mbox series

[v3,0/6] introduce mirrored memory support for arm64

Message ID 20220607093805.1354256-1-mawupeng1@huawei.com (mailing list archive)
Headers show
Series introduce mirrored memory support for arm64 | expand

Message

mawupeng June 7, 2022, 9:37 a.m. UTC
From: Ma Wupeng <mawupeng1@huawei.com>

Commit b05b9f5f9dcf ("x86, mirror: x86 enabling - find mirrored memory ranges")
introduced mirrored memory support for x86. This support rely on UEFI to
report mirrored memory address ranges.  See UEFI 2.5 spec pages 157-158:

  http://www.uefi.org/sites/default/files/resources/UEFI%202_5.pdf

Memory mirroring is a technique used to separate memory into two separate
channels, usually on a memory device, like a server. In memory mirroring,
one channel is copied to another to create redundancy. This method makes
input/output (I/O) registers and memory appear with more than one address
range because the same physical byte is accessible at more than one
address. Using memory mirroring, higher memory reliability and a higher
level of memory consolidation are possible.

These EFI memory regions have various attributes, and the "mirrored"
attribute is one of them. The physical memory region whose descriptors
in EFI memory map has EFI_MEMORY_MORE_RELIABLE attribute (bit: 16) are
mirrored. The address range mirroring feature of the kernel arranges such
mirrored regions into normal zones and other regions into movable zones.

Arm64 can support this too. So mirrored memory support is added to support
arm64.

The main purpose of this patch set is to introduce mirrored support for
arm64 and we have already fixed the problems we had which is shown in
patch #5 to patch #8 and try to bring total isolation in patch #9 which
will disable mirror feature if kernelcore is not specified.

In order to test this support in arm64:
- patch this patch set
- add kernelcore=mirror in kernel parameter
- start you kernel

Patch #1-#2 introduce mirrored memory support form arm64.
Patch #3-#5 fix some bugs for arm64 if memory reliable is enabled.
Patch #6 disable mirror feature if kernelcore is not specified.

Thanks to Ard Biesheuvel's hard work [1], now kernel will perfer mirrored
memory if kaslr is enabled.

[1] https://lore.kernel.org/linux-arm-kernel/CAMj1kXEPVEzMgOM4+Yj6PxHA-jFuDOAUdDJSiSxy_XaP4P7LSw@mail.gmail.com/T/

Changelog since v2:
- remove efi_fake_mem support
- remove Commit ("remove some redundant code in ia64 efi_init") since
  efi_print_memmap() is not public
- add mirror flag back on initrd memory

Changelog since v1:
- update changelog in cover letter
- use PHYS_PFN in patch #7

Ma Wupeng (6):
  efi: Make efi_find_mirror() public
  arm64/mirror: arm64 enabling - find mirrored memory ranges
  mm: Ratelimited mirrored memory related warning messages
  mm: Demote warning message in vmemmap_verify() to debug level
  mm: Add mirror flag back on initrd memory
  efi: Disable mirror feature if kernelcore is not specified

 .../admin-guide/kernel-parameters.txt         |  2 +-
 arch/arm64/kernel/setup.c                     |  1 +
 arch/arm64/mm/init.c                          |  9 +++++++
 arch/x86/include/asm/efi.h                    |  4 ---
 arch/x86/platform/efi/efi.c                   | 23 ----------------
 drivers/firmware/efi/efi.c                    | 26 +++++++++++++++++++
 include/linux/efi.h                           |  3 +++
 include/linux/memblock.h                      |  1 +
 include/linux/mm.h                            |  2 ++
 mm/memblock.c                                 | 24 +++++++++++++++--
 mm/page_alloc.c                               |  2 +-
 mm/sparse-vmemmap.c                           |  2 +-
 12 files changed, 67 insertions(+), 32 deletions(-)

Comments

Ard Biesheuvel June 10, 2022, 11:23 a.m. UTC | #1
On Tue, 7 Jun 2022 at 11:16, Wupeng Ma <mawupeng1@huawei.com> wrote:
>
> From: Ma Wupeng <mawupeng1@huawei.com>
>
> Commit b05b9f5f9dcf ("x86, mirror: x86 enabling - find mirrored memory ranges")
> introduced mirrored memory support for x86. This support rely on UEFI to
> report mirrored memory address ranges.  See UEFI 2.5 spec pages 157-158:
>
>   http://www.uefi.org/sites/default/files/resources/UEFI%202_5.pdf
>
> Memory mirroring is a technique used to separate memory into two separate
> channels, usually on a memory device, like a server. In memory mirroring,
> one channel is copied to another to create redundancy. This method makes
> input/output (I/O) registers and memory appear with more than one address
> range because the same physical byte is accessible at more than one
> address. Using memory mirroring, higher memory reliability and a higher
> level of memory consolidation are possible.
>
> These EFI memory regions have various attributes, and the "mirrored"
> attribute is one of them. The physical memory region whose descriptors
> in EFI memory map has EFI_MEMORY_MORE_RELIABLE attribute (bit: 16) are
> mirrored. The address range mirroring feature of the kernel arranges such
> mirrored regions into normal zones and other regions into movable zones.
>
> Arm64 can support this too. So mirrored memory support is added to support
> arm64.
>
> The main purpose of this patch set is to introduce mirrored support for
> arm64 and we have already fixed the problems we had which is shown in
> patch #5 to patch #8 and try to bring total isolation in patch #9 which
> will disable mirror feature if kernelcore is not specified.
>
> In order to test this support in arm64:
> - patch this patch set
> - add kernelcore=mirror in kernel parameter
> - start you kernel
>
> Patch #1-#2 introduce mirrored memory support form arm64.
> Patch #3-#5 fix some bugs for arm64 if memory reliable is enabled.
> Patch #6 disable mirror feature if kernelcore is not specified.
>
> Thanks to Ard Biesheuvel's hard work [1], now kernel will perfer mirrored
> memory if kaslr is enabled.
>
> [1] https://lore.kernel.org/linux-arm-kernel/CAMj1kXEPVEzMgOM4+Yj6PxHA-jFuDOAUdDJSiSxy_XaP4P7LSw@mail.gmail.com/T/
>
> Changelog since v2:
> - remove efi_fake_mem support
> - remove Commit ("remove some redundant code in ia64 efi_init") since
>   efi_print_memmap() is not public
> - add mirror flag back on initrd memory
>
> Changelog since v1:
> - update changelog in cover letter
> - use PHYS_PFN in patch #7
>
> Ma Wupeng (6):
>   efi: Make efi_find_mirror() public
>   arm64/mirror: arm64 enabling - find mirrored memory ranges
>   mm: Ratelimited mirrored memory related warning messages
>   mm: Demote warning message in vmemmap_verify() to debug level
>   mm: Add mirror flag back on initrd memory
>   efi: Disable mirror feature if kernelcore is not specified
>

I have tested these changes on QEMU/arm64 with the patch below, and
things seem to work as expected. We have some minor issues to work out
but the general shape of this code is good.

As for the mm/ changes: does anyone mind if I take those through the
EFI tree as well? I don't think the EFI and -mm changes depend on each
other, so they can go into -mm separately as well.
Ard Biesheuvel June 10, 2022, 11:24 a.m. UTC | #2
On Fri, 10 Jun 2022 at 13:23, Ard Biesheuvel <ardb@kernel.org> wrote:
>
> On Tue, 7 Jun 2022 at 11:16, Wupeng Ma <mawupeng1@huawei.com> wrote:
> >
> > From: Ma Wupeng <mawupeng1@huawei.com>
> >
> > Commit b05b9f5f9dcf ("x86, mirror: x86 enabling - find mirrored memory ranges")
> > introduced mirrored memory support for x86. This support rely on UEFI to
> > report mirrored memory address ranges.  See UEFI 2.5 spec pages 157-158:
> >
> >   http://www.uefi.org/sites/default/files/resources/UEFI%202_5.pdf
> >
> > Memory mirroring is a technique used to separate memory into two separate
> > channels, usually on a memory device, like a server. In memory mirroring,
> > one channel is copied to another to create redundancy. This method makes
> > input/output (I/O) registers and memory appear with more than one address
> > range because the same physical byte is accessible at more than one
> > address. Using memory mirroring, higher memory reliability and a higher
> > level of memory consolidation are possible.
> >
> > These EFI memory regions have various attributes, and the "mirrored"
> > attribute is one of them. The physical memory region whose descriptors
> > in EFI memory map has EFI_MEMORY_MORE_RELIABLE attribute (bit: 16) are
> > mirrored. The address range mirroring feature of the kernel arranges such
> > mirrored regions into normal zones and other regions into movable zones.
> >
> > Arm64 can support this too. So mirrored memory support is added to support
> > arm64.
> >
> > The main purpose of this patch set is to introduce mirrored support for
> > arm64 and we have already fixed the problems we had which is shown in
> > patch #5 to patch #8 and try to bring total isolation in patch #9 which
> > will disable mirror feature if kernelcore is not specified.
> >
> > In order to test this support in arm64:
> > - patch this patch set
> > - add kernelcore=mirror in kernel parameter
> > - start you kernel
> >
> > Patch #1-#2 introduce mirrored memory support form arm64.
> > Patch #3-#5 fix some bugs for arm64 if memory reliable is enabled.
> > Patch #6 disable mirror feature if kernelcore is not specified.
> >
> > Thanks to Ard Biesheuvel's hard work [1], now kernel will perfer mirrored
> > memory if kaslr is enabled.
> >
> > [1] https://lore.kernel.org/linux-arm-kernel/CAMj1kXEPVEzMgOM4+Yj6PxHA-jFuDOAUdDJSiSxy_XaP4P7LSw@mail.gmail.com/T/
> >
> > Changelog since v2:
> > - remove efi_fake_mem support
> > - remove Commit ("remove some redundant code in ia64 efi_init") since
> >   efi_print_memmap() is not public
> > - add mirror flag back on initrd memory
> >
> > Changelog since v1:
> > - update changelog in cover letter
> > - use PHYS_PFN in patch #7
> >
> > Ma Wupeng (6):
> >   efi: Make efi_find_mirror() public
> >   arm64/mirror: arm64 enabling - find mirrored memory ranges
> >   mm: Ratelimited mirrored memory related warning messages
> >   mm: Demote warning message in vmemmap_verify() to debug level
> >   mm: Add mirror flag back on initrd memory
> >   efi: Disable mirror feature if kernelcore is not specified
> >
>
> I have tested these changes on QEMU/arm64 with the patch below, and
> things seem to work as expected. We have some minor issues to work out
> but the general shape of this code is good.
>
> As for the mm/ changes: does anyone mind if I take those through the
> EFI tree as well? I don't think the EFI and -mm changes depend on each
> other, so they can go into -mm separately as well.

diff --git a/drivers/firmware/efi/libstub/efi-stub.c
b/drivers/firmware/efi/libstub/efi-stub.c
index f515394cce6e..1d4dd8aca3e6 100644
--- a/drivers/firmware/efi/libstub/efi-stub.c
+++ b/drivers/firmware/efi/libstub/efi-stub.c
@@ -136,6 +136,7 @@ efi_status_t __efiapi efi_pe_entry(efi_handle_t handle,
        enum efi_secureboot_mode secure_boot;
        struct screen_info *si;
        efi_properties_table_t *prop_tbl;
+       const efi_dxe_services_table_t *efi_dxe_table;

        efi_system_table = sys_table_arg;

@@ -161,6 +162,20 @@ efi_status_t __efiapi efi_pe_entry(efi_handle_t handle,
                goto fail;
        }

