mbox series

[v3,0/3] arm64: support page mapping percpu first chunk allocator

Message ID 20210809093750.131091-1-wangkefeng.wang@huawei.com (mailing list archive)
Headers show
Series arm64: support page mapping percpu first chunk allocator | expand

Message

Kefeng Wang Aug. 9, 2021, 9:37 a.m. UTC
Percpu embedded first chunk allocator is the firstly option, but it
could fails on ARM64, eg,
  "percpu: max_distance=0x5fcfdc640000 too large for vmalloc space 0x781fefff0000"
  "percpu: max_distance=0x600000540000 too large for vmalloc space 0x7dffb7ff0000"
  "percpu: max_distance=0x5fff9adb0000 too large for vmalloc space 0x5dffb7ff0000"

then we could meet "WARNING: CPU: 15 PID: 461 at vmalloc.c:3087 pcpu_get_vm_areas+0x488/0x838",
even the system could not boot successfully.

Let's implement page mapping percpu first chunk allocator as a fallback
to the embedding allocator to increase the robustness of the system.

Also fix a crash when both NEED_PER_CPU_PAGE_FIRST_CHUNK and KASAN_VMALLOC enabled.

Tested on ARM64 qemu with cmdline "percpu_alloc=page" based on v5.14-rc5.

v3:
- search for a range that fits instead of always picking the end from
  vmalloc area suggested by Catalin.
- use NUMA_NO_NODE to avoid "virt_to_phys used for non-linear address:"
  issue in arm64 kasan_populate_early_vm_area_shadow().
- add Acked-by: Marco Elver <elver@google.com> to patch v3

V2:
- fix build error when CONFIG_KASAN disabled, found by lkp@intel.com
- drop wrong __weak comment from kasan_populate_early_vm_area_shadow(),
  found by Marco Elver <elver@google.com>

Kefeng Wang (3):
  vmalloc: Choose a better start address in vm_area_register_early()
  arm64: Support page mapping percpu first chunk allocator
  kasan: arm64: Fix pcpu_page_first_chunk crash with KASAN_VMALLOC

 arch/arm64/Kconfig         |  4 ++
 arch/arm64/mm/kasan_init.c | 16 ++++++++
 drivers/base/arch_numa.c   | 82 +++++++++++++++++++++++++++++++++-----
 include/linux/kasan.h      |  6 +++
 mm/kasan/init.c            |  5 +++
 mm/vmalloc.c               | 17 +++++---
 6 files changed, 115 insertions(+), 15 deletions(-)

Comments

Kefeng Wang Aug. 12, 2021, 6:07 a.m. UTC | #1
Hi Catalin and Will,

The drivers/base/arch_numa.c is only shared by riscv and arm64,

and the change from patch2 won't broke riscv.

Could all patches be merged by arm64 tree? or any new comments?

Many thanks.

On 2021/8/9 17:37, Kefeng Wang wrote:
> Percpu embedded first chunk allocator is the firstly option, but it
> could fails on ARM64, eg,
>    "percpu: max_distance=0x5fcfdc640000 too large for vmalloc space 0x781fefff0000"
>    "percpu: max_distance=0x600000540000 too large for vmalloc space 0x7dffb7ff0000"
>    "percpu: max_distance=0x5fff9adb0000 too large for vmalloc space 0x5dffb7ff0000"
>
> then we could meet "WARNING: CPU: 15 PID: 461 at vmalloc.c:3087 pcpu_get_vm_areas+0x488/0x838",
> even the system could not boot successfully.
>
> Let's implement page mapping percpu first chunk allocator as a fallback
> to the embedding allocator to increase the robustness of the system.
>
> Also fix a crash when both NEED_PER_CPU_PAGE_FIRST_CHUNK and KASAN_VMALLOC enabled.
>
> Tested on ARM64 qemu with cmdline "percpu_alloc=page" based on v5.14-rc5.
>
> v3:
> - search for a range that fits instead of always picking the end from
>    vmalloc area suggested by Catalin.
> - use NUMA_NO_NODE to avoid "virt_to_phys used for non-linear address:"
>    issue in arm64 kasan_populate_early_vm_area_shadow().
> - add Acked-by: Marco Elver <elver@google.com> to patch v3
>
> V2:
> - fix build error when CONFIG_KASAN disabled, found by lkp@intel.com
> - drop wrong __weak comment from kasan_populate_early_vm_area_shadow(),
>    found by Marco Elver <elver@google.com>
>
> Kefeng Wang (3):
>    vmalloc: Choose a better start address in vm_area_register_early()
>    arm64: Support page mapping percpu first chunk allocator
>    kasan: arm64: Fix pcpu_page_first_chunk crash with KASAN_VMALLOC
>
>   arch/arm64/Kconfig         |  4 ++
>   arch/arm64/mm/kasan_init.c | 16 ++++++++
>   drivers/base/arch_numa.c   | 82 +++++++++++++++++++++++++++++++++-----
>   include/linux/kasan.h      |  6 +++
>   mm/kasan/init.c            |  5 +++
>   mm/vmalloc.c               | 17 +++++---
>   6 files changed, 115 insertions(+), 15 deletions(-)
>
Kefeng Wang Aug. 23, 2021, 2:29 a.m. UTC | #2
On 2021/8/12 14:07, Kefeng Wang wrote:
> Hi Catalin and Will,
>
> The drivers/base/arch_numa.c is only shared by riscv and arm64,
>
> and the change from patch2 won't broke riscv.
>
> Could all patches be merged by arm64 tree? or any new comments?

Kindly ping...

>
> Many thanks.
>
> On 2021/8/9 17:37, Kefeng Wang wrote:
>> Percpu embedded first chunk allocator is the firstly option, but it
>> could fails on ARM64, eg,
>>    "percpu: max_distance=0x5fcfdc640000 too large for vmalloc space 
>> 0x781fefff0000"
>>    "percpu: max_distance=0x600000540000 too large for vmalloc space 
>> 0x7dffb7ff0000"
>>    "percpu: max_distance=0x5fff9adb0000 too large for vmalloc space 
>> 0x5dffb7ff0000"
>>
>> then we could meet "WARNING: CPU: 15 PID: 461 at vmalloc.c:3087 
>> pcpu_get_vm_areas+0x488/0x838",
>> even the system could not boot successfully.
>>
>> Let's implement page mapping percpu first chunk allocator as a fallback
>> to the embedding allocator to increase the robustness of the system.
>>
>> Also fix a crash when both NEED_PER_CPU_PAGE_FIRST_CHUNK and 
>> KASAN_VMALLOC enabled.
>>
>> Tested on ARM64 qemu with cmdline "percpu_alloc=page" based on 
>> v5.14-rc5.
>>
>> v3:
>> - search for a range that fits instead of always picking the end from
>>    vmalloc area suggested by Catalin.
>> - use NUMA_NO_NODE to avoid "virt_to_phys used for non-linear address:"
>>    issue in arm64 kasan_populate_early_vm_area_shadow().
>> - add Acked-by: Marco Elver <elver@google.com> to patch v3
>>
>> V2:
>> - fix build error when CONFIG_KASAN disabled, found by lkp@intel.com
>> - drop wrong __weak comment from kasan_populate_early_vm_area_shadow(),
>>    found by Marco Elver <elver@google.com>
>>
>> Kefeng Wang (3):
>>    vmalloc: Choose a better start address in vm_area_register_early()
>>    arm64: Support page mapping percpu first chunk allocator
>>    kasan: arm64: Fix pcpu_page_first_chunk crash with KASAN_VMALLOC
>>
>>   arch/arm64/Kconfig         |  4 ++
>>   arch/arm64/mm/kasan_init.c | 16 ++++++++
>>   drivers/base/arch_numa.c   | 82 +++++++++++++++++++++++++++++++++-----
>>   include/linux/kasan.h      |  6 +++
>>   mm/kasan/init.c            |  5 +++
>>   mm/vmalloc.c               | 17 +++++---
>>   6 files changed, 115 insertions(+), 15 deletions(-)
>>
Catalin Marinas Aug. 25, 2021, 6:11 p.m. UTC | #3
On Thu, Aug 12, 2021 at 02:07:36PM +0800, Kefeng Wang wrote:
> The drivers/base/arch_numa.c is only shared by riscv and arm64,
> 
> and the change from patch2 won't broke riscv.
> 
> Could all patches be merged by arm64 tree? or any new comments?

The series touches drivers/ and mm/ but missing acks from both Greg and
Andrew (cc'ing them).

I'm also happy for the series to go in via the mm tree in case Andrew
wants to take it.