due to a breakage like the following:
                    ^^^^^^^^
||    migrate_to_reboot_cpu();
||    cpu_hotplug_enable();
||                           --> comes a cpu_down(this_cpu) on other cpu
||    machine_shutdown();
|| 
|| Although the kexec-reboot task can get through a cpu_down() on its cpu,
                                      ^^^^^^^^^^^

|| this code looks a little confusing.

Should I rephrase it?

Thanks,

Pingfan

Quoted 88 lines
Quoted 87 lines

Note: this patch only affects the kexec-reboot on arches, which rely on cpu hotplug mechanism.

Cc Eric Biederman <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc Vincent Donnefort <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Mark Rutland <[email protected]>
Cc YueHaibing <[email protected]>
Cc Baokun Li <[email protected]>
Cc Randy Dunlap <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc kexec at lists.infradead.org

To: linux-kernel at vger.kernel.org
---
v1 -> v2:
improve commit log

 kernel/cpu.c        | 16 ++++++++++------
 kernel/kexec_core.c | 10 ++++------
 2 files changed, 14 insertions(+), 12 deletions(-)
kernel/cpu.c +10 -6
index 9c92147f0812..87bdf21de950 100644
@@ -1240,20 +1240,24 @@ int remove_cpu(unsigned int cpu)
 }
 EXPORT_SYMBOL_GPL(remove_cpu);
 
+/* primary_cpu keeps unchanged after migrate_to_reboot_cpu() */
 void smp_shutdown_nonboot_cpus(unsigned int primary_cpu)
 {
 	unsigned int cpu;
 	int error;
 
+	/*
+	 * Block other cpu hotplug event, so primary_cpu is always online if
+	 * it is not touched by us
+	 */
 	cpu_maps_update_begin();
-
 	/*
-	 * Make certain the cpu I'm about to reboot on is online.
-	 *
-	 * This is inline to what migrate_to_reboot_cpu() already do.
+	 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
+	 * no further code needs to use CPU hotplug (which is true in
+	 * the reboot case). However, the kexec path depends on using
+	 * CPU hotplug again; so re-enable it here.
 	 */
-	if (!cpu_online(primary_cpu))
-		primary_cpu = cpumask_first(cpu_online_mask);
+	__cpu_hotplug_enable();
 
 	for_each_online_cpu(cpu) {
 		if (cpu == primary_cpu)
kernel/kexec_core.c +4 -6
index 68480f731192..db4fa6b174e3 100644
@@ -1168,14 +1168,12 @@ int kernel_kexec(void)
 		kexec_in_progress = true;
 		kernel_restart_prepare("kexec reboot");
 		migrate_to_reboot_cpu();
-
 		/*
-		 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
-		 * no further code needs to use CPU hotplug (which is true in
-		 * the reboot case). However, the kexec path depends on using
-		 * CPU hotplug again; so re-enable it here.
+		 * migrate_to_reboot_cpu() disables CPU hotplug. If an arch
+		 * relies on the cpu teardown to achieve reboot, it needs to
+		 * re-enable CPU hotplug there.
 		 */
-		cpu_hotplug_enable();
+
 		pr_notice("Starting new kernel\n");
 		machine_shutdown();
 	}
Signature · 1 line
2.31.1
#3 lore raw
Quoted 45 lines · Pingfan Liu

On Thu, Jan 27, 2022 at 05:41:44PM +0800, Baoquan He wrote:
Hi Baoquan,

Thanks for reviewing, please see comment inlined

Quoted 18 lines

Hi Pingfan,

Quoted 15 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Quoted 9 lines
Quoted 3 lines

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

If I didn't go through code wrongly, you may miss the x86 case. Several ARCHes do call smp_shutdown_nonboot_cpus() in machine_shutdown() in kexec reboot code path, while x86 doesn't. If I am right, you may need reconsider if this patch is needed or need be adjustd.

Citing the code piece in kernel_kexec()

                migrate_to_reboot_cpu();

                /*
                 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
                 * no further code needs to use CPU hotplug (which is true in
                 * the reboot case). However, the kexec path depends on using
                 * CPU hotplug again; so re-enable it here.
                 */
                cpu_hotplug_enable();
                pr_notice("Starting new kernel\n");
                machine_shutdown();
Quoted 3 lines

So maybe it can be considered in such way: "cpu_hotplug_enable()" is not needed by x86 and ppc, so this patch removes it, while re-displace it in a more appropriate place for arm64/riscv ...

OK, so the thing is:

==
In the current code of kexec, we disable cpu hotplug and check reboot cpu validity in migrate_to_reboot_cpu(), then enable cpu hotplug. Then in machine_shutdown()->smp_shutdown_nonboot_cpus(), check the reboot cpu and disable cpu hotplug again.

In this patch, it disables cpu hotplug in migrate_to_reboot_cpu() and keep it till entering into smp_shutdown_nonboot_cpus() to shutdown all other cpu with hotplug mechanism, then disable it again. With this change, it won't need to double check the reboot cpu validity.

This change only affect ARCHes relying on hotplug to shutdown cpu before kexec reboot, e.g arm64, risc-v. Other ARCH like x86 is not affected. ==

Do I got it right?

#4 lore raw
Quoted 50 lines · Baoquan He
Quoted 45 lines · Pingfan Liu

On Thu, Jan 27, 2022 at 05:41:44PM +0800, Baoquan He wrote:
Hi Baoquan,

Thanks for reviewing, please see comment inlined

Quoted 18 lines

Hi Pingfan,

Quoted 15 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Quoted 9 lines
Quoted 3 lines

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

If I didn't go through code wrongly, you may miss the x86 case. Several ARCHes do call smp_shutdown_nonboot_cpus() in machine_shutdown() in kexec reboot code path, while x86 doesn't. If I am right, you may need reconsider if this patch is needed or need be adjustd.

Citing the code piece in kernel_kexec()

                migrate_to_reboot_cpu();

                /*
                 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
                 * no further code needs to use CPU hotplug (which is true in
                 * the reboot case). However, the kexec path depends on using
                 * CPU hotplug again; so re-enable it here.
                 */
                cpu_hotplug_enable();
                pr_notice("Starting new kernel\n");
                machine_shutdown();
Quoted 3 lines

So maybe it can be considered in such way: "cpu_hotplug_enable()" is not needed by x86 and ppc, so this patch removes it, while re-displace it in a more appropriate place for arm64/riscv ...

Quoted 7 lines

OK, so the thing is:

==
In the current code of kexec, we disable cpu hotplug and check reboot cpu validity in migrate_to_reboot_cpu(), then enable cpu hotplug. Then in machine_shutdown()->smp_shutdown_nonboot_cpus(), check the reboot cpu and disable cpu hotplug again.

^^^
No, there is no any new call to cpu_hotplug_disable() after migrate_to_reboot_cpu(). smp_shutdown_nonboot_cpus() just leaves cpu_hotplug_disabled==0 as it is.

Quoted 4 lines

In this patch, it disables cpu hotplug in migrate_to_reboot_cpu() and keep it till entering into smp_shutdown_nonboot_cpus() to shutdown all other cpu with hotplug mechanism, then disable it again. With this

                                         ^^^
Here is enable, i.e. smp_shutdown_nonboot_cpus() makes
cpu_hotplug_disabled switch from 1 to 0, so the following fn can work
||	static int cpu_down_maps_locked(unsigned int cpu, enum cpuhp_state target)
||	{
||		if (cpu_hotplug_disabled)
||			return -EBUSY;
||		return _cpu_down(cpu, 0, target);
||	}
Quoted 4 lines

change, it won't need to double check the reboot cpu validity.

This change only affect ARCHes relying on hotplug to shutdown cpu before kexec reboot, e.g arm64, risc-v. Other ARCH like x86 is not affected.

Yes.

Quoted 4 lines

==

Do I got it right?

Here, neither the original code nor this patch has another cpu_hotplug_disable(), the only change is the shift of cpu_hotplug_enable() to an onwards place.

And I think you should get my idea except this.

Thanks,

Pingfan

#5 lore raw
Quoted 58 lines · Pingfan Liu
Quoted 50 lines · Baoquan He
Quoted 45 lines · Pingfan Liu

On Thu, Jan 27, 2022 at 05:41:44PM +0800, Baoquan He wrote:
Hi Baoquan,

Thanks for reviewing, please see comment inlined

Quoted 18 lines

Hi Pingfan,

Quoted 15 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

	if (!cpu_online(primary_cpu))
		primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Quoted 9 lines
Quoted 3 lines

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

If I didn't go through code wrongly, you may miss the x86 case. Several ARCHes do call smp_shutdown_nonboot_cpus() in machine_shutdown() in kexec reboot code path, while x86 doesn't. If I am right, you may need reconsider if this patch is needed or need be adjustd.

Citing the code piece in kernel_kexec()

                migrate_to_reboot_cpu();

                /*
                 * migrate_to_reboot_cpu() disables CPU hotplug assuming that
                 * no further code needs to use CPU hotplug (which is true in
                 * the reboot case). However, the kexec path depends on using
                 * CPU hotplug again; so re-enable it here.
                 */
                cpu_hotplug_enable();
                pr_notice("Starting new kernel\n");
                machine_shutdown();
Quoted 3 lines

So maybe it can be considered in such way: "cpu_hotplug_enable()" is not needed by x86 and ppc, so this patch removes it, while re-displace it in a more appropriate place for arm64/riscv ...

Quoted 7 lines

OK, so the thing is:

==
In the current code of kexec, we disable cpu hotplug and check reboot cpu validity in migrate_to_reboot_cpu(), then enable cpu hotplug. Then in machine_shutdown()->smp_shutdown_nonboot_cpus(), check the reboot cpu and disable cpu hotplug again.

Quoted 4 lines

^^^
No, there is no any new call to cpu_hotplug_disable() after migrate_to_reboot_cpu(). smp_shutdown_nonboot_cpus() just leaves cpu_hotplug_disabled==0 as it is.

Hmm, there's 'cpu_hotplug_disabled++;' at the end of smp_shutdown_nonboot_cpus(). It's disabling cpu hotplug again, I think.

