<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Aug 13, 2015 at 8:48 AM, Rui Ueyama <span dir="ltr"><<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><p dir="ltr">Noticed that at least on Linux, the file offset must be a multiple of the page size, or the kernel returns an error to the system call. I still don't think this has a negative impact on performance because this file structure is the same as MSVC linker (that aligns sections to 512 bytes boundaries on disk) though.</p></blockquote><div>They may have a hack in their kernel that makes this not a problem. Still, I would highly recommend measuring this to be sure.</div><div><br></div><div>For example, the kernel is *guaranteed* to have to make a copy after LLD outputs an unaligned file, since LLD has the section mapped at a different offset modulo the page size than the process will when it is executing (not that this matters very much since LLD only outputs the file on the build machine).</div><div><br></div><div>-- Sean Silva</div><div><br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="gmail_quote">2015/08/13 16:51 "Rui Ueyama" <<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>>:<div><div class="h5"><br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">I think I do understand how the paging mechanism works. :) We are talking about different things. My question is why you think a file offset must be at a 4K boundary in order to map it efficiently to memory. To me you seems to be claiming that mmap(0, /*length*/4096, PROT_READ|PROT_EXEC, 0, SomeFD, /*file offset*/5120) is much inefficient than mmap(0, /*length*/4096, PROT_READ|PROT_EXEC, 0, SomeFD, /*file offset*/4096) because of the file offset of the former mmap call is not a multiple of 4096. And I'm saying that that's not true.<div class="gmail_extra"><br><div class="gmail_quote">On Thu, Aug 13, 2015 at 4:10 PM, Sean Silva <span dir="ltr"><<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span>On Wed, Aug 12, 2015 at 10:59 PM, Rui Ueyama <span dir="ltr"><<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div>What I don't understand is that why the offset from the beginning of the file must be multiple of page size in order to avoid full copy. Windows requires all sections to be aligned at least 4K in memory and 512 bytes on file, and I don't see any problem there.</div><div><br></div><div>Let's say we have two sections, A and B, whose sizes are 1024B and 4096B, respectively. We also assume that A's offset from the beginning of file is 4096, and B's 5120. The loader can map offset 4096 to 8192 of the file to some page, and 5120 to 9216 to other page. Why can't that?</div></div></blockquote><div><br></div></span><div>From the kernel's perspective of mapping memory (on x86), memory is divided into aligned 4K pieces. 5120 % 4096 == 1024, so in order to map it at an address that is 4K aligned, it must do a full memmove in order to move all the memory by 1024 bytes so that it is 4K aligned. This image maybe helps to understand how a 32-bit x86 CPU understands a virtual memory address:</div><div><a href="https://upload.wikimedia.org/wikipedia/commons/8/8e/X86_Paging_4K.svg" target="_blank">https://upload.wikimedia.org/wikipedia/commons/8/8e/X86_Paging_4K.svg</a></div><div><br></div><div>IIRC the resources I learned from are:</div><div><a href="http://duartes.org/gustavo/blog/post/how-the-kernel-manages-your-memory/" target="_blank">http://duartes.org/gustavo/blog/post/how-the-kernel-manages-your-memory/</a><br></div><div><a href="http://duartes.org/gustavo/blog/post/the-thing-king/" target="_blank">http://duartes.org/gustavo/blog/post/the-thing-king/</a><br></div><div>(that web page has many other very, *very* good posts. A list can be seen at: <a href="http://duartes.org/gustavo/blog/category/internals/" target="_blank">http://duartes.org/gustavo/blog/category/internals/</a>)</div><div><br></div><div>I think you will find that understanding virtual memory (and TLB) will greatly help you optimize LLD, since many operations in LLD have very high pressure on the virtual memory system.