From: Laszlo Ersek <lersek@redhat.com>
To: Jian J Wang <jian.j.wang@intel.com>, edk2-devel@lists.01.org
Cc: Jiewen Yao <jiewen.yao@intel.com>,
Eric Dong <eric.dong@intel.com>,
Jeff Fan <vanjeff_919@hotmail.com>
Subject: Re: [PATCH] UefiCpuPkg/MpInitLib: fix wrong base address set as Stack Guard
Date: Wed, 3 Jan 2018 18:33:22 +0100 [thread overview]
Message-ID: <d2cd1cd0-a2e0-edd8-ebd3-52c44d24faab@redhat.com> (raw)
In-Reply-To: <20171229083631.16652-1-jian.j.wang@intel.com>
(CC Jeff)
Sorry about the delay.
I have some light comments below; I expect at least a few of them to be
incorrect :)
On 12/29/17 09:36, Jian J Wang wrote:
> The reason is that DXE part initialization will reuse the stack allocated
> at PEI phase, if MP was initialized before. Some code added to check this
> situation and use stack base address saved in HOB passed from PEI.
>
> Cc: Jiewen Yao <jiewen.yao@intel.com>
> Cc: Eric Dong <eric.dong@intel.com>
> Cc: Laszlo Ersek <lersek@redhat.com>
> Contributed-under: TianoCore Contribution Agreement 1.1
> Signed-off-by: Jian J Wang <jian.j.wang@intel.com>
> ---
> UefiCpuPkg/Library/MpInitLib/DxeMpLib.c | 17 +++++++++++++++--
> 1 file changed, 15 insertions(+), 2 deletions(-)
>
> diff --git a/UefiCpuPkg/Library/MpInitLib/DxeMpLib.c b/UefiCpuPkg/Library/MpInitLib/DxeMpLib.c
> index 40c1bf407a..05484c9ff3 100644
> --- a/UefiCpuPkg/Library/MpInitLib/DxeMpLib.c
> +++ b/UefiCpuPkg/Library/MpInitLib/DxeMpLib.c
> @@ -295,6 +295,7 @@ InitMpGlobalData (
> UINTN Index;
> EFI_GCD_MEMORY_SPACE_DESCRIPTOR MemDesc;
> UINTN StackBase;
> + CPU_INFO_IN_HOB *CpuInfoInHob;
>
> SaveCpuMpData (CpuMpData);
>
> @@ -314,9 +315,18 @@ InitMpGlobalData (
> ASSERT (FALSE);
> }
>
> - for (Index = 0; Index < CpuMpData->CpuCount; ++Index) {
> - StackBase = CpuMpData->Buffer + Index * CpuMpData->CpuApStackSize;
> + //
> + // DXE will reuse stack allocated for APs at PEI phase if it's available.
> + // Let's check it here.
> + //
> + CpuInfoInHob = (CPU_INFO_IN_HOB *)(UINTN)CpuMpData->CpuInfoInHob;
> + if (CpuInfoInHob != NULL && CpuInfoInHob->ApTopOfStack != 0) {
> + StackBase = CpuInfoInHob->ApTopOfStack;
> + } else {
> + StackBase = CpuMpData->Buffer;
> + }
So, if the HOB is not found, then StackBase is set okay.
However, I'm unsure about the other case. The
CPU_INFO_IN_HOB.ApTopOfStack field identifies the *top* of the stack
(highest address, and the stack grows down); however the loop below
*increments* StackBase. Given the incrementing nature of the loop,
shouldn't we first calculate the actual base (= lowest address) from the
CPU_INFO_IN_HOB.ApTopOfStack field?
Actually... I'm even more confused. The CpuMpData->CpuInfoInHob field
points to an *array* of CPU_INFO_IN_HOB structures. Therefore, for any
given processor #N, we should not calculate the stack base as
CpuMpData->CpuInfoInHob->ApTopOfStack + N * CpuMpData->CpuApStackSize
instead we should calculate the stack base as something like:
CpuMpData->CpuInfoInHob[N].ApTopOfStack - CpuMpData->CpuApStackSize
See
- the InitializeApData() function,
- and its call site in the ApWakeupFunction() function.
(To my surprise, I personally modified InitializeApData() earlier, in
commit dd3fa0cd72de ("UefiCpuPkg/MpInitLib: support 64-bit AP stack
addresses", 2016-11-17) -- I've totally forgotten about that by now!)
What do you think?
>
> + for (Index = 0; Index < CpuMpData->CpuCount; ++Index) {
> Status = gDS->GetMemorySpaceDescriptor (StackBase, &MemDesc);
> ASSERT_EFI_ERROR (Status);
>
> @@ -326,6 +336,9 @@ InitMpGlobalData (
> MemDesc.Attributes | EFI_MEMORY_RP
> );
> ASSERT_EFI_ERROR (Status);
> +
> + DEBUG ((DEBUG_VERBOSE, "Stack Guard set at %x [cpu%d]!\n", StackBase, Index));
StackBase has type UINTN, and so it should not be printed with %x. It
should be cast to (UINT64), and then printed with %Lx.
Similarly, Index has type UINTN. It should not be printed with %d. It
should be cast to (UINT64) and printed with %Lu.
> + StackBase += CpuMpData->CpuApStackSize;
Again, I don't think the simple increment applies when the
CpuMpData->CpuInfoInHob array exists.
> }
> }
>
>
Thanks,
Laszlo
next prev parent reply other threads:[~2018-01-03 17:28 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-12-29 8:36 [PATCH] UefiCpuPkg/MpInitLib: fix wrong base address set as Stack Guard Jian J Wang
2018-01-03 7:05 ` Wang, Jian J
2018-01-03 17:33 ` Laszlo Ersek [this message]
2018-01-04 0:41 ` Wang, Jian J
2018-01-04 1:09 ` Wang, Jian J
2018-01-04 1:45 ` Wang, Jian J
2018-01-04 12:21 ` Laszlo Ersek
2018-01-05 0:52 ` Wang, Jian J
2018-01-05 1:40 ` 答复: " Fan Jeff
2018-01-05 1:57 ` Wang, Jian J
2018-01-05 2:48 ` 答复: " Fan Jeff
2018-01-05 2:49 ` Fan Jeff
2018-01-05 2:54 ` Chaganty, Rangasai V
2018-01-05 2:56 ` Wang, Jian J
2018-01-05 2:55 ` Wang, Jian J
2018-01-05 2:57 ` Yao, Jiewen
2018-01-05 3:04 ` 答复: " Fan Jeff
2018-01-05 3:06 ` Wang, Jian J
2018-01-04 12:18 ` 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=d2cd1cd0-a2e0-edd8-ebd3-52c44d24faab@redhat.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