From mboxrd@z Thu Jan 1 00:00:00 1970 Subject: Re: [edk2-devel] [PATCH 2/5] UefiCpuPkg: Contiguous memory allocation and code clean-up. To: Yuanhao Xie ,devel@edk2.groups.io From: =?UTF-8?B?TWFydmluIEjDpHVzZXI=?= X-Originating-Location: Berlin, Land Berlin, DE (104.28.45.25) X-Originating-Platform: Mac Safari 16.3 User-Agent: GROUPS.IO Web Poster MIME-Version: 1.0 Date: Tue, 07 Feb 2023 08:40:41 -0800 References: <20230207134939.273-3-yuanhao.xie@intel.com> In-Reply-To: <20230207134939.273-3-yuanhao.xie@intel.com> Message-ID: <32073.1675788041796635561@groups.io> Content-Type: multipart/alternative; boundary="J7dQcgKZKLrOvgcBMfl5" --J7dQcgKZKLrOvgcBMfl5 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Hi=C2=A0Yuanhao, 1) The code comments and copy code suggest that the stacks are located low = and the function is located high (good). However, the SetMemorySpaceAttribu= tes() call un-XP's Address, which is the low address. So, do I misunderstan= d the changes, or are you un-XP'ing the first stack (and keep the function = XP'd)? 2) The same SetMemorySpaceAttributes() call, you now pass ApLoopFuncSize ov= er ApSafeBufferSize. The latter was explicitly page-aligned, while the form= er is not. How is it guaranteed it is indeed aligned? If it is not, I don't= think this is supported, at least universally. 3) Similar to 2), the stack size is much smaller than the page size, no? Ho= w do you guarantee the function is on a page boundary for memory protection= ? 4) A proper W^X flow should be to wait with un-XP till the CopyMem() for th= e function code has returned. Right before that, the copied code should be = marked read-only. Best regards, Marvin --J7dQcgKZKLrOvgcBMfl5 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable Hi Yuanhao,

1) The code comments and copy code suggest that= the stacks are located low and the function is located high (good). Howeve= r, the SetMemorySpaceAttributes= () call un-XP's Address, which is the low address. So, do I misunderstand t= he changes, or are you un-XP'ing the first stack (and keep the function XP'= d)?

2) The same Se= tMemorySpaceAttributes() call, you now pass ApLoopFuncSize over ApSafeBufferSize. The l= atter was explicitly page-aligned, while the former is not. How is it guara= nteed it is indeed aligned? If it is not, I don't think this is supported, = at least universally.

3) Similar to 2), the stack size is much s= maller than the page size, no? How do you guarantee the function is on a pa= ge boundary for memory protection?

4) A proper W^X flow should b= e to wait with un-XP till the CopyMem() for the function code has returned. Right before that, the copie= d code should be marked read-only.

Best regards,
Marvin --J7dQcgKZKLrOvgcBMfl5--