</div><span><font color="#888888"><div><br></div><div>-- Sean Silva</div></font></span><div><div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div><div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Aug 13, 2015 at 2:44 PM, Sean Silva <span dir="ltr"><<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span>On Wed, Aug 12, 2015 at 7:17 PM, Rui Ueyama <span dir="ltr"><<a href="mailto:ruiu@google.com" target="_blank">ruiu@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">I didn't test to see if this change ever has a negative impact on memory usage, but my guess is that's very unlikely because this layout is the same as what MSVC linker creates. If this is inefficient, virtually all Windows executables are being suffered by that, which is unlikely. My understanding is that the kernel maps each section separately to a memory address, so file offset of each section can be given independently from other sections.</div></blockquote><div><br></div></span><div>The mapping is done at the granularity of aligned 4K pages minimum (this is just how the x86 hardware page table mechanism works). A piece of the file cannot be moved by an amount that is not a multiple of 4K without a full copy.</div><div><br></div><div>The only way this could (in the usual case) not have a large overhead is for the kernel to do a crazy hack like have special paging semantics for files that are executables. This means that when LLD finishes working on a memory mapped file, if a section is not 4K aligned at least, then the kernel has to then do a copy to make the file conform the the actual memory layout it needs to have in the paging subsystem.</div><div><br></div><div>Or does windows make full copies of sections always? In other words processes don't share e.g. readonly text?</div><div><br></div><div>In ELF, the offset in the file and the offset in memory are required to be congruent modulo the alignment (see the documentation of p_align in <a href="http://www.sco.com/developers/gabi/latest/ch5.pheader.html" target="_blank">http://www.sco.com/developers/gabi/latest/ch5.pheader.html</a>), precisely to avoid the need to do crazy hacks like this when loading the program.</div><div><br></div><div>You can see that Linux will reject the binary:</div><div><a href="http://lxr.free-electrons.com/source/fs/binfmt_elf.c#L664" target="_blank">http://lxr.free-electrons.com/source/fs/binfmt_elf.c#L664</a> (load_elf_binary)<br></div><div>-> <a href="http://lxr.free-electrons.com/source/fs/binfmt_elf.c#L336" target="_blank">http://lxr.free-electrons.com/source/fs/binfmt_elf.c#L336</a> (elf_map)<br></div><div>-> <a href="http://lxr.free-electrons.com/source/mm/util.c#L306" target="_blank">http://lxr.free-electrons.com/source/mm/util.c#L306</a> (vm_mmap)</div><div>Notice:</div><div><div>312 if (unlikely(offset & ~PAGE_MASK))</div><div>313 return -EINVAL;</div></div><div><br></div><div>FreeBSD is more lenient, but you can see that the kernel does not like the situation when this is violated:</div><div><a href="http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#593" target="_blank">http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#593</a> (__elfN(load_file))<br></div><div>--> <a href="http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#467" target="_blank">http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#467</a> (__elfN(load_section))<br></div><div>--> <a href="http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#398" target="_blank">http://src.illumos.org/source/xref/freebsd-head/sys/kern/imgact_elf.c#398</a> (__elfN(map_insert))</div><div><div>423<span style="white-space:pre-wrap"> </span>/*</div><div>424<span style="white-space:pre-wrap"> </span> * The mapping is not page aligned. This means we have</div><div>425<span style="white-space:pre-wrap"> </span> * to copy the data. Sigh.</div><div>426<span style="white-space:pre-wrap"> </span> */</div></div><span><font color="#888888"><div><br></div><div>-- Sean Silva</div></font></span><div><div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div><div><div><div><br></div><div><div class="gmail_extra"><div class="gmail_quote">On Wed, Aug 12, 2015 at 5:31 PM, Sean Silva <span dir="ltr"><<a href="mailto:chisophugis@gmail.com" target="_blank">chisophugis@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span>On Tue, Aug 11, 2015 at 4:09 PM, Rui Ueyama via llvm-commits <span dir="ltr"><<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Author: ruiu<br>
