arm64: paravirt: Initialize steal time when cpu is online
Subsystems (4)
PATCH 1/1 arm64: paravirt: Initialize steal time when cpu is online
Steal time initialization requires mapping a memory region which invokes a memory allocation. Doing this at CPU starting time results in the following trace when CONFIG_DEBUG_ATOMIC_SLEEP is enabled:
BUG: sleeping function called from invalid context at mm/slab.h:498 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1 CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.9.0-rc5+ #1 Call trace: dump_backtrace+0x0/0x208 show_stack+0x1c/0x28 dump_stack+0xc4/0x11c ___might_sleep+0xf8/0x130 __might_sleep+0x58/0x90 slab_pre_alloc_hook.constprop.101+0xd0/0x118 kmem_cache_alloc_node_trace+0x84/0x270 __get_vm_area_node+0x88/0x210 get_vm_area_caller+0x38/0x40 __ioremap_caller+0x70/0xf8 ioremap_cache+0x78/0xb0 memremap+0x9c/0x1a8 init_stolen_time_cpu+0x54/0xf0 cpuhp_invoke_callback+0xa8/0x720 notify_cpu_starting+0xc8/0xd8 secondary_start_kernel+0x114/0x180
CPU1: Booted secondary processor 0x0000000001 [0x431f0a11]
However we don't need to initialize steal time at CPU starting time. We can simply wait until CPU online time, just sacrificing a bit of accuracy by returning zero for steal time until we know better.
While at it, add __init to the functions that are only called by pv_time_init() which is __init.
--- arch/arm64/kernel/paravirt.c | 26 +++++++++++++++----------- include/linux/cpuhotplug.h | 1 - 2 files changed, 15 insertions(+), 12 deletions(-)
arch/arm64/kernel/paravirt.c +15 -11
@@ -50,16 +50,19 @@ static u64 pv_steal_clock(int cpu)struct pv_time_stolen_time_region *reg; reg =per_cpu_ptr (&stolen_time_region, cpu); -if (!reg->kaddr) { -pr_warn_once ("stolen time enabled but not configured for cpu %d\n" , - cpu); + +/* +* paravirt_steal_clock() may be called before the CPU +* online notification callback runs. Until the callback +* has run we just return zero. +*/ +if (!reg->kaddr)return 0 ; - }return le64_to_cpu (READ_ONCE (reg->kaddr->stolen_time)); } -static int stolen_time_dying_cpu (unsigned int cpu) +static int stolen_time_cpu_down_prepare (unsigned int cpu) {struct pv_time_stolen_time_region *reg;
@@ -73,7 +76,7 @@ static int stolen_time_dying_cpu(unsigned int cpu)return 0 ; } -static int init_stolen_time_cpu (unsigned int cpu) +static int stolen_time_cpu_online (unsigned int cpu) {struct pv_time_stolen_time_region *reg;struct arm_smccc_res res;
@@ -103,19 +106,20 @@ static int init_stolen_time_cpu(unsigned int cpu)return 0 ; } -static int pv_time_init_stolen_time (void ) +static int __init pv_time_init_stolen_time (void ) {int ret; - ret =cpuhp_setup_state (CPUHP_AP_ARM_KVMPV_STARTING , -"hypervisor/arm/pvtime:starting" , - init_stolen_time_cpu, stolen_time_dying_cpu); + ret =cpuhp_setup_state (CPUHP_AP_ONLINE_DYN , +"hypervisor/arm/pvtime:online" , + stolen_time_cpu_online, + stolen_time_cpu_down_prepare);if (ret <0 )return ret;return 0 ; } -static bool has_pv_steal_clock (void ) +static bool __init has_pv_steal_clock (void ) {struct arm_smccc_res res;
include/linux/cpuhotplug.h +0 -1
@@ -142,7 +142,6 @@ enum cpuhp_state {/* Must be the last timer callback */ CPUHP_AP_DUMMY_TIMER_STARTING ,CPUHP_AP_ARM_XEN_STARTING , -CPUHP_AP_ARM_KVMPV_STARTING ,CPUHP_AP_ARM_CORESIGHT_STARTING ,CPUHP_AP_ARM_CORESIGHT_CTI_STARTING , CPUHP_AP_ARM64_ISNDEP_STARTING,
Signature · 6 lines
2.27.0 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 26 lines · Andrew Jones
Steal time initialization requires mapping a memory region which invokes a memory allocation. Doing this at CPU starting time results in the following trace when
CONFIG_DEBUG_ATOMIC_SLEEPis enabled:BUG: sleeping function called from invalid context at mm/slab.h:498 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1 CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.9.0-rc5+ #1 Call trace: dump_backtrace+0x0/0x208 show_stack+0x1c/0x28 dump_stack+0xc4/0x11c ___might_sleep+0xf8/0x130 __might_sleep+0x58/0x90 slab_pre_alloc_hook.constprop.101+0xd0/0x118 kmem_cache_alloc_node_trace+0x84/0x270 __get_vm_area_node+0x88/0x210 get_vm_area_caller+0x38/0x40 __ioremap_caller+0x70/0xf8 ioremap_cache+0x78/0xb0 memremap+0x9c/0x1a8 init_stolen_time_cpu+0x54/0xf0 cpuhp_invoke_callback+0xa8/0x720 notify_cpu_starting+0xc8/0xd8 secondary_start_kernel+0x114/0x180CPU1: Booted secondary processor
0x0000000001[0x431f0a11]
Quoted 8 lines
However we don't need to initialize steal time at CPU starting time. We can simply wait until CPU online time, just sacrificing a bit of accuracy by returning zero for steal time until we know better.
