Re: [PATCH 00/43] OvmfPkg: remove Xen support from OvmfPkg*.dsc, in favor of OvmfXen.dsc

["Laszlo Ersek" <lersek@redhat.com>]

On 05/27/21 09:34, Ard Biesheuvel wrote:
> On Wed, 26 May 2021 at 22:15, Laszlo Ersek <lersek@redhat.com> wrote:
>>
>> Bugzilla: https://bugzilla.tianocore.org/show_bug.cgi?id=2122
>> Repo:     https://pagure.io/lersek/edk2.git
>> Branch:   xen_split_bz_2122
>>
>> This patch set removes dynamic Xen enlightenment from the following
>> platforms:
>>
>>   OvmfPkg/OvmfPkgIa32.dsc
>>   OvmfPkg/OvmfPkgIa32X64.dsc
>>   OvmfPkg/OvmfPkgX64.dsc
>>
>> In Xen guests, the following platform should be used:
>>
>>   OvmfPkg/OvmfXen.dsc
>>
>> Please see more details / references in the bugzilla ticket.
>>
>> NOOPT build savings:
>>
>> - Ia32:    PEIFV 1536 bytes, DXEFV 130288 bytes
>> - Ia32X64: PEIFV 1536 bytes, DXEFV 140912 bytes
>> - X64:     PEIFV 1664 bytes, DXEFV 140912 bytes
>> - Xen:     PEIFV  256 bytes, DXEFV  69504 bytes
>>
>> Functional testing:
>>
>> - Booted a Fedora guest on OvmfPkgIa32X64 on QEMU/KVM, compared verbose
>>   logs before-after. Memory allocations were satisfied at different
>>   addresses, as expected, plus the Xen drivers were absent. No
>>   differences otherwise.
>>
>> - Booted a RHEL guest on ArmVirtQemu on AARCH64. Memory allocations were
>>   satisfied at different addresses, as expected.
>>
>> - Xen regression-testing was not done; I'm requesting feedback.
>>
>> Build testing / bisectability: at every stage, the series builds with
>> the following script:
>>
>>> #!/bin/bash
>>> set -e -u -C
>>>
>>> build -b DEBUG -t GCC5 -p ArmVirtPkg/ArmVirtKvmTool.dsc            -a AARCH64
>>> build -b DEBUG -t GCC5 -p ArmVirtPkg/ArmVirtKvmTool.dsc    -a ARM
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtQemu.dsc               -a AARCH64
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtQemu.dsc       -a ARM
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtQemuKernel.dsc         -a AARCH64
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtQemuKernel.dsc -a ARM
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtXen.dsc                -a AARCH64
>>> build -b NOOPT -t GCC5 -p ArmVirtPkg/ArmVirtXen.dsc        -a ARM
>>> build -b NOOPT -t GCC5 -p OvmfPkg/AmdSev/AmdSevX64.dsc             -a X64
>>> build -b NOOPT -t GCC5 -p OvmfPkg/Bhyve/BhyveX64.dsc               -a X64
>>> build -b NOOPT -t GCC5 -p OvmfPkg/OvmfPkgIa32.dsc          -a IA32
>>> build -b NOOPT -t GCC5 -p OvmfPkg/OvmfPkgIa32X64.dsc       -a IA32 -a X64
>>> build -b NOOPT -t GCC5 -p OvmfPkg/OvmfPkgX64.dsc                   -a X64
>>> build -b NOOPT -t GCC5 -p OvmfPkg/OvmfXen.dsc                      -a X64
>>
>> The patches in the series were formatted with the following options, for
>> posting:
>>
>>   --stat=1000 --stat-graph-width=20 --find-copies-harder -U6
>>
>> (The option "--find-copies-harder" is not the best for presenting every
>> single patch in the series, in isolation, but taken globally for the
>> entire series, it is the most helpful option.)
>>
>> Some patches advance with really small steps, in order to cut down on a
>> subsequent "meaty" patch. Personally I don't like reviewing code
>> movement patches, so I did my best to (a) keep that to a minimum, and
>> (b) present it as unintrusively as possible.
>>
>> The CC list is a bit long; the reason is that I kept touching up
>> "Maintainers.txt", and the "OvmfPkg/Bhyve" and "OvmfPkg/AmdSev"
>> platforms as well (whenever it made sense).

> Thanks for taking this on.
> 
> For the series,
> 
> Reviewed-by: Ard Biesheuvel <ardb@kernel.org>

Thank you, Ard!

I hope the large patch count (with the small patch bodies and the
not-so-short commit messages) didn't get on your nerves! I really did
make an effort to find a good balance. E.g. in the "clean up and sort
#includes & INF files" patches.

Thank you again!
Laszlo