mbox series

[v4,0/3] arm64: support page mapping percpu first chunk allocator

Message ID 20210910053354.26721-1-wangkefeng.wang@huawei.com (mailing list archive)
Headers show
Series arm64: support page mapping percpu first chunk allocator | expand

Message

Kefeng Wang Sept. 10, 2021, 5:33 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.

V4:
- add ACK/RB
- address comments about patch1 from Catalin
- add Greg and Andrew into list suggested by Catalin

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               | 19 ++++++---
 6 files changed, 116 insertions(+), 16 deletions(-)

Comments

Kefeng Wang Sept. 15, 2021, 8:33 a.m. UTC | #1
Hi Greg and Andrew, as Catalin saids,the series touches drivers/ and mm/ 
but missing

acks from both of you,could you take a look of this patchset(patch1 
change mm/vmalloc.c

and patch2 changes drivers/base/arch_numa.c).

And Catalin, is there any other comments? I hope this could be merged 
into next version,

Many thanks all of you.

On 2021/9/10 13:33, 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.
>
> V4:
> - add ACK/RB
> - address comments about patch1 from Catalin
> - add Greg and Andrew into list suggested by Catalin
>
> 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               | 19 ++++++---
>   6 files changed, 116 insertions(+), 16 deletions(-)
>
Greg KH Sept. 16, 2021, 3:41 p.m. UTC | #2
On Wed, Sep 15, 2021 at 04:33:09PM +0800, Kefeng Wang wrote:
> Hi Greg and Andrew, as Catalin saids,the series touches drivers/ and mm/
> but missing
> 
> acks from both of you,could you take a look of this patchset(patch1 change
> mm/vmalloc.c

What patchset?

> and patch2 changes drivers/base/arch_numa.c).

that file is not really owned by anyone it seems :(

Can you provide a link to the real patch please?

thanks,

greg k-h
Greg KH Sept. 17, 2021, 6:24 a.m. UTC | #3
On Fri, Sep 17, 2021 at 09:11:38AM +0800, Kefeng Wang wrote:
> 
> On 2021/9/16 23:41, Greg KH wrote:
> > On Wed, Sep 15, 2021 at 04:33:09PM +0800, Kefeng Wang wrote:
> > > Hi Greg and Andrew, as Catalin saids,the series touches drivers/ and mm/
> > > but missing
> > > 
> > > acks from both of you,could you take a look of this patchset(patch1 change
> > > mm/vmalloc.c
> > What patchset?
> 
> [PATCH v4 1/3] vmalloc: Choose a better start address in
> vm_area_register_early()  <https://lore.kernel.org/linux-arm-kernel/20210910053354.26721-2-wangkefeng.wang@huawei.com/>
> [PATCH v4 2/3] arm64: Support page mapping percpu first chunk allocator  <https://lore.kernel.org/linux-arm-kernel/20210910053354.26721-3-wangkefeng.wang@huawei.com/>
> [PATCH v4 3/3] kasan: arm64: Fix pcpu_page_first_chunk crash with
> KASAN_VMALLOC  <https://lore.kernel.org/linux-arm-kernel/20210910053354.26721-4-wangkefeng.wang@huawei.com/>
> [PATCH v4 0/3] arm64: support page mapping percpu first chunk allocator  <https://lore.kernel.org/linux-arm-kernel/c06faf6c-3d21-04f2-6855-95c86e96cf5a@huawei.com/>
> 
> > > and patch2 changes drivers/base/arch_numa.c).
> patch2 :
> 
> [PATCH v4 2/3] arm64: Support page mapping percpu first chunk allocator  <https://lore.kernel.org/linux-arm-kernel/20210910053354.26721-3-wangkefeng.wang@huawei.com/#r>
> 
> > that file is not really owned by anyone it seems :(
> > 
> > Can you provide a link to the real patch please?
> 
> Yes, arch_numa.c is moved into drivers/base to support riscv numa, it is
> shared by arm64/riscv,
> 
> my changes(patch2) only support NEED_PER_CPU_PAGE_FIRST_CHUNK on ARM64.
> 
> here is the link:
> 
> https://lore.kernel.org/linux-arm-kernel/20210910053354.26721-1-wangkefeng.wang@huawei.com/

Now reviewed.
Kefeng Wang Sept. 28, 2021, 7:48 a.m. UTC | #4
Hi Catalin and Andrew, kindly ping again, any comments, thanks.

On 2021/9/10 13:33, 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.
> 
> V4:
> - add ACK/RB
> - address comments about patch1 from Catalin
> - add Greg and Andrew into list suggested by Catalin
> 
> 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               | 19 ++++++---
>   6 files changed, 116 insertions(+), 16 deletions(-)
>
Kefeng Wang Oct. 8, 2021, 1:33 p.m. UTC | #5
On 2021/9/28 15:48, Kefeng Wang wrote:
> Hi Catalin and Andrew, kindly ping again, any comments, thanks.

Looks no more comments, Catalin and Andrew, ping again, any one of you
could merge this patchset, many thanks.

> 
> On 2021/9/10 13:33, 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.
>>
>> V4:
>> - add ACK/RB
>> - address comments about patch1 from Catalin
>> - add Greg and Andrew into list suggested by Catalin
>>
>> 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               | 19 ++++++---
>>   6 files changed, 116 insertions(+), 16 deletions(-)
>>
Andrew Morton Oct. 10, 2021, 9:36 p.m. UTC | #6
On Fri, 10 Sep 2021 13:33:51 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> 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.

How serious are these problems in real-world situations?  Do people
feel that a -stable backport is needed, or is a 5.16-rc1 merge
sufficient?
Kefeng Wang Oct. 11, 2021, 1:09 a.m. UTC | #7
On 2021/10/11 5:36, Andrew Morton wrote:
> On Fri, 10 Sep 2021 13:33:51 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> 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.
> 
> How serious are these problems in real-world situations?  Do people
> feel that a -stable backport is needed, or is a 5.16-rc1 merge
> sufficient?
> .
Thanks Andrew.

A specific memory layout is required(also with KASAN enabled), we met 
this issue at qemu and real hardware, due to KASAN enabled, so I think
5.16-rc1 is sufficient.