From: "Michael Brown" <mcb30@ipxe.org>
To: "Zhou, Jianfeng" <jianfeng.zhou@intel.com>,
"devel@edk2.groups.io" <devel@edk2.groups.io>
Cc: "Ni, Ray" <ray.ni@intel.com>, Laszlo Ersek <lersek@redhat.com>,
"Kumar, Rahul R" <rahul.r.kumar@intel.com>,
Gerd Hoffmann <kraxel@redhat.com>,
Pedro Falcato <pedro.falcato@gmail.com>,
"Zhang, Di" <di.zhang@intel.com>, "Tan, Dun" <dun.tan@intel.com>
Subject: Re: [edk2-devel] [PATCH] UefiCpuPkg/CpuPageTableLib: qualify page table accesses as volatile
Date: Fri, 23 Feb 2024 15:51:10 +0000 [thread overview]
Message-ID: <0102018dd6a9d291-d987d495-f53a-4e39-8782-336e3c2e13da-000000@eu-west-1.amazonses.com> (raw)
In-Reply-To: <PH7PR11MB6673275D6F03F32A89A43103EF552@PH7PR11MB6673.namprd11.prod.outlook.com>
On 23/02/2024 15:12, Zhou, Jianfeng wrote:
>> While it may well cause the compiler to generate less optimised code, there is absolutely no way that this volatile declaration on a local stack variable can possibly change the outcome of the code.
>> There can never be any meaningful side-effects from reading or writing a stack variable.
>> I would suggest dropping the volatile on LocalPte4K, since its *only* possible impact is to confuse a future reader of the code.
>
> The change is for preventing compiler from optimizing.
> As a temporary variable, LocalPte4K may be replaced by function parameter Pte4K.
No, it can't. If Pte4K is marked as a volatile pointer, then the
compiler is not allowed to unilaterally decide to treat it as a
non-volatile pointer.
> In this case, code like "LocalPte4K.Bits.Present = Attribute->Bits.Present" may lead to unexpected result, as it is not atomic. Assembly code look like:
> mov eax, [r8]
> and dword [rcx], 0xfffffffe // this instruction clear the present bit and may leads to unexpected result.
> and eax, 0x1
> or [rcx], eax
Please test with Pte4K marked as volatile and LocalPte4K marked as
non-volatile. If you can still generate assembly code that writes to
*Pte4K more than once, then that would be a serious compiler bug.
As a separate note, I would also suggest removing the unnecessary second
read through Pte4K, since once Pte4K is marked as volatile the compiler
will generate an extra read from that address:
--- a/UefiCpuPkg/Library/CpuPageTableLib/CpuPageTableMap.c
+++ b/UefiCpuPkg/Library/CpuPageTableLib/CpuPageTableMap.c
@@ -30,7 +30,7 @@ PageTableLibSetPte4K (
LocalPte4K.Uint64 = Pte4K->Uint64;
if (Mask->Bits.PageTableBaseAddressLow ||
Mask->Bits.PageTableBaseAddressHigh) {
- LocalPte4K.Uint64 = (IA32_MAP_ATTRIBUTE_PAGE_TABLE_BASE_ADDRESS
(Attribute) + Offset) | (Pte4K->Uint64 & ~IA32_PE_BASE_ADDRESS_MASK_40);
+ LocalPte4K.Uint64 = (IA32_MAP_ATTRIBUTE_PAGE_TABLE_BASE_ADDRESS
(Attribute) + Offset) | (LocalPte4K.Uint64 & ~IA32_PE_BASE_ADDRESS_MASK_40);
}
if (Mask->Bits.Present) {
Michael
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#115888): https://edk2.groups.io/g/devel/message/115888
Mute This Topic: https://groups.io/mt/104524857/7686176
Group Owner: devel+owner@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [rebecca@openfw.io]
-=-=-=-=-=-=-=-=-=-=-=-
next prev parent reply other threads:[~2024-02-23 15:51 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-22 8:41 [edk2-devel] [PATCH] UefiCpuPkg/CpuPageTableLib: qualify page table accesses as volatile Zhou Jianfeng
2024-02-23 11:59 ` Michael Brown
2024-02-23 15:12 ` Zhou, Jianfeng
2024-02-23 15:51 ` Michael Brown [this message]
2024-02-25 13:47 ` Laszlo Ersek
-- strict thread matches above, loose matches on Subject: below --
2024-03-01 2:54 Zhou Jianfeng
2024-03-01 11:50 ` Michael Brown
2024-03-01 12:21 ` Laszlo Ersek
2024-03-01 18:56 ` 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=0102018dd6a9d291-d987d495-f53a-4e39-8782-336e3c2e13da-000000@eu-west-1.amazonses.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