Skip to contentSkip to conversation

ppc: fix vcpu hotunplug leak in spapr_realize_vcpu

PATCH Superseded Daniel Henrique Barboza 3 participants 6 messages Mar 28, 2022 – Mar 29, 2022

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover ppc: fix vcpu hotunplug leak in spapr_realize_vcpu

Hi,

This is a memory leak found by Valgrind when testing vcpu hotplug/unplug in pSeries guests.

Other vcpu hotplug/unplug leaks are still present in the common code (one in the KVM thread loop and another in cpu_address_space via cpu->cpu_ases) but these are already being handled by Mark Kanda and Phillipe.

Daniel Henrique Barboza (2):
hw/ppc/ppc.c: add cpu_ppc_tb_free() hw/ppc: free env->tb_env in spapr_unrealize_vcpu()

 hw/ppc/ppc.c            | 7 +++++++
 hw/ppc/spapr_cpu_core.c | 3 +++
 include/hw/ppc/ppc.h    | 1 +
 3 files changed, 11 insertions(+)
Signature · 1 line
2.35.1
#2 lore raw

PATCH 2/2 hw/ppc: free env->tb_env in spapr_unrealize_vcpu()

The timebase is allocated during spapr_realize_vcpu() and it's not freed. This results in memory leaks when doing vcpu unplugs:

==636935==
==636935== 144 (96 direct, 48 indirect) bytes in 1 blocks are definitely lost in loss record 6 ,461 of 8,135

==636935==    at 0x4897468: calloc (vg_replace_malloc.c:760)
==636935==    by 0x5077213: g_malloc0 (in /usr/lib64/libglib-2.0.so.0.6400.4)
==636935==    by 0x507757F: g_malloc0_n (in /usr/lib64/libglib-2.0.so.0.6400.4)
==636935==    by 0x93C3FB: cpu_ppc_tb_init (ppc.c:1066)
==636935==    by 0x97BC2B: spapr_realize_vcpu (spapr_cpu_core.c:268)
==636935==    by 0x97C01F: spapr_cpu_core_realize (spapr_cpu_core.c:337)
==636935==    by 0xD4626F: device_set_realized (qdev.c:531)
==636935==    by 0xD55273: property_set_bool (object.c:2273)
==636935==    by 0xD523DF: object_property_set (object.c:1408)
==636935==    by 0xD588B7: object_property_set_qobject (qom-qobject.c:28)
==636935==    by 0xD52897: object_property_set_bool (object.c:1477)
==636935==    by 0xD4579B: qdev_realize (qdev.c:333)

==636935==

Fix it by calling cpu_ppc_tb_free() in spapr_unrealize_vcpu().

Fixes 6f4b5c3ec590 ("spapr: CPU hot unplug support")
---
 hw/ppc/spapr_cpu_core.c | 3 +++
 1 file changed, 3 insertions(+)
hw/ppc/spapr_cpu_core.c +3 -0
index ed84713960..8a4861f45a 100644
@@ -189,10 +189,13 @@ static const VMStateDescription vmstate_spapr_cpu_state = {
 
 static void spapr_unrealize_vcpu(PowerPCCPU *cpu, SpaprCpuCore *sc)
 {
+    CPUPPCState *env = &cpu->env;
+
     if (!sc->pre_3_0_migration) {
         vmstate_unregister(NULL, &vmstate_spapr_cpu_state, cpu->machine_data);
     }
     spapr_irq_cpu_intc_destroy(SPAPR_MACHINE(qdev_get_machine()), cpu);
+    cpu_ppc_tb_free(env);
     qdev_unrealize(DEVICE(cpu));
 }
 
Signature · 1 line
2.35.1
#1 lore raw

PATCH 1/2 hw/ppc/ppc.c: add cpu_ppc_tb_free()

In reply to Daniel Henrique Barboza

The timebase is created using cpu_ppc_tb_init(). At this moment we don't have a way to free it. Add cpu_ppc_tb_free() to allow env->tb_env to be freed.

---
 hw/ppc/ppc.c         | 7 +++++++
 include/hw/ppc/ppc.h | 1 +
 2 files changed, 8 insertions(+)
hw/ppc/ppc.c +7 -0
index faa02d6710..fea70df45e 100644
@@ -1083,6 +1083,13 @@ clk_setup_cb cpu_ppc_tb_init (CPUPPCState *env, uint32_t freq)
     return &cpu_ppc_set_tb_clk;
 }
 
+void cpu_ppc_tb_free(CPUPPCState *env)
+{
+    timer_free(env->tb_env->decr_timer);
+    timer_free(env->tb_env->hdecr_timer);
+    g_free(env->tb_env);
+}
+
 /* cpu_ppc_hdecr_init may be used if the timer is not used by HDEC emulation */
 void cpu_ppc_hdecr_init(CPUPPCState *env)
 {
include/hw/ppc/ppc.h +1 -0
index b0ba4bd6b9..364f165b4b 100644
@@ -54,6 +54,7 @@ struct ppc_tb_t {
 
 uint64_t cpu_ppc_get_tb(ppc_tb_t *tb_env, uint64_t vmclk, int64_t tb_offset);
 clk_setup_cb cpu_ppc_tb_init (CPUPPCState *env, uint32_t freq);
+void cpu_ppc_tb_free(CPUPPCState *env);
 void cpu_ppc_hdecr_init(CPUPPCState *env);
 void cpu_ppc_hdecr_exit(CPUPPCState *env);
 
Signature · 1 line
2.35.1
#3 lore raw
In reply to Daniel Henrique Barboza
Quoted 9 lines · Daniel Henrique Barboza

Hi,

This is a memory leak found by Valgrind when testing vcpu hotplug/unplug in pSeries guests.

Other vcpu hotplug/unplug leaks are still present in the common code (one in the KVM thread loop and another in cpu_address_space via cpu->cpu_ases) but these are already being handled by Mark Kanda and Phillipe.

Changes LGTM, but I don't see much reason to split this into two patches. They're both small, and are part of the same logical change.

Quoted 11 lines

Daniel Henrique Barboza (2):
hw/ppc/ppc.c: add cpu_ppc_tb_free() hw/ppc: free env->tb_env in spapr_unrealize_vcpu()

 hw/ppc/ppc.c            | 7 +++++++
 hw/ppc/spapr_cpu_core.c | 3 +++
 include/hw/ppc/ppc.h    | 1 +
 3 files changed, 11 insertions(+)
Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
#4 lore raw
Quoted 6 lines · David Gibson
Quoted 5 lines · Daniel Henrique Barboza

Hi,

This is a memory leak found by Valgrind when testing vcpu hotplug/unplug in pSeries guests.

Quoted 7 lines
Quoted 4 lines

Other vcpu hotplug/unplug leaks are still present in the common code (one in the KVM thread loop and another in cpu_address_space via cpu->cpu_ases) but these are already being handled by Mark Kanda and Phillipe.

Changes LGTM, but I don't see much reason to split this into two patches. They're both small, and are part of the same logical change.

And it could be a 7.0 candidate. Are we ok with that ?

Thanks,

C.

#5 lore raw
Quoted 7 lines · Cédric Le Goater
Quoted 6 lines · David Gibson
Quoted 5 lines · Daniel Henrique Barboza

Hi,

This is a memory leak found by Valgrind when testing vcpu hotplug/unplug in pSeries guests.

Quoted 7 lines
Quoted 7 lines
Quoted 4 lines

Other vcpu hotplug/unplug leaks are still present in the common code (one in the KVM thread loop and another in cpu_address_space via cpu->cpu_ases) but these are already being handled by Mark Kanda and Phillipe.

Changes LGTM, but I don't see much reason to split this into two patches.  They're both small, and are part of the same logical change.

I did it in separated patches because I tried to find other instances where the timebase would need to be freed. Didn't find any.

I am ok with squashing them in a single patch. I'll send a v2.

Quoted 2 lines

And it could be a 7.0 candidate. Are we ok with that ?

Since it's a memory leak we are now aware of and have a fix for, yeah, I think it qualifies for 7.0.

Daniel

Quoted 4 lines

Thanks,

C.