Date: Tue Aug 11 18:09:00 2015<br>
New Revision: 244691<br>
<br>
URL: <a href="http://llvm.org/viewvc/llvm-project?rev=244691&view=rev" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project?rev=244691&view=rev</a><br>
Log:<br>
COFF: Align sections to 512-byte boundaries on disk.<br>
<br>
Sections must start at page boundaries in memory, but they<br>
can be aligned to sector boundaries (512-bytes) on disk.<br>
We aligned them to 4096-byte boundaries even on disk, so we<br>
wasted disk space a bit.<br></blockquote><div><br></div></span><div>This will likely force the kernel to copy or otherwise do unnecessary work when loading. Are you sure that isn't happening? The kernel ideally wants to just create a couple page table entries. But if it needs to move things around at <4K granularity to make them properly aligned to their load address when loading (like this patch I think causes) then it will need to do copies.</div><div><br></div><div>This can likely be checked by looking for an increase in real memory usage for the system when the new binaries are loaded (vs. the old page-aligned ones), since the kernel will have a copy sitting in page cache and a copy for alignment mapped into the process address space; alternatively, you can check for the slowdown from the kernel copies when faulting the memory into the process's address space (or (less likely) it may do the copies eagerly which should be easy to measure too).</div><span><font color="#888888"><div><br></div><div>-- Sean Silva</div></font></span><div><div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Modified:<br>
lld/trunk/COFF/Writer.cpp<br>
lld/trunk/test/COFF/baserel.test<br>
lld/trunk/test/COFF/hello32.test<br>
<br>
Modified: lld/trunk/COFF/Writer.cpp<br>
URL: <a href="http://llvm.org/viewvc/llvm-project/lld/trunk/COFF/Writer.cpp?rev=244691&r1=244690&r2=244691&view=diff" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project/lld/trunk/COFF/Writer.cpp?rev=244691&r1=244690&r2=244691&view=diff</a><br>
==============================================================================<br>
--- lld/trunk/COFF/Writer.cpp (original)<br>
+++ lld/trunk/COFF/Writer.cpp Tue Aug 11 18:09:00 2015<br>
@@ -37,8 +37,7 @@ using namespace lld;<br>
using namespace lld::coff;<br>
<br>
static const int PageSize = 4096;<br>
-static const int FileAlignment = 512;<br>
-static const int SectionAlignment = 4096;<br>
+static const int SectorSize = 512;<br>
static const int DOSStubSize = 64;<br>
static const int NumberfOfDataDirectory = 16;<br>
<br>
@@ -174,7 +173,7 @@ void OutputSection::addChunk(Chunk *C) {<br>
Off += C->getSize();<br>
Header.VirtualSize = Off;<br>
if (C->hasData())<br>
- Header.SizeOfRawData = RoundUpToAlignment(Off, FileAlignment);<br>
+ Header.SizeOfRawData = RoundUpToAlignment(Off, SectorSize);<br>
}<br>
<br>
void OutputSection::addPermissions(uint32_t C) {<br>
@@ -507,15 +506,14 @@ void Writer::createSymbolAndStringTable(<br>
// We position the symbol table to be adjacent to the end of the last section.<br>
uint64_t FileOff =<br>
LastSection->getFileOff() +<br>
- RoundUpToAlignment(LastSection->getRawSize(), FileAlignment);<br>
+ RoundUpToAlignment(LastSection->getRawSize(), SectorSize);<br>
if (!OutputSymtab.empty()) {<br>
PointerToSymbolTable = FileOff;<br>
FileOff += OutputSymtab.size() * sizeof(coff_symbol16);<br>
}<br>
if (!Strtab.empty())<br>
FileOff += Strtab.size() + 4;<br>
- FileSize = SizeOfHeaders +<br>
- RoundUpToAlignment(FileOff - SizeOfHeaders, FileAlignment);<br>
+ FileSize = RoundUpToAlignment(FileOff, SectorSize);<br>
}<br>
<br>
// Visits all sections to assign incremental, non-overlapping RVAs and<br>
@@ -526,9 +524,9 @@ void Writer::assignAddresses() {<br>
sizeof(coff_section) * OutputSections.size();<br>
SizeOfHeaders +=<br>
Config->is64() ? sizeof(pe32plus_header) : sizeof(pe32_header);<br>