Quoted 7 lines
Quoted 4 lines

In this patch, it disables cpu hotplug in migrate_to_reboot_cpu() and keep it till entering into smp_shutdown_nonboot_cpus() to shutdown all other cpu with hotplug mechanism, then disable it again. With this

                                         ^^^

Here is enable, i.e. smp_shutdown_nonboot_cpus() makes cpu_hotplug_disabled switch from 1 to 0, so the following fn can work

Right, you keep the cpu hotplug disabled till in smp_shutdown_nonboot_cpus(), then enable it to shut down all other cpus.

Quoted 26 lines
||	static int cpu_down_maps_locked(unsigned int cpu, enum cpuhp_state target)
||	{
||		if (cpu_hotplug_disabled)
||			return -EBUSY;
||		return _cpu_down(cpu, 0, target);
||	}
Quoted 4 lines

change, it won't need to double check the reboot cpu validity.

This change only affect ARCHes relying on hotplug to shutdown cpu before kexec reboot, e.g arm64, risc-v. Other ARCH like x86 is not affected.

Yes.

Quoted 4 lines

==

Do I got it right?

Here, neither the original code nor this patch has another cpu_hotplug_disable(), the only change is the shift of cpu_hotplug_enable() to an onwards place.

And I think you should get my idea except this.

Thanks,

Pingfan

#6 lore raw
In reply to Pingfan Liu

Gently ping, maintainers, could you share your opinions?

Thanks

Quoted 106 lines · Pingfan Liu

The following identical code piece appears in both migrate_to_reboot_cpu() and smp_shutdown_nonboot_cpus():

        if (!cpu_online(primary_cpu))
                primary_cpu = cpumask_first(cpu_online_mask);

This is due to a breakage like the following:

   migrate_to_reboot_cpu();
   cpu_hotplug_enable();
                          --> comes a cpu_down(this_cpu) on other cpu
   machine_shutdown();

Although the kexec-reboot task can get through a cpu_down() on its cpu, this code looks a little confusing.

Make things straight forward by keeping cpu hotplug disabled until smp_shutdown_nonboot_cpus() holds cpu_add_remove_lock. By this way, the breakage is squashed out and the rebooting cpu can keep unchanged.

Note: this patch only affects the kexec-reboot on arches, which rely on cpu hotplug mechanism.

Cc Eric Biederman <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc Vincent Donnefort <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Mark Rutland <[email protected]>
Cc YueHaibing <[email protected]>
Cc Baokun Li <[email protected]>
Cc Randy Dunlap <[email protected]>
Cc Valentin Schneider <[email protected]>
Cc kexec at lists.infradead.org

To: linux-kernel at vger.kernel.org
---
v1 -> v2:
improve commit log

 kernel/cpu.c        | 16 ++++++++++------
 kernel/kexec_core.c | 10 ++++------
 2 files changed, 14 insertions(+), 12 deletions(-)
kernel/cpu.c +10 -6
index 9c92147f0812..87bdf21de950 100644
@@ -1240,20 +1240,24 @@ int remove_cpu(unsigned int cpu)
 }
 EXPORT_SYMBOL_GPL(remove_cpu);
 
+/* primary_cpu keeps unchanged after migrate_to_reboot_cpu() */
 void smp_shutdown_nonboot_cpus(unsigned int primary_cpu)
 {
        unsigned int cpu;
        int error;
 
+       /*
+        * Block other cpu hotplug event, so primary_cpu is always online if
+        * it is not touched by us
+        */
        cpu_maps_update_begin();
-
        /*
-        * Make certain the cpu I'm about to reboot on is online.
-        *
-        * This is inline to what migrate_to_reboot_cpu() already do.
+        * migrate_to_reboot_cpu() disables CPU hotplug assuming that
+        * no further code needs to use CPU hotplug (which is true in
+        * the reboot case). However, the kexec path depends on using
+        * CPU hotplug again; so re-enable it here.
         */
-       if (!cpu_online(primary_cpu))
-               primary_cpu = cpumask_first(cpu_online_mask);
+       __cpu_hotplug_enable();
 
        for_each_online_cpu(cpu) {
                if (cpu == primary_cpu)
kernel/kexec_core.c +4 -7
index 68480f731192..db4fa6b174e3 100644
@@ -1168,14 +1168,12 @@ int kernel_kexec(void)
                kexec_in_progress = true;
                kernel_restart_prepare("kexec reboot");
                migrate_to_reboot_cpu();
-
                /*
-                * migrate_to_reboot_cpu() disables CPU hotplug assuming that
-                * no further code needs to use CPU hotplug (which is true in
-                * the reboot case). However, the kexec path depends on using
-                * CPU hotplug again; so re-enable it here.
+                * migrate_to_reboot_cpu() disables CPU hotplug. If an arch
+                * relies on the cpu teardown to achieve reboot, it needs to
+                * re-enable CPU hotplug there.
                 */
-               cpu_hotplug_enable();
+
                pr_notice("Starting new kernel\n");
                machine_shutdown();
        }
--

2.31.1