+       efi_dxe_table = get_efi_config_table(EFI_DXE_SERVICES_TABLE_GUID);
+       if (efi_dxe_table) {
+               efi_physical_addr_t pa = 0x60000000;
+               efi_gcd_memory_space_desc_t desc;
+
+               status = efi_dxe_call(get_memory_space_descriptor, pa, &desc);
+               if (status != EFI_SUCCESS)
+                       efi_err("Failed to get memory space
descriptor: %lx\n", status);
+               status = efi_dxe_call(set_memory_space_capabilities, pa, SZ_1G,
+                                     desc.capabilities |
EFI_MEMORY_MORE_RELIABLE);
+               if (status != EFI_SUCCESS)
+                       efi_err("Failed to set memory space
capabilities: %lx\n", status);
+       }
+
        /*
         * Get the command line from EFI, using the LOADED_IMAGE
         * protocol. We are going to copy the command line into the
diff --git a/drivers/firmware/efi/libstub/efistub.h
b/drivers/firmware/efi/libstub/efistub.h
index b0ae0a454404..bf11d85bf9b4 100644
--- a/drivers/firmware/efi/libstub/efistub.h
+++ b/drivers/firmware/efi/libstub/efistub.h
@@ -378,7 +378,8 @@ union efi_dxe_services_table {
                void *schedule;
                void *trust;
                void *process_firmware_volume;
-               void *set_memory_space_capabilities;
+               efi_status_t (__efiapi
*set_memory_space_capabilities)(efi_physical_addr_t,
+
u64, u64);
        };
        struct {
                efi_table_hdr_t hdr;
Mike Rapoport June 11, 2022, 9:56 a.m. UTC | #3
On Fri, Jun 10, 2022 at 01:23:34PM +0200, Ard Biesheuvel wrote:
> On Tue, 7 Jun 2022 at 11:16, Wupeng Ma <mawupeng1@huawei.com> wrote:
> >
> > From: Ma Wupeng <mawupeng1@huawei.com>
> >
> > Commit b05b9f5f9dcf ("x86, mirror: x86 enabling - find mirrored memory ranges")
> > introduced mirrored memory support for x86. This support rely on UEFI to
> > report mirrored memory address ranges.  See UEFI 2.5 spec pages 157-158:
> >
> >   http://www.uefi.org/sites/default/files/resources/UEFI%202_5.pdf
> >
> > Memory mirroring is a technique used to separate memory into two separate
> > channels, usually on a memory device, like a server. In memory mirroring,
> > one channel is copied to another to create redundancy. This method makes
> > input/output (I/O) registers and memory appear with more than one address
> > range because the same physical byte is accessible at more than one
> > address. Using memory mirroring, higher memory reliability and a higher
> > level of memory consolidation are possible.
> >
> > These EFI memory regions have various attributes, and the "mirrored"
> > attribute is one of them. The physical memory region whose descriptors
> > in EFI memory map has EFI_MEMORY_MORE_RELIABLE attribute (bit: 16) are
> > mirrored. The address range mirroring feature of the kernel arranges such
> > mirrored regions into normal zones and other regions into movable zones.
> >
> > Arm64 can support this too. So mirrored memory support is added to support
> > arm64.
> >
> > The main purpose of this patch set is to introduce mirrored support for
> > arm64 and we have already fixed the problems we had which is shown in
> > patch #5 to patch #8 and try to bring total isolation in patch #9 which
> > will disable mirror feature if kernelcore is not specified.
> >
> > In order to test this support in arm64:
> > - patch this patch set
> > - add kernelcore=mirror in kernel parameter
> > - start you kernel
> >
> > Patch #1-#2 introduce mirrored memory support form arm64.
> > Patch #3-#5 fix some bugs for arm64 if memory reliable is enabled.
> > Patch #6 disable mirror feature if kernelcore is not specified.
> >
> > Thanks to Ard Biesheuvel's hard work [1], now kernel will perfer mirrored
> > memory if kaslr is enabled.
> >
> > [1] https://lore.kernel.org/linux-arm-kernel/CAMj1kXEPVEzMgOM4+Yj6PxHA-jFuDOAUdDJSiSxy_XaP4P7LSw@mail.gmail.com/T/
> >
> > Changelog since v2:
> > - remove efi_fake_mem support
> > - remove Commit ("remove some redundant code in ia64 efi_init") since
> >   efi_print_memmap() is not public
> > - add mirror flag back on initrd memory
> >
> > Changelog since v1:
> > - update changelog in cover letter
> > - use PHYS_PFN in patch #7
> >
> > Ma Wupeng (6):
> >   efi: Make efi_find_mirror() public
> >   arm64/mirror: arm64 enabling - find mirrored memory ranges
> >   mm: Ratelimited mirrored memory related warning messages
> >   mm: Demote warning message in vmemmap_verify() to debug level
> >   mm: Add mirror flag back on initrd memory
> >   efi: Disable mirror feature if kernelcore is not specified
> >
> 
> I have tested these changes on QEMU/arm64 with the patch below, and
> things seem to work as expected. We have some minor issues to work out
> but the general shape of this code is good.
> 
> As for the mm/ changes: does anyone mind if I take those through the
> EFI tree as well?

No objections from me.

> I don't think the EFI and -mm changes depend on each other, so they
> can go into -mm separately as well.

--
Sincerely yours,
Mike.