powerpc/pseries: Only wait for dying CPU after call to rtas_stop_self()
PATCH v3 1/1 powerpc/pseries: Only wait for dying CPU after call to rtas_stop_self()
When testing DLPAR CPU add/remove on a system under stress, pseries_cpu_die() doesn't wait long enough for a CPU to die:
[ 446.983944] cpu 148 (hwid 148) Ready to die... [ 446.984062] cpu 149 (hwid 149) Ready to die... [ 446.993518] cpu 150 (hwid 150) Ready to die... [ 446.993543] Querying DEAD? cpu 150 (150) shows 2 [ 446.994098] cpu 151 (hwid 151) Ready to die... [ 447.133726] cpu 136 (hwid 136) Ready to die... [ 447.403532] cpu 137 (hwid 137) Ready to die... [ 447.403772] cpu 138 (hwid 138) Ready to die... [ 447.403839] cpu 139 (hwid 139) Ready to die... [ 447.403887] cpu 140 (hwid 140) Ready to die... [ 447.403937] cpu 141 (hwid 141) Ready to die... [ 447.403979] cpu 142 (hwid 142) Ready to die... [ 447.404038] cpu 143 (hwid 143) Ready to die... [ 447.513546] cpu 128 (hwid 128) Ready to die... [ 447.693533] cpu 129 (hwid 129) Ready to die... [ 447.693999] cpu 130 (hwid 130) Ready to die... [ 447.703530] cpu 131 (hwid 131) Ready to die... [ 447.704087] Querying DEAD? cpu 132 (132) shows 2 [ 447.704102] cpu 132 (hwid 132) Ready to die... [ 447.713534] cpu 133 (hwid 133) Ready to die... [ 447.714064] Querying DEAD? cpu 134 (134) shows 2
This is a race between one CPU stopping and another one calling pseries_cpu_die() to wait for it to stop. That function does a short busy loop calling RTAS query-cpu-stopped-state on the stopping CPU to verify that it is stopped, but I think there's a lot for the stopping CPU to do which may take longer than this loop allows.
As can be seen in the dmesg right before or after the "Querying DEAD?" messages, if pseries_cpu_die() waited a little longer it would have seen the CPU in the stopped state.
What I think is going on is that CPU 134 was inactive at the time it was unplugged. In that case, dlpar_offline_cpu() calls H_PROD on that CPU and immediately calls pseries_cpu_die(). Meanwhile, the prodded CPU activates and start the process of stopping itself. The busy loop is not long enough to allow for the CPU to wake up and complete the stopping process.
This can be a problem because if the busy loop finishes too early, then the kernel may offline another CPU before the previous one finished dying, which would lead to two concurrent calls to rtas-stop-self, which is prohibited by the PAPR.
We can make the race a lot more even if we only start querying if the CPU is stopped when the stopping CPU is close to call rtas_stop_self(). Since pseries_mach_cpu_die() sets the CPU current state to offline almost immediately before calling rtas_stop_self(), we use that as a signal that it is either already stopped or very close to that point, and we can start the busy loop.
As suggested by Michael Ellerman, this patch also changes the busy loop to wait for a fixed amount of wall time. Based on the measurements that Gautham did on a POWER9 system, in successful cases of smp_query_cpu_stopped(cpu) returning affirmative, the maximum time spent inside the loop was was 10 ms. This patch loops for 20 ms just be sure.
--- arch/powerpc/platforms/pseries/hotplug-cpu.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)
I have seen this problem since v4.8. Should this patch go to stable as well?
Changes since v2:
- Increaded busy loop to 200 iterations so that it can last up to 20 ms (suggested by Gautham).
