Re: [PATCH v2] powerpc/lockdep: fix a false positive warning

[Qian Cai]

> On Sep 7, 2019, at 3:05 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> 
> 
> * Qian Cai <cai@xxxxxx> wrote:
> 
>> The commit 108c14858b9e ("locking/lockdep: Add support for dynamic
>> keys") introduced a boot warning on powerpc below, because since the
>> commit 2d4f567103ff ("KVM: PPC: Introduce kvm_tmp framework") adds
>> kvm_tmp[] into the .bss section and then free the rest of unused spaces
>> back to the page allocator.
>> 
>> kernel_init
>>  kvm_guest_init
>>    kvm_free_tmp
>>      free_reserved_area
>>        free_unref_page
>>          free_unref_page_prepare
>> 
>> Later, alloc_workqueue() happens to allocate some pages from there and
>> trigger the warning at,
>> 
>> if (WARN_ON_ONCE(static_obj(key)))
>> 
>> Fix it by adding a generic helper arch_is_bss_hole() to skip those areas
>> in static_obj(). Since kvm_free_tmp() is only done early during the
>> boot, just go lockless to make the implementation simple for now.
>> 
>> WARNING: CPU: 0 PID: 13 at kernel/locking/lockdep.c:1120
>> Workqueue: events work_for_cpu_fn
>> Call Trace:
>>  lockdep_register_key+0x68/0x200
>>  wq_init_lockdep+0x40/0xc0
>>  trunc_msg+0x385f9/0x4c30f (unreliable)
>>  wq_init_lockdep+0x40/0xc0
>>  alloc_workqueue+0x1e0/0x620
>>  scsi_host_alloc+0x3d8/0x490
>>  ata_scsi_add_hosts+0xd0/0x220 [libata]
>>  ata_host_register+0x178/0x400 [libata]
>>  ata_host_activate+0x17c/0x210 [libata]
>>  ahci_host_activate+0x84/0x250 [libahci]
>>  ahci_init_one+0xc74/0xdc0 [ahci]
>>  local_pci_probe+0x78/0x100
>>  work_for_cpu_fn+0x40/0x70
>>  process_one_work+0x388/0x750
>>  process_scheduled_works+0x50/0x90
>>  worker_thread+0x3d0/0x570
>>  kthread+0x1b8/0x1e0
>>  ret_from_kernel_thread+0x5c/0x7c
>> 
>> Fixes: 108c14858b9e ("locking/lockdep: Add support for dynamic keys")
>> Signed-off-by: Qian Cai <cai@xxxxxx>
>> ---
>> 
>> v2: No need to actually define arch_is_bss_hole() powerpc64 only.
>> 
>> arch/powerpc/include/asm/sections.h | 11 +++++++++++
>> arch/powerpc/kernel/kvm.c           |  5 +++++
>> include/asm-generic/sections.h      |  7 +++++++
>> kernel/locking/lockdep.c            |  3 +++
>> 4 files changed, 26 insertions(+)
>> 
>> diff --git a/arch/powerpc/include/asm/sections.h b/arch/powerpc/include/asm/sections.h
>> index 4a1664a8658d..4f5d69c42017 100644
>> --- a/arch/powerpc/include/asm/sections.h
>> +++ b/arch/powerpc/include/asm/sections.h
>> @@ -5,8 +5,19 @@
>> 
>> #include <linux/elf.h>
>> #include <linux/uaccess.h>
>> +
>> +#define arch_is_bss_hole arch_is_bss_hole
>> +
>> #include <asm-generic/sections.h>
>> 
>> +extern void *bss_hole_start, *bss_hole_end;
>> +
>> +static inline int arch_is_bss_hole(unsigned long addr)
>> +{
>> +	return addr >= (unsigned long)bss_hole_start &&
>> +	       addr < (unsigned long)bss_hole_end;
>> +}
>> +
>> extern char __head_end[];
>> 
>> #ifdef __powerpc64__
>> diff --git a/arch/powerpc/kernel/kvm.c b/arch/powerpc/kernel/kvm.c
>> index b7b3a5e4e224..89e0e522e125 100644
>> --- a/arch/powerpc/kernel/kvm.c
>> +++ b/arch/powerpc/kernel/kvm.c
>> @@ -66,6 +66,7 @@
>> static bool kvm_patching_worked = true;
>> char kvm_tmp[1024 * 1024];
>> static int kvm_tmp_index;
>> +void *bss_hole_start, *bss_hole_end;
>> 
>> static inline void kvm_patch_ins(u32 *inst, u32 new_inst)
>> {
>> @@ -707,6 +708,10 @@ static __init void kvm_free_tmp(void)
>> 	 */
>> 	kmemleak_free_part(&kvm_tmp[kvm_tmp_index],
>> 			   ARRAY_SIZE(kvm_tmp) - kvm_tmp_index);
>> +
>> +	bss_hole_start = &kvm_tmp[kvm_tmp_index];
>> +	bss_hole_end = &kvm_tmp[ARRAY_SIZE(kvm_tmp)];
>> +
>> 	free_reserved_area(&kvm_tmp[kvm_tmp_index],
>> 			   &kvm_tmp[ARRAY_SIZE(kvm_tmp)], -1, NULL);
>> }
>> diff --git a/include/asm-generic/sections.h b/include/asm-generic/sections.h
>> index d1779d442aa5..4d8b1f2c5fd9 100644
>> --- a/include/asm-generic/sections.h
>> +++ b/include/asm-generic/sections.h
>> @@ -91,6 +91,13 @@ static inline int arch_is_kernel_initmem_freed(unsigned long addr)
>> }
>> #endif
>> 
>> +#ifndef arch_is_bss_hole
>> +static inline int arch_is_bss_hole(unsigned long addr)
>> +{
>> +	return 0;
>> +}
>> +#endif
>> +
>> /**
>>  * memory_contains - checks if an object is contained within a memory region
>>  * @begin: virtual address of the beginning of the memory region
>> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
>> index 4861cf8e274b..cd75b51f15ce 100644
>> --- a/kernel/locking/lockdep.c
>> +++ b/kernel/locking/lockdep.c
>> @@ -675,6 +675,9 @@ static int static_obj(const void *obj)
>> 	if (arch_is_kernel_initmem_freed(addr))
>> 		return 0;
>> 
>> +	if (arch_is_bss_hole(addr))
>> +		return 0;
> 
> arch_is_bss_hole() should use a 'bool' - but other than that, this 
> looks good to me, if the PowerPC maintainers agree too.

I thought about making it a bool in the first place, but since all other similar helpers
(arch_is_kernel_initmem_freed(), arch_is_kernel_text(), arch_is_kernel_data() etc)
could be bool too but are not, I kept arch_is_bss_hole() just to be âintâ for consistent.

Although then there is is_kernel_rodata() which is bool. I suppose Iâll change
arch_is_bss_hole() to bool, and then could have a follow-up patch to covert all similar
helpers to return boo instead.