- SizeOfHeaders = RoundUpToAlignment(SizeOfHeaders, PageSize);<br>
+ SizeOfHeaders = RoundUpToAlignment(SizeOfHeaders, SectorSize);<br>
uint64_t RVA = 0x1000; // The first page is kept unmapped.<br>
- uint64_t FileOff = SizeOfHeaders;<br>
+ FileSize = SizeOfHeaders;<br>
// Move DISCARDABLE (or non-memory-mapped) sections to the end of file because<br>
// the loader cannot handle holes.<br>
std::stable_partition(<br>
@@ -539,13 +537,11 @@ void Writer::assignAddresses() {<br>
if (Sec->getName() == ".reloc")<br>
addBaserels(Sec);<br>
Sec->setRVA(RVA);<br>
- Sec->setFileOffset(FileOff);<br>
+ Sec->setFileOffset(FileSize);<br>
RVA += RoundUpToAlignment(Sec->getVirtualSize(), PageSize);<br>
- FileOff += RoundUpToAlignment(Sec->getRawSize(), FileAlignment);<br>
+ FileSize += RoundUpToAlignment(Sec->getRawSize(), SectorSize);<br>
}<br>
SizeOfImage = SizeOfHeaders + RoundUpToAlignment(RVA - 0x1000, PageSize);<br>
- FileSize = SizeOfHeaders +<br>
- RoundUpToAlignment(FileOff - SizeOfHeaders, FileAlignment);<br>
}<br>
<br>
template <typename PEHeaderTy> void Writer::writeHeader() {<br>
@@ -584,8 +580,8 @@ template <typename PEHeaderTy> void Writ<br>
Buf += sizeof(*PE);<br>
PE->Magic = Config->is64() ? PE32Header::PE32_PLUS : PE32Header::PE32;<br>
PE->ImageBase = Config->ImageBase;<br>
- PE->SectionAlignment = SectionAlignment;<br>
- PE->FileAlignment = FileAlignment;<br>
+ PE->SectionAlignment = PageSize;<br>
+ PE->FileAlignment = SectorSize;<br>
PE->MajorImageVersion = Config->MajorImageVersion;<br>
PE->MinorImageVersion = Config->MinorImageVersion;<br>
PE->MajorOperatingSystemVersion = Config->MajorOSVersion;<br>
<br>
Modified: lld/trunk/test/COFF/baserel.test<br>
URL: <a href="http://llvm.org/viewvc/llvm-project/lld/trunk/test/COFF/baserel.test?rev=244691&r1=244690&r2=244691&view=diff" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project/lld/trunk/test/COFF/baserel.test?rev=244691&r1=244690&r2=244691&view=diff</a><br>
==============================================================================<br>
--- lld/trunk/test/COFF/baserel.test (original)<br>
+++ lld/trunk/test/COFF/baserel.test Tue Aug 11 18:09:00 2015<br>
@@ -61,7 +61,7 @@<br>
# BASEREL-HEADER-NEXT: VirtualSize: 0x20<br>
# BASEREL-HEADER-NEXT: VirtualAddress: 0x5000<br>
# BASEREL-HEADER-NEXT: RawDataSize: 512<br>
-# BASEREL-HEADER-NEXT: PointerToRawData: 0x1800<br>
+# BASEREL-HEADER-NEXT: PointerToRawData: 0xC00<br>
# BASEREL-HEADER-NEXT: PointerToRelocations: 0x0<br>
# BASEREL-HEADER-NEXT: PointerToLineNumbers: 0x0<br>
# BASEREL-HEADER-NEXT: RelocationCount: 0<br>
<br>
Modified: lld/trunk/test/COFF/hello32.test<br>
URL: <a href="http://llvm.org/viewvc/llvm-project/lld/trunk/test/COFF/hello32.test?rev=244691&r1=244690&r2=244691&view=diff" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project/lld/trunk/test/COFF/hello32.test?rev=244691&r1=244690&r2=244691&view=diff</a><br>
==============================================================================<br>
--- lld/trunk/test/COFF/hello32.test (original)<br>
+++ lld/trunk/test/COFF/hello32.test Tue Aug 11 18:09:00 2015<br>
@@ -38,8 +38,8 @@ HEADER-NEXT: MajorImageVersion: 0<br>
HEADER-NEXT: MinorImageVersion: 0<br>
HEADER-NEXT: MajorSubsystemVersion: 6<br>
HEADER-NEXT: MinorSubsystemVersion: 0<br>
-HEADER-NEXT: SizeOfImage: 20480<br>
-HEADER-NEXT: SizeOfHeaders: 4096<br>
+HEADER-NEXT: SizeOfImage: 16896<br>
+HEADER-NEXT: SizeOfHeaders: 512<br>
HEADER-NEXT: Subsystem: IMAGE_SUBSYSTEM_WINDOWS_CUI (0x3)<br>
HEADER-NEXT: Characteristics [ (0x8140)<br>
HEADER-NEXT: IMAGE_DLL_CHARACTERISTICS_DYNAMIC_BASE (0x40)<br>
<br>
<br>
_______________________________________________<br>
llvm-commits mailing list<br>
<a href="mailto:llvm-commits@lists.llvm.org" target="_blank">llvm-commits@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits</a><br>
</blockquote></div></div></div><br></div></div>
</blockquote></div><br></div></div></div></div></div></div>
</blockquote></div></div></div><br></div></div>
</blockquote></div><br></div>
</div></div></blockquote></div></div></div><br></div></div>
</blockquote></div><br></div></div>
</blockquote></div></div></div>
</blockquote></div><br></div></div>