- Changed commit message to include Gautham's remarks.
arch/powerpc/platforms/pseries/hotplug-cpu.c +11 -2
@@ -214,13 +214,22 @@ static void pseries_cpu_die(unsigned int cpu)msleep (1 ); } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { +/* +* If the current state is not offline yet, it means that the +* dying CPU (which is either in pseries_mach_cpu_die() or in +* the process of getting there) didn't have a chance yet to +* call rtas_stop_self() and therefore it's too early to query +* if the CPU is stopped. +*/ +spin_event_timeout (get_cpu_current_state (cpu) ==CPU_STATE_OFFLINE , +100000 ,100 ); -for (tries =0 ; tries <25 ; tries++) { +for (tries =0 ; tries <200 ; tries++) { cpu_status =smp_query_cpu_stopped (pcpu);if (cpu_status ==QCSS_STOPPED || cpu_status ==QCSS_HARDWARE_ERROR )break ; -cpu_relax (); +udelay (100 ); } }
Hello Thiago,
Quoted 53 lines · Thiago Jung Bauermann
When testing DLPAR CPU add/remove on a system under stress,
pseries_cpu_die()doesn't wait long enough for a CPU to die:[ 446.983944] cpu 148 (hwid 148) Ready to die... [ 446.984062] cpu 149 (hwid 149) Ready to die... [ 446.993518] cpu 150 (hwid 150) Ready to die... [ 446.993543] Querying DEAD? cpu 150 (150) shows 2 [ 446.994098] cpu 151 (hwid 151) Ready to die... [ 447.133726] cpu 136 (hwid 136) Ready to die... [ 447.403532] cpu 137 (hwid 137) Ready to die... [ 447.403772] cpu 138 (hwid 138) Ready to die... [ 447.403839] cpu 139 (hwid 139) Ready to die... [ 447.403887] cpu 140 (hwid 140) Ready to die... [ 447.403937] cpu 141 (hwid 141) Ready to die... [ 447.403979] cpu 142 (hwid 142) Ready to die... [ 447.404038] cpu 143 (hwid 143) Ready to die... [ 447.513546] cpu 128 (hwid 128) Ready to die... [ 447.693533] cpu 129 (hwid 129) Ready to die... [ 447.693999] cpu 130 (hwid 130) Ready to die... [ 447.703530] cpu 131 (hwid 131) Ready to die... [ 447.704087] Querying DEAD? cpu 132 (132) shows 2 [ 447.704102] cpu 132 (hwid 132) Ready to die... [ 447.713534] cpu 133 (hwid 133) Ready to die... [ 447.714064] Querying DEAD? cpu 134 (134) shows 2This is a race between one CPU stopping and another one calling
pseries_cpu_die()to wait for it to stop. That function does a short busy loop calling RTAS query-cpu-stopped-state on the stopping CPU to verify that it is stopped, but I think there's a lot for the stopping CPU to do which may take longer than this loop allows.As can be seen in the dmesg right before or after the "Querying DEAD?" messages, if
pseries_cpu_die()waited a little longer it would have seen the CPU in the stopped state.What I think is going on is that CPU 134 was inactive at the time it was unplugged. In that case,
dlpar_offline_cpu()callsH_PRODon that CPU and immediately callspseries_cpu_die(). Meanwhile, the prodded CPU activates and start the process of stopping itself. The busy loop is not long enough to allow for the CPU to wake up and complete the stopping process.This can be a problem because if the busy loop finishes too early, then the kernel may offline another CPU before the previous one finished dying, which would lead to two concurrent calls to rtas-stop-self, which is prohibited by the PAPR.
We can make the race a lot more even if we only start querying if the CPU is stopped when the stopping CPU is close to call
rtas_stop_self(). Sincepseries_mach_cpu_die()sets the CPU current state to offline almost immediately before callingrtas_stop_self(), we use that as a signal that it is either already stopped or very close to that point, and we can start the busy loop.
Quoted 7 lines
As suggested by Michael Ellerman, this patch also changes the busy loop to wait for a fixed amount of wall time. Based on the measurements that Gautham did on a POWER9 system, in successful cases of
smp_query_cpu_stopped(cpu)returning affirmative, the maximum time spent inside the loop was was 10 ms. This patch loops for 20 ms just be sure.Signed-off-by Thiago Jung Bauermann <[email protected]>
Thanks for this version. I have tested the patch and we no longer see the "Querying DEAD? cpu X (Y) shows 2" message.
--
Thanks and Regards
gautham.
Quoted 7 lines · Gautham R Shenoy
Quoted 1 line
Signed-off-by Thiago Jung Bauermann <[email protected]>Thanks for this version. I have tested the patch and we no longer see the "Querying DEAD? cpu X (Y) shows 2" message.
Tested-and-Reviewed-by Gautham R. Shenoy <[email protected]>
Thanks for reviewing and testing the patch!
Signature · 2 lines
Thiago Jung Bauermann IBM Linux Technology Center
Hello,
Ping?
Signature · 2 lines
Thiago Jung Bauermann IBM Linux Technology Center
Quoted 102 lines · Thiago Jung Bauermann
When testing DLPAR CPU add/remove on a system under stress,
pseries_cpu_die()doesn't wait long enough for a CPU to die:[ 446.983944] cpu 148 (hwid 148) Ready to die... [ 446.984062] cpu 149 (hwid 149) Ready to die... [ 446.993518] cpu 150 (hwid 150) Ready to die... [ 446.993543] Querying DEAD? cpu 150 (150) shows 2 [ 446.994098] cpu 151 (hwid 151) Ready to die... [ 447.133726] cpu 136 (hwid 136) Ready to die... [ 447.403532] cpu 137 (hwid 137) Ready to die... [ 447.403772] cpu 138 (hwid 138) Ready to die... [ 447.403839] cpu 139 (hwid 139) Ready to die... [ 447.403887] cpu 140 (hwid 140) Ready to die... [ 447.403937] cpu 141 (hwid 141) Ready to die... [ 447.403979] cpu 142 (hwid 142) Ready to die... [ 447.404038] cpu 143 (hwid 143) Ready to die... [ 447.513546] cpu 128 (hwid 128) Ready to die... [ 447.693533] cpu 129 (hwid 129) Ready to die... [ 447.693999] cpu 130 (hwid 130) Ready to die... [ 447.703530] cpu 131 (hwid 131) Ready to die... [ 447.704087] Querying DEAD? cpu 132 (132) shows 2 [ 447.704102] cpu 132 (hwid 132) Ready to die... [ 447.713534] cpu 133 (hwid 133) Ready to die... [ 447.714064] Querying DEAD? cpu 134 (134) shows 2This is a race between one CPU stopping and another one calling
pseries_cpu_die()to wait for it to stop. That function does a short busy loop calling RTAS query-cpu-stopped-state on the stopping CPU to verify that it is stopped, but I think there's a lot for the stopping CPU to do which may take longer than this loop allows.As can be seen in the dmesg right before or after the "Querying DEAD?" messages, if
pseries_cpu_die()waited a little longer it would have seen the CPU in the stopped state.What I think is going on is that CPU 134 was inactive at the time it was unplugged. In that case,
dlpar_offline_cpu()callsH_PRODon that CPU and immediately callspseries_cpu_die(). Meanwhile, the prodded CPU activates and start the process of stopping itself. The busy loop is not long enough to allow for the CPU to wake up and complete the stopping process.This can be a problem because if the busy loop finishes too early, then the kernel may offline another CPU before the previous one finished dying, which would lead to two concurrent calls to rtas-stop-self, which is prohibited by the PAPR.
We can make the race a lot more even if we only start querying if the CPU is stopped when the stopping CPU is close to call
rtas_stop_self(). Sincepseries_mach_cpu_die()sets the CPU current state to offline almost immediately before callingrtas_stop_self(), we use that as a signal that it is either already stopped or very close to that point, and we can start the busy loop.As suggested by Michael Ellerman, this patch also changes the busy loop to wait for a fixed amount of wall time. Based on the measurements that Gautham did on a POWER9 system, in successful cases of
smp_query_cpu_stopped(cpu)returning affirmative, the maximum time spent inside the loop was was 10 ms. This patch loops for 20 ms just be sure.Signed-off-by Thiago Jung Bauermann <[email protected]>Analyzed-by Gautham R Shenoy <[email protected]>--- arch/powerpc/platforms/pseries/hotplug-cpu.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)I have seen this problem since v4.8. Should this patch go to stable as well?
Changes since v2:
- Increaded busy loop to 200 iterations so that it can last up to 20 ms (suggested by Gautham).
- Changed commit message to include Gautham's remarks.arch/powerpc/platforms/pseries/hotplug-cpu.c +11
-2@@ -214,13 +214,22 @@ static void pseries_cpu_die(unsigned int cpu)msleep (1 ); } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { +/* +* If the current state is not offline yet, it means that the +* dying CPU (which is either in pseries_mach_cpu_die() or in +* the process of getting there) didn't have a chance yet to +* call rtas_stop_self() and therefore it's too early to query +* if the CPU is stopped. +*/ +spin_event_timeout (get_cpu_current_state (cpu) ==CPU_STATE_OFFLINE , +100000 ,100 ); -for (tries =0 ; tries <25 ; tries++) { +for (tries =0 ; tries <200 ; tries++) { cpu_status =smp_query_cpu_stopped (pcpu);if (cpu_status ==QCSS_STOPPED || cpu_status ==QCSS_HARDWARE_ERROR )break ; -cpu_relax (); +udelay (100 ); } }
Nicholas Piggin
reviewerThiago Jung Bauermann's on April 11, 2019 9:08 am:
Quoted 102 lines
Hello,
Ping?
Signature · 2 lines
Thiago Jung Bauermann IBM Linux Technology CenterQuoted 80 lines · Thiago Jung Bauermann
When testing DLPAR CPU add/remove on a system under stress,
pseries_cpu_die()doesn't wait long enough for a CPU to die:[ 446.983944] cpu 148 (hwid 148) Ready to die... [ 446.984062] cpu 149 (hwid 149) Ready to die... [ 446.993518] cpu 150 (hwid 150) Ready to die... [ 446.993543] Querying DEAD? cpu 150 (150) shows 2 [ 446.994098] cpu 151 (hwid 151) Ready to die... [ 447.133726] cpu 136 (hwid 136) Ready to die... [ 447.403532] cpu 137 (hwid 137) Ready to die... [ 447.403772] cpu 138 (hwid 138) Ready to die... [ 447.403839] cpu 139 (hwid 139) Ready to die... [ 447.403887] cpu 140 (hwid 140) Ready to die... [ 447.403937] cpu 141 (hwid 141) Ready to die... [ 447.403979] cpu 142 (hwid 142) Ready to die... [ 447.404038] cpu 143 (hwid 143) Ready to die... [ 447.513546] cpu 128 (hwid 128) Ready to die... [ 447.693533] cpu 129 (hwid 129) Ready to die... [ 447.693999] cpu 130 (hwid 130) Ready to die... [ 447.703530] cpu 131 (hwid 131) Ready to die... [ 447.704087] Querying DEAD? cpu 132 (132) shows 2 [ 447.704102] cpu 132 (hwid 132) Ready to die... [ 447.713534] cpu 133 (hwid 133) Ready to die... [ 447.714064] Querying DEAD? cpu 134 (134) shows 2This is a race between one CPU stopping and another one calling
pseries_cpu_die()to wait for it to stop. That function does a short busy loop calling RTAS query-cpu-stopped-state on the stopping CPU to verify that it is stopped, but I think there's a lot for the stopping CPU to do which may take longer than this loop allows.As can be seen in the dmesg right before or after the "Querying DEAD?" messages, if
pseries_cpu_die()waited a little longer it would have seen the CPU in the stopped state.What I think is going on is that CPU 134 was inactive at the time it was unplugged. In that case,
dlpar_offline_cpu()callsH_PRODon that CPU and immediately callspseries_cpu_die(). Meanwhile, the prodded CPU activates and start the process of stopping itself. The busy loop is not long enough to allow for the CPU to wake up and complete the stopping process.This can be a problem because if the busy loop finishes too early, then the kernel may offline another CPU before the previous one finished dying, which would lead to two concurrent calls to rtas-stop-self, which is prohibited by the PAPR.
We can make the race a lot more even if we only start querying if the CPU is stopped when the stopping CPU is close to call
rtas_stop_self(). Sincepseries_mach_cpu_die()sets the CPU current state to offline almost immediately before callingrtas_stop_self(), we use that as a signal that it is either already stopped or very close to that point, and we can start the busy loop.As suggested by Michael Ellerman, this patch also changes the busy loop to wait for a fixed amount of wall time. Based on the measurements that Gautham did on a POWER9 system, in successful cases of
smp_query_cpu_stopped(cpu)returning affirmative, the maximum time spent inside the loop was was 10 ms. This patch loops for 20 ms just be sure.Signed-off-by Thiago Jung Bauermann <[email protected]>Analyzed-by Gautham R Shenoy <[email protected]>--- arch/powerpc/platforms/pseries/hotplug-cpu.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)I have seen this problem since v4.8. Should this patch go to stable as well?
Changes since v2:
- Increaded busy loop to 200 iterations so that it can last up to 20 ms (suggested by Gautham).
- Changed commit message to include Gautham's remarks.arch/powerpc/platforms/pseries/hotplug-cpu.c +0
-0@@ -214,13 +214,22 @@ static void pseries_cpu_die(unsigned int cpu)msleep (1 ); }arch/powerpc/platforms/pseries/hotplug-cpu.c static void pseries_cpu_die(unsigned int cpu) · line 225
Quoted 10 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c +9
-0}else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { +/* +* If the current state is not offline yet, it means that the +* dying CPU (which is either in pseries_mach_cpu_die() or in +* the process of getting there) didn't have a chance yet to +* call rtas_stop_self() and therefore it's too early to query +* if the CPU is stopped. +*/ +spin_event_timeout (get_cpu_current_state (cpu) ==CPU_STATE_OFFLINE , +100000 ,100 );
If the CPU state does not go to offline here, you should give up and return online, right? Otherwise I think query-cpu-stopped-state can get confused by CPUs in idle and you get a false positive.
That race can still happen, we would really need a sequence count check over current CPU state to ensure we got a race-free qcss value, but at least a check here should make the race implausible to hit.
Thanks,
Nick
Quoted 14 lines
Quoted 12 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c +2
-2-for (tries =0 ; tries <25 ; tries++) { +for (tries =0 ; tries <200 ; tries++) { cpu_status =smp_query_cpu_stopped (pcpu);if (cpu_status ==QCSS_STOPPED || cpu_status ==QCSS_HARDWARE_ERROR )break ; -cpu_relax (); +udelay (100 ); } }
Hello Nick,
Thank you very much for reviewing this patch!
Quoted 17 lines · Nicholas Piggin
Thiago Jung Bauermann's on April 11, 2019 9:08 am:
Quoted 16 lines
Quoted 13 lines · Thiago Jung Bauermann
arch/powerpc/platforms/pseries/hotplug-cpu.c +5
-0@@ -214,13 +214,22 @@ static void pseries_cpu_die(unsigned int cpu)msleep (1 ); } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { +/* +* If the current state is not offline yet, it means that the +* dying CPU (which is either in pseries_mach_cpu_die() or in +* the process of getting there) didn't have a chance yet to +* call rtas_stop_self() and therefore it's too early to query
Quoted 8 lines
Quoted 4 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c static void pseries_cpu_die(unsigned int cpu) · line 225
Quoted 4 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c +4
-0+* if the CPU is stopped. +*/ +spin_event_timeout (get_cpu_current_state (cpu) ==CPU_STATE_OFFLINE , +100000 ,100 );If the CPU state does not go to offline here, you should give up and return online, right? Otherwise I think query-cpu-stopped-state can get confused by CPUs in idle and you get a false positive.
Can it get confused? My impression from reading the definition for query-cpu-stopped-state in the PAPR is that it will simply return a CPU_status value of 2 in that case, meaning that "the processor thread is not in the RTAS stopped state", but I don't know much about this.
Quoted 3 lines
That race can still happen, we would really need a sequence count check over current CPU state to ensure we got a race-free qcss value, but at least a check here should make the race implausible to hit.
Actually, since rtas_stop_self() panics if the processor fails to stop and also since callers of pseries_cpu_die()¹ already assume that it is going to succeed in stopping the CPU (given that the function returns void and can't signal an error), a more straightforward way of eliminating the race is to simply do this:
arch/powerpc/platforms/pseries/hotplug-cpu.c +1 -1
@@ -215,7 +215,7 @@ static void pseries_cpu_die(unsigned int cpu) } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { -for (tries =0 ; tries <25 ; tries++) { +while (true) { cpu_status =smp_query_cpu_stopped (pcpu);if (cpu_status ==QCSS_STOPPED || cpu_status ==QCSS_HARDWARE_ERROR )
What do you think?
--
Thiago Jung Bauermann
IBM Linux Technology Center
¹ dlpar_offline_cpu() and takedown_cpu() in generic code
Nicholas Piggin
reviewerThiago Jung Bauermann's on April 18, 2019 11:00 am:
Quoted 33 lines
Hello Nick,
Thank you very much for reviewing this patch!
Quoted 17 lines · Nicholas Piggin
Thiago Jung Bauermann's on April 11, 2019 9:08 am:
Quoted 16 lines
Quoted 13 lines · Thiago Jung Bauermann
arch/powerpc/platforms/pseries/hotplug-cpu.c +5
-0@@ -214,13 +214,22 @@ static void pseries_cpu_die(unsigned int cpu)msleep (1 ); } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { +/* +* If the current state is not offline yet, it means that the +* dying CPU (which is either in pseries_mach_cpu_die() or in +* the process of getting there) didn't have a chance yet to +* call rtas_stop_self() and therefore it's too early to query Quoted 8 lines
Quoted 4 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c static void pseries_cpu_die(unsigned int cpu) · line 225
Quoted 4 lines
arch/powerpc/platforms/pseries/hotplug-cpu.c +4
-0+* if the CPU is stopped. +*/ +spin_event_timeout (get_cpu_current_state (cpu) ==CPU_STATE_OFFLINE , +100000 ,100 );If the CPU state does not go to offline here, you should give up and return online, right? Otherwise I think query-cpu-stopped-state can get confused by CPUs in idle and you get a false positive.
Quoted 4 lines
Can it get confused? My impression from reading the definition for query-cpu-stopped-state in the PAPR is that it will simply return a
CPU_statusvalue of 2 in that case, meaning that "the processor thread is not in the RTAS stopped state", but I don't know much about this.
In QEMU (non-KVM) mode, qcss I think may get confused between H_CEDE and rtas-stop-self. KVM mode may be okay because H_CEDE is handled in the kernel.
Quoted 25 lines
Quoted 3 lines
That race can still happen, we would really need a sequence count check over current CPU state to ensure we got a race-free qcss value, but at least a check here should make the race implausible to hit.
Actually, since
rtas_stop_self()panics if the processor fails to stop and also since callers of pseries_cpu_die()¹ already assume that it is going to succeed in stopping the CPU (given that the function returns void and can't signal an error), a more straightforward way of eliminating the race is to simply do this:arch/powerpc/platforms/pseries/hotplug-cpu.c +1
-1@@ -215,7 +215,7 @@ static void pseries_cpu_die(unsigned int cpu) } }else if (get_preferred_offline_state (cpu) ==CPU_STATE_OFFLINE ) { -for (tries =0 ; tries <25 ; tries++) { +while (true) { cpu_status =smp_query_cpu_stopped (pcpu);if (cpu_status ==QCSS_STOPPED || cpu_status ==QCSS_HARDWARE_ERROR )
Quoted 1 line
What do you think?
Yeah I think that may be a good idea, just makes things much simpler.
Thanks,
Nick