While at it, add
__initto the functions that are only called bypv_time_init()which is__init.Signed-off-by Andrew Jones <[email protected]>
The changes look good to me - it simplifies things delaying the startup of stolen time and I doubt anyone will notice the accuracy loss.
Do we need a CC: stable (or Fixes:)?
Steve
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainer+ Catalin,
Quoted 26 lines · Andrew Jones
Steal time initialization requires mapping a memory region which invokes a memory allocation. Doing this at CPU starting time results in the following trace when
CONFIG_DEBUG_ATOMIC_SLEEPis enabled:BUG: sleeping function called from invalid context at mm/slab.h:498 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1 CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.9.0-rc5+ #1 Call trace: dump_backtrace+0x0/0x208 show_stack+0x1c/0x28 dump_stack+0xc4/0x11c ___might_sleep+0xf8/0x130 __might_sleep+0x58/0x90 slab_pre_alloc_hook.constprop.101+0xd0/0x118 kmem_cache_alloc_node_trace+0x84/0x270 __get_vm_area_node+0x88/0x210 get_vm_area_caller+0x38/0x40 __ioremap_caller+0x70/0xf8 ioremap_cache+0x78/0xb0 memremap+0x9c/0x1a8 init_stolen_time_cpu+0x54/0xf0 cpuhp_invoke_callback+0xa8/0x720 notify_cpu_starting+0xc8/0xd8 secondary_start_kernel+0x114/0x180CPU1: Booted secondary processor
0x0000000001[0x431f0a11]
Quoted 8 lines
However we don't need to initialize steal time at CPU starting time. We can simply wait until CPU online time, just sacrificing a bit of accuracy by returning zero for steal time until we know better.
While at it, add
__initto the functions that are only called bypv_time_init()which is__init.Signed-off-by Andrew Jones <[email protected]>
Catalin, any chance you could queue this if you are OK with it?
Thanks,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 24 lines · Andrew Jones
Steal time initialization requires mapping a memory region which invokes a memory allocation. Doing this at CPU starting time results in the following trace when
CONFIG_DEBUG_ATOMIC_SLEEPis enabled:BUG: sleeping function called from invalid context at mm/slab.h:498 in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 0, name: swapper/1 CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.9.0-rc5+ #1 Call trace: dump_backtrace+0x0/0x208 show_stack+0x1c/0x28 dump_stack+0xc4/0x11c ___might_sleep+0xf8/0x130 __might_sleep+0x58/0x90 slab_pre_alloc_hook.constprop.101+0xd0/0x118 kmem_cache_alloc_node_trace+0x84/0x270 __get_vm_area_node+0x88/0x210 get_vm_area_caller+0x38/0x40 __ioremap_caller+0x70/0xf8 ioremap_cache+0x78/0xb0 memremap+0x9c/0x1a8 init_stolen_time_cpu+0x54/0xf0 cpuhp_invoke_callback+0xa8/0x720 notify_cpu_starting+0xc8/0xd8 secondary_start_kernel+0x114/0x180
Quoted 3 lines
CPU1: Booted secondary processor
0x0000000001[0x431f0a11][...]
Applied to arm64 (for-next/fixes), thanks!
[1/1] arm64: paravirt: Initialize steal time when cpu is online
https://git.kernel.org/arm64/c/75df529bec91
Signature · 6 lines
Catalin _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm