Re: [Qemu-devel] [ARM SMBIOS V1 PATCH 0/6] SMBIOS Support for ARM

[Peter Maydell]

On 5 August 2015 at 18:16, Laszlo Ersek <address@hidden> wrote:
> On 07/28/15 08:00, Wei Huang wrote:
>> SMBIOS tables present userful system hardware info to management
>> applications, such as DMI tools. Even though SMBIOS was originally
>> developed for Intel x86, it has been extended to both Itanium and
>> ARM (32bit & 64bit). More and more ARM server releases, such as
>> RHEL Server for ARM, start to integrate support for SMBIOS.
>>
>> This patchset is intendted to provid SMBIOS tables for ARM mach-virt
>> machine. The SMBIOS tables are created and stored in fw_cfg, relying on
>> OVMF (AAVMF) to parse/present SMBIOS entry.
>>
>> RFC version have been tested by Laszlo using his customized version of
>> AAVMF. We were able to detect SMBIOS 2.8 tables using dmidecode inside
>> an AArch64 guest VM. Moving forward, it is better to support SMBIOS 3.0
>> for ARM guest VM. This new version (V1) integrates SMBIOS 3.0 support
>> for ARM mach-virt. I have tested this version by forcing SMBIOS 2.1
>> format (i.e. passing SMBIOS_21_ENTRY_POINT to smbios_set_defaults()).
>> SMBIOS 3.0 hasn't been tested yet as it requires AAVMF to install 3.0 entry.
>>
>> RFC->V1:
>>  * Add SMBIOS 3.0 support for buidling SMBIOS
>>  * Switch from SMBIOS 2.1 to 3.0 for ARM mach-virt
>>  * RFC version Tested-by Laszlo Ersek and Acked-by Gabriel Somlo
>>
>> Thanks,
>> -Wei
>>
>> Wei Huang (6):
>>   smbios: extract x86 smbios building code into a function
>>   smbios: remove dependency on x86 e820 tables
>>   smbios: pass ram size as a parameter to build smbios tables
>>   smbios: move smbios code into a common folder
>>   smbios: add smbios 3.0 support
>>   smbios: implement smbios support for mach-virt
>>
>>  arch_init.c                          |  2 +-
>>  default-configs/arm-softmmu.mak      |  1 +
>>  default-configs/i386-softmmu.mak     |  1 +
>>  default-configs/x86_64-softmmu.mak   |  1 +
>>  hw/Makefile.objs                     |  1 +
>>  hw/arm/virt.c                        | 24 +++++++++
>>  hw/i386/Makefile.objs                |  2 +-
>>  hw/i386/pc.c                         | 56 ++++++++++++++-------
>>  hw/i386/pc_piix.c                    |  5 +-
>>  hw/i386/pc_q35.c                     |  5 +-
>>  hw/smbios/Makefile.objs              |  1 +
>>  hw/{i386 => smbios}/smbios.c         | 96 
>> +++++++++++++++++++++++-------------
>>  include/hw/arm/virt-acpi-build.h     |  1 +
>>  include/hw/{i386 => smbios}/smbios.h | 42 ++++++++++++++--
>>  tests/bios-tables-test.c             |  2 +-
>>  vl.c                                 |  2 +-
>>  16 files changed, 179 insertions(+), 63 deletions(-)
>>  create mode 100644 hw/smbios/Makefile.objs
>>  rename hw/{i386 => smbios}/smbios.c (93%)
>>  rename include/hw/{i386 => smbios}/smbios.h (84%)
>>
>
> I was hoping there would be a focused review from the subsystem
> maintainers / feature owners for this patchset. Thus far only Shannon
> commented on the series, plus I tested it and reported a small bug (with
> a fix).
>
> Peter: if I review this series (and version 2 that Wei is already
> planning to post, in order to address the notes above, plus anything
> that further review might turn up), will my review suffice for you to
> apply this series (after 2.4 is out)?

Maybe. I haven't looked at the series at all, because it fell
under "not for 2.4 and not something I know enough about to
easily and quickly review" (and besides 5 out of 6 patches are
not ARM-related but just about refactoring the x86 code).

What is SMBIOS supposed to provide for ARM virt anyway?
I would have expected all the information a guest needs
to be in the dtb or ACPI tables...

Is support for this all in the mainline kernel yet?

thanks
-- PMM