From: Ard Biesheuvel <ard.biesheuvel@linaro.org>
To: Laszlo Ersek <lersek@redhat.com>
Cc: edk2-devel-01 <edk2-devel@lists.01.org>,
Anthony Perard <anthony.perard@citrix.com>,
Jordan Justen <jordan.l.justen@intel.com>,
Julien Grall <julien.grall@linaro.org>
Subject: Re: [PATCH v2 3/8] OvmfPkg: conditionally disable reserved memory varstore emulation at build
Date: Fri, 1 Dec 2017 10:53:38 +0000 [thread overview]
Message-ID: <CAKv+Gu-YXXycOwDKP51HA10h1_Eyt26NE-ROU1MXF0V6azhrTQ@mail.gmail.com> (raw)
In-Reply-To: <ccb09a7b-c576-bcdc-5986-beb0b7370972@redhat.com>
On 1 December 2017 at 10:52, Laszlo Ersek <lersek@redhat.com> wrote:
> On 12/01/17 09:51, Ard Biesheuvel wrote:
>> On 30 November 2017 at 16:30, Laszlo Ersek <lersek@redhat.com> wrote:
>>> (All of the below is only relevant for SMM_REQUIRE=FALSE.)
>>>
>>> For the emulated variable store, PlatformPei allocates reserved memory (as
>>> early as possible, so that the address remains the same during reboot),
>>> and PcdEmuVariableNvStoreReserved carries the address to
>>> EmuVariableFvbRuntimeDxe.
>>>
>>> In addition, QemuFlashFvbServicesRuntimeDxe is always launched before
>>> EmuVariableFvbRuntimeDxe, so that if flash variables are available,
>>> QemuFlashFvbServicesRuntimeDxe can set PcdFlashNvStorageVariableBase64
>>> first, and EmuVariableFvbRuntimeDxe can exit early. This ordering is
>>> currently enforced by adding QemuFlashFvbServicesRuntimeDxe to the APRIORI
>>> DXE file.
>>>
>>> All of this is unnecessary when MEM_VARSTORE_EMU_ENABLE is set to FALSE.
>>> In such a build,
>>>
>>> - (almost) remove the dynamic default for PcdEmuVariableNvStoreReserved
>>> (we can't really do this because the PcdSet64() in
>>> ReserveEmuVariableNvStore() wouldn't compile),
>>>
>>
>> If that is the only concern, and the value is irrelevant, you could
>> make it a patchable PCD instead
>
> This sounds interesting; I've never used patchable PCDs. Can you please
> elaborate?
>
> Do you mean that for (SMM_REQUIRE==TRUE ||
> MEM_VARSTORE_EMU_ENABLE==FALSE), the DSC file should list
> PcdEmuVariableNvStoreReserved somewhere else?
>
Yes.
> Does PcdSet work on patchable PCDs?
>
Yes. It's a bit icky because the value does not actually propagate to
other modules, but that doesn't matter in this case.
next prev parent reply other threads:[~2017-12-01 10:49 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-30 16:30 [PATCH v2 0/8] OvmfPkg: add the Variable PEIM, defragment the UEFI memmap Laszlo Ersek
2017-11-30 16:30 ` [PATCH v2 1/8] OvmfPkg: introduce PcdMemVarstoreEmuEnable feature flag Laszlo Ersek
2017-12-01 8:44 ` Ard Biesheuvel
2017-11-30 16:30 ` [PATCH v2 2/8] OvmfPkg/QemuFlashFvbServicesRuntimeDxe: check PcdMemVarstoreEmuEnable Laszlo Ersek
2017-12-01 8:44 ` Ard Biesheuvel
2017-11-30 16:30 ` [PATCH v2 3/8] OvmfPkg: conditionally disable reserved memory varstore emulation at build Laszlo Ersek
2017-12-01 8:51 ` Ard Biesheuvel
2017-12-01 10:52 ` Laszlo Ersek
2017-12-01 10:53 ` Ard Biesheuvel [this message]
2017-12-01 11:41 ` Laszlo Ersek
2017-12-02 8:53 ` Ard Biesheuvel
2017-11-30 16:30 ` [PATCH v2 4/8] OvmfPkg: introduce FlashNvStorageAddressLib Laszlo Ersek
2017-12-01 8:55 ` Ard Biesheuvel
2017-12-01 11:03 ` Laszlo Ersek
2017-12-01 11:28 ` Laszlo Ersek
2017-11-30 16:30 ` [PATCH v2 5/8] OvmfPkg: include FaultTolerantWritePei and VariablePei Laszlo Ersek
2017-11-30 16:30 ` [PATCH v2 6/8] OvmfPkg/QemuFlashFvbServicesRuntimeDxe: don't set flash PCDs if SMM or no-emu Laszlo Ersek
2017-11-30 16:30 ` [PATCH v2 7/8] OvmfPkg/PlatformPei: remedy UEFI memory map fragmentation Laszlo Ersek
2017-11-30 16:30 ` [PATCH v2 8/8] OvmfPkg/README: document MEM_VARSTORE_EMU_ENABLE and memmap defrag Laszlo Ersek
2017-11-30 19:00 ` [PATCH v2 0/8] OvmfPkg: add the Variable PEIM, defragment the UEFI memmap Jordan Justen
2017-12-01 8:42 ` Ard Biesheuvel
2017-12-01 10:48 ` Laszlo Ersek
2017-12-01 10:44 ` Laszlo Ersek
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-list from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAKv+Gu-YXXycOwDKP51HA10h1_Eyt26NE-ROU1MXF0V6azhrTQ@mail.gmail.com \
--to=devel@edk2.groups.io \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox