* [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages @ 2023-05-18 7:31 duntan 2023-05-18 8:55 ` Ni, Ray 0 siblings, 1 reply; 4+ messages in thread From: duntan @ 2023-05-18 7:31 UTC (permalink / raw) To: devel; +Cc: Ray Ni, Michael D Kinney, Liming Gao, Zhiguang Liu This commit is code optimization to InternalAllocateAlignedPages of SmmMemoryAllocationLib which can reduce free memory fragments. Also it can reduce one pre-allocation page. Let's take a simple example: The expected pages size is 8KB, Alignment value is 8KB. In original InternalAllocateAlignedPages(), the first step is to allocate 4 pages and then find the first 8KB-aligned address in allocated 4 pages. If the upper limit address of allocated 4 pages is already 8KB aligned, then the allocated 4 pages contains two 8KB-aligned 8KB ranges. The lower 2 pages will be selected and removed from free pages. Then the higher 2 pages will be free. Since the whole memory allocation is from high address to low address, then the higher 2 pages cann't be merged with other free pages, causing the free memory fragments. However, when only allocate 3(2+2-1) pages, we can avoid the free memory fragments in specific case. Also 3 pages must contain a 8KB-aligned 8KB range, which meets the requirement. If the upper limit address of allocated 3 pages is 8KB-aligned, then the higher 2 pages range of allocated 3 pages is 8KB-aligned and will be selected and removed from free pages. The remaining lower one page of allocated 3 pages will be free and merged with left lower free memory. This can reduce free memory fragments in smm. Signed-off-by: Dun Tan <dun.tan@intel.com> Cc: Ray Ni <ray.ni@intel.com> Cc: Michael D Kinney <michael.d.kinney@intel.com> Cc: Liming Gao <gaoliming@byosoft.com.cn> Cc: Zhiguang Liu <zhiguang.liu@intel.com> --- MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c index 3ab2c1ebfd..99a8259325 100644 --- a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c +++ b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c @@ -322,7 +322,7 @@ InternalAllocateAlignedPages ( // Calculate the total number of pages since alignment is larger than page size. // AlignmentMask = Alignment - 1; - RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment); + RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment) - 1; // // Make sure that Pages plus EFI_SIZE_TO_PAGES (Alignment) does not overflow. // -- 2.31.1.windows.1 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages 2023-05-18 7:31 [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages duntan @ 2023-05-18 8:55 ` Ni, Ray 2023-05-18 15:37 ` Michael D Kinney 0 siblings, 1 reply; 4+ messages in thread From: Ni, Ray @ 2023-05-18 8:55 UTC (permalink / raw) To: Tan, Dun, devel@edk2.groups.io Cc: Kinney, Michael D, Gao, Liming, Liu, Zhiguang Reviewed-by: Ray Ni <ray.ni@intel.com> > -----Original Message----- > From: Tan, Dun <dun.tan@intel.com> > Sent: Thursday, May 18, 2023 3:31 PM > To: devel@edk2.groups.io > Cc: Ni, Ray <ray.ni@intel.com>; Kinney, Michael D > <michael.d.kinney@intel.com>; Gao, Liming <gaoliming@byosoft.com.cn>; Liu, > Zhiguang <zhiguang.liu@intel.com> > Subject: [Patch V2] MdePkg: Code optimization to SMM > InternalAllocateAlignedPages > > This commit is code optimization to InternalAllocateAlignedPages of > SmmMemoryAllocationLib which can reduce free memory fragments. Also > it can reduce one pre-allocation page. > > Let's take a simple example: > The expected pages size is 8KB, Alignment value is 8KB. > > In original InternalAllocateAlignedPages(), the first step is to > allocate 4 pages and then find the first 8KB-aligned address in > allocated 4 pages. If the upper limit address of allocated 4 pages > is already 8KB aligned, then the allocated 4 pages contains two > 8KB-aligned 8KB ranges. The lower 2 pages will be selected and > removed from free pages. Then the higher 2 pages will be free. > Since the whole memory allocation is from high address to low > address, then the higher 2 pages cann't be merged with other free > pages, causing the free memory fragments. > > However, when only allocate 3(2+2-1) pages, we can avoid the free > memory fragments in specific case. Also 3 pages must contain a > 8KB-aligned 8KB range, which meets the requirement. If the upper > limit address of allocated 3 pages is 8KB-aligned, then the higher > 2 pages range of allocated 3 pages is 8KB-aligned and will be > selected and removed from free pages. The remaining lower one page > of allocated 3 pages will be free and merged with left lower free > memory. This can reduce free memory fragments in smm. > > Signed-off-by: Dun Tan <dun.tan@intel.com> > Cc: Ray Ni <ray.ni@intel.com> > Cc: Michael D Kinney <michael.d.kinney@intel.com> > Cc: Liming Gao <gaoliming@byosoft.com.cn> > Cc: Zhiguang Liu <zhiguang.liu@intel.com> > --- > MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > index 3ab2c1ebfd..99a8259325 100644 > --- a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > +++ b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > @@ -322,7 +322,7 @@ InternalAllocateAlignedPages ( > // Calculate the total number of pages since alignment is larger than page size. > // > AlignmentMask = Alignment - 1; > - RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment); > + RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment) - 1; > // > // Make sure that Pages plus EFI_SIZE_TO_PAGES (Alignment) does not > overflow. > // > -- > 2.31.1.windows.1 ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages 2023-05-18 8:55 ` Ni, Ray @ 2023-05-18 15:37 ` Michael D Kinney 2023-05-18 23:30 ` Ni, Ray 0 siblings, 1 reply; 4+ messages in thread From: Michael D Kinney @ 2023-05-18 15:37 UTC (permalink / raw) To: Ni, Ray, Tan, Dun, devel@edk2.groups.io Cc: Gao, Liming, Liu, Zhiguang, Kinney, Michael D Is this considered a critical bug for stable tag release? Is there are HSD and if it is critical, all memory allocation lobs should be fixed. Right? Mike > -----Original Message----- > From: Ni, Ray <ray.ni@intel.com> > Sent: Thursday, May 18, 2023 1:55 AM > To: Tan, Dun <dun.tan@intel.com>; devel@edk2.groups.io > Cc: Kinney, Michael D <michael.d.kinney@intel.com>; Gao, Liming > <gaoliming@byosoft.com.cn>; Liu, Zhiguang <zhiguang.liu@intel.com> > Subject: RE: [Patch V2] MdePkg: Code optimization to SMM > InternalAllocateAlignedPages > > Reviewed-by: Ray Ni <ray.ni@intel.com> > > > -----Original Message----- > > From: Tan, Dun <dun.tan@intel.com> > > Sent: Thursday, May 18, 2023 3:31 PM > > To: devel@edk2.groups.io > > Cc: Ni, Ray <ray.ni@intel.com>; Kinney, Michael D > > <michael.d.kinney@intel.com>; Gao, Liming <gaoliming@byosoft.com.cn>; > Liu, > > Zhiguang <zhiguang.liu@intel.com> > > Subject: [Patch V2] MdePkg: Code optimization to SMM > > InternalAllocateAlignedPages > > > > This commit is code optimization to InternalAllocateAlignedPages of > > SmmMemoryAllocationLib which can reduce free memory fragments. Also > > it can reduce one pre-allocation page. > > > > Let's take a simple example: > > The expected pages size is 8KB, Alignment value is 8KB. > > > > In original InternalAllocateAlignedPages(), the first step is to > > allocate 4 pages and then find the first 8KB-aligned address in > > allocated 4 pages. If the upper limit address of allocated 4 pages > > is already 8KB aligned, then the allocated 4 pages contains two > > 8KB-aligned 8KB ranges. The lower 2 pages will be selected and > > removed from free pages. Then the higher 2 pages will be free. > > Since the whole memory allocation is from high address to low > > address, then the higher 2 pages cann't be merged with other free > > pages, causing the free memory fragments. > > > > However, when only allocate 3(2+2-1) pages, we can avoid the free > > memory fragments in specific case. Also 3 pages must contain a > > 8KB-aligned 8KB range, which meets the requirement. If the upper > > limit address of allocated 3 pages is 8KB-aligned, then the higher > > 2 pages range of allocated 3 pages is 8KB-aligned and will be > > selected and removed from free pages. The remaining lower one page > > of allocated 3 pages will be free and merged with left lower free > > memory. This can reduce free memory fragments in smm. > > > > Signed-off-by: Dun Tan <dun.tan@intel.com> > > Cc: Ray Ni <ray.ni@intel.com> > > Cc: Michael D Kinney <michael.d.kinney@intel.com> > > Cc: Liming Gao <gaoliming@byosoft.com.cn> > > Cc: Zhiguang Liu <zhiguang.liu@intel.com> > > --- > > MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git > a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > index 3ab2c1ebfd..99a8259325 100644 > > --- a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > +++ b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > @@ -322,7 +322,7 @@ InternalAllocateAlignedPages ( > > // Calculate the total number of pages since alignment is larger than > page size. > > // > > AlignmentMask = Alignment - 1; > > - RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment); > > + RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment) - 1; > > // > > // Make sure that Pages plus EFI_SIZE_TO_PAGES (Alignment) does not > > overflow. > > // > > -- > > 2.31.1.windows.1 ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages 2023-05-18 15:37 ` Michael D Kinney @ 2023-05-18 23:30 ` Ni, Ray 0 siblings, 0 replies; 4+ messages in thread From: Ni, Ray @ 2023-05-18 23:30 UTC (permalink / raw) To: Kinney, Michael D, Tan, Dun, devel@edk2.groups.io Cc: Gao, Liming, Liu, Zhiguang Mike, No. The code is there for couple years. With that fact, I don't think it's a critical bug. Yes. all memory allocation libs should be fixed. We start with the SMM one first to see feedbacks. Thanks, Ray > -----Original Message----- > From: Kinney, Michael D <michael.d.kinney@intel.com> > Sent: Thursday, May 18, 2023 11:38 PM > To: Ni, Ray <ray.ni@intel.com>; Tan, Dun <dun.tan@intel.com>; > devel@edk2.groups.io > Cc: Gao, Liming <gaoliming@byosoft.com.cn>; Liu, Zhiguang > <zhiguang.liu@intel.com>; Kinney, Michael D <michael.d.kinney@intel.com> > Subject: RE: [Patch V2] MdePkg: Code optimization to SMM > InternalAllocateAlignedPages > > Is this considered a critical bug for stable tag release? > > Is there are HSD and if it is critical, all memory allocation lobs should be fixed. > Right? > > Mike > > > -----Original Message----- > > From: Ni, Ray <ray.ni@intel.com> > > Sent: Thursday, May 18, 2023 1:55 AM > > To: Tan, Dun <dun.tan@intel.com>; devel@edk2.groups.io > > Cc: Kinney, Michael D <michael.d.kinney@intel.com>; Gao, Liming > > <gaoliming@byosoft.com.cn>; Liu, Zhiguang <zhiguang.liu@intel.com> > > Subject: RE: [Patch V2] MdePkg: Code optimization to SMM > > InternalAllocateAlignedPages > > > > Reviewed-by: Ray Ni <ray.ni@intel.com> > > > > > -----Original Message----- > > > From: Tan, Dun <dun.tan@intel.com> > > > Sent: Thursday, May 18, 2023 3:31 PM > > > To: devel@edk2.groups.io > > > Cc: Ni, Ray <ray.ni@intel.com>; Kinney, Michael D > > > <michael.d.kinney@intel.com>; Gao, Liming <gaoliming@byosoft.com.cn>; > > Liu, > > > Zhiguang <zhiguang.liu@intel.com> > > > Subject: [Patch V2] MdePkg: Code optimization to SMM > > > InternalAllocateAlignedPages > > > > > > This commit is code optimization to InternalAllocateAlignedPages of > > > SmmMemoryAllocationLib which can reduce free memory fragments. Also > > > it can reduce one pre-allocation page. > > > > > > Let's take a simple example: > > > The expected pages size is 8KB, Alignment value is 8KB. > > > > > > In original InternalAllocateAlignedPages(), the first step is to > > > allocate 4 pages and then find the first 8KB-aligned address in > > > allocated 4 pages. If the upper limit address of allocated 4 pages > > > is already 8KB aligned, then the allocated 4 pages contains two > > > 8KB-aligned 8KB ranges. The lower 2 pages will be selected and > > > removed from free pages. Then the higher 2 pages will be free. > > > Since the whole memory allocation is from high address to low > > > address, then the higher 2 pages cann't be merged with other free > > > pages, causing the free memory fragments. > > > > > > However, when only allocate 3(2+2-1) pages, we can avoid the free > > > memory fragments in specific case. Also 3 pages must contain a > > > 8KB-aligned 8KB range, which meets the requirement. If the upper > > > limit address of allocated 3 pages is 8KB-aligned, then the higher > > > 2 pages range of allocated 3 pages is 8KB-aligned and will be > > > selected and removed from free pages. The remaining lower one page > > > of allocated 3 pages will be free and merged with left lower free > > > memory. This can reduce free memory fragments in smm. > > > > > > Signed-off-by: Dun Tan <dun.tan@intel.com> > > > Cc: Ray Ni <ray.ni@intel.com> > > > Cc: Michael D Kinney <michael.d.kinney@intel.com> > > > Cc: Liming Gao <gaoliming@byosoft.com.cn> > > > Cc: Zhiguang Liu <zhiguang.liu@intel.com> > > > --- > > > MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git > > a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > > b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > > index 3ab2c1ebfd..99a8259325 100644 > > > --- a/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > > +++ b/MdePkg/Library/SmmMemoryAllocationLib/MemoryAllocationLib.c > > > @@ -322,7 +322,7 @@ InternalAllocateAlignedPages ( > > > // Calculate the total number of pages since alignment is larger than > > page size. > > > // > > > AlignmentMask = Alignment - 1; > > > - RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment); > > > + RealPages = Pages + EFI_SIZE_TO_PAGES (Alignment) - 1; > > > // > > > // Make sure that Pages plus EFI_SIZE_TO_PAGES (Alignment) does not > > > overflow. > > > // > > > -- > > > 2.31.1.windows.1 ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2023-05-18 23:30 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2023-05-18 7:31 [Patch V2] MdePkg: Code optimization to SMM InternalAllocateAlignedPages duntan 2023-05-18 8:55 ` Ni, Ray 2023-05-18 15:37 ` Michael D Kinney 2023-05-18 23:30 ` Ni, Ray
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox