Commit Graph

271 Commits

Author SHA1 Message Date
Igor Kudrin
f6f45476f5 [ELF2] Remove Writer::getVAStart, use TargetInfo::getVAStart instead.
llvm-svn: 252593
2015-11-10 08:39:27 +00:00
Rafael Espindola
ea7a1e9092 Round up the memsize of PT_TLS.
This is cleaner than computing relocations as if we had done it.

While at it, keep a single Phdr variable instead of multiple fields of it.

llvm-svn: 252352
2015-11-06 22:14:44 +00:00
Igor Kudrin
15cd9ffd1e [ELF2] Add GOT section for MIPS target.
This patch implements R_MIPS_GOT16 relocation for global symbols in order to
generate some entries in GOT. Only reserved and global entries are supported
for now. For the detailed description about GOT in MIPS, see "Global Offset
Table" in Chapter 5 in the followin document:
ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf

In addition, the platform specific symbol "_gp" is added, see "Global Data
Symbols" in Chapter 6 in the aforementioned document.

Differential revision: http://reviews.llvm.org/D14211

llvm-svn: 252275
2015-11-06 07:43:03 +00:00
Michael J. Spencer
6c34eff536 [elf2] Fix R_X86_64_TPOFF32 handling.
For x86-64 the initial executable TLS block is placed directly before the
thread specific data register so compilers can directly access it via
R_X86_64_TPOFF32. Generate the correct (negative) offset for this case.

llvm-svn: 252131
2015-11-05 02:00:35 +00:00
Rafael Espindola
5fbe008eb7 Simplify the logic to avoid the Closed set.
IMHO this makes the code easier to read: at each iteration we add
a section to a PT_LOAD and increase its size.

llvm-svn: 252011
2015-11-03 23:19:42 +00:00
Michael J. Spencer
2c6f213bcd [elf2] Use value-initialization instead of memset.
llvm-svn: 252006
2015-11-03 22:43:11 +00:00
Michael J. Spencer
d77f0d2526 [elf2] Implement R_X86_64_TPOFF32.
This does not support TPOFF32 relocations to local symbols as the address calculations are separate. Support for this will be a separate patch.

llvm-svn: 251998
2015-11-03 22:39:09 +00:00
Rafael Espindola
61362ce876 Handle 0 sized sections like any other section.
This is a case where there is inconsistency among ELF linkers:

* The spec says nothing special about empty sections.
* BFD ld removes them.
* Gold handles them like regular sections.

We were outputting them but sometimes ignoring them. This would create
odd looking outputs where a rw section could be in a ro segment for example.

The bfd way of doing things is also strange for the case where a symbol
points to the empty section.

Now we match gold and what seems to be the intention of the spec.

llvm-svn: 251988
2015-11-03 22:08:08 +00:00
Rafael Espindola
443f50a833 Create .bss only when needed.
This is a small complication, but produces nicer output and is a step to
handling zero size sections uniformly.

llvm-svn: 251980
2015-11-03 21:35:14 +00:00
Rafael Espindola
b8c3b1523f Fix thinko.
This would match SHF_ALLOC or SHF_TLS. We want both.

llvm-svn: 251939
2015-11-03 16:30:27 +00:00
Rafael Espindola
9eb3e57042 Simplify and fix style.
llvm-svn: 251927
2015-11-03 15:46:15 +00:00
Rafael Espindola
9b89608698 Remove a redundant boolean.
llvm-svn: 251921
2015-11-03 14:34:11 +00:00
Rafael Espindola
115f0f365d Two small fixes to copy relocation processing.
* We can determine the alignment requirement.
* We have to update the bss alignment with it.

llvm-svn: 251914
2015-11-03 14:13:40 +00:00
Michael J. Spencer
78aa1de3ae [elf2] Generate PT_TLS.
Differential Revision: http://reviews.llvm.org/D14167

llvm-svn: 251872
2015-11-03 00:34:39 +00:00
Igor Kudrin
2169b1bd26 [ELF2] Ensure that .dynsym section is finalized before .gnu.hash.
It is required to fill up the GNU hash table section before its
finalize() method is called.

Differential Revision: http://reviews.llvm.org/D14196

llvm-svn: 251789
2015-11-02 10:46:14 +00:00
George Rimar
bc590feb2b [ELF2] R_X86_64_COPY relocation implemented
Differential revision: http://reviews.llvm.org/D14090.

llvm-svn: 251526
2015-10-28 16:48:58 +00:00
Michael J. Spencer
4633a373e0 [elf2] Don't allocate VA space for TLS NOBITS sections.
Differential Revision: http://reviews.llvm.org/D13838

llvm-svn: 251454
2015-10-27 21:22:54 +00:00
Rui Ueyama
9eb7ed0423 ELF2: Move Writer class to the top of the file. NFC.
llvm-svn: 251195
2015-10-24 18:22:59 +00:00
Rui Ueyama
6621d8ed7c ELF2: Add comments and simplify. NFC.
llvm-svn: 251194
2015-10-24 17:57:40 +00:00
Rui Ueyama
157c433a83 ELF2: Remove setNameOffset and combine that with writeHeader. NFC.
llvm-svn: 251193
2015-10-24 17:57:39 +00:00
Rui Ueyama
9fbb3d8850 ELF2: Rename StringTableSection::getFileOff -> getOffset.
getFileOff functions defined for other classes return an offset
from beginning of the file. StringTableSection's getFileOff however
returned an offset from beginning of the section. That was confusing.

llvm-svn: 251192
2015-10-24 17:44:52 +00:00
George Rimar
5dad7c1d4c Reapply r251144: fixed test.
llvm-svn: 251184
2015-10-24 08:52:46 +00:00
Rui Ueyama
c27d2f266e ELF2: Remove FIXMEs which we are not sure if we have to fix.
llvm-svn: 251173
2015-10-24 01:46:40 +00:00
Rui Ueyama
e08cd678de ELF2: Simplify file magic setter.
llvm-svn: 251159
2015-10-23 22:44:39 +00:00
Rui Ueyama
bcb2d0f94b ELF2: Factor out getEntryAddr from writeHeader. NFC.
llvm-svn: 251158
2015-10-23 22:44:37 +00:00
Rui Ueyama
c611d97006 ELF2: Use character literals for ELF file magic.
This is consitent with Support/ELF.h and more obvious than before.

llvm-svn: 251155
2015-10-23 22:31:47 +00:00
Rui Ueyama
803195ed6d ELF2: Simplify Writer::assignAddresses. NFC.
This patch is an attempt to simplify assignAddresses function by splitting
it and using less variables. I tried to split the code to create PHDRs from
the code to assign addresses, but it didn't make this code simpler, so I
didn't do that in this patch.

llvm-svn: 251152
2015-10-23 21:45:59 +00:00
Rui Ueyama
b621d1ae58 Revert "[ELF2] --strip-all/-s command line implemented"
This reverts commit r251144 because the new test is failing.

llvm-svn: 251148
2015-10-23 21:08:41 +00:00
George Rimar
236bf3bfe4 [ELF2] --strip-all/-s command line implemented
-s, --strip-all - Strip all symbols
Implementation removes .strtab and .symtab sections from output.

Differential Revision: http://reviews.llvm.org/D13934

llvm-svn: 251144
2015-10-23 20:24:44 +00:00
Rui Ueyama
c4aaed9255 ELF2: Implement --gc-sections.
Section garbage collection is a feature to remove unused sections
from outputs. Unused sections are sections that cannot be reachable
from known GC-root symbols or sections. Naturally the feature is
implemented as a mark-sweep garbage collector.

In this patch, I added Live bit to InputSectionBase. If and only
if Live bit is on, the section will be written to the output.
Starting from GC-root symbols or sections, a new function, markLive(),
visits all reachable sections and sets their Live bits. Writer then
ignores sections whose Live bit is off, so that such sections are
excluded from the output.

This change has small negative impact on performance if you use
the feature because making sections means more work. The time to
link Clang changes from 0.356s to 0.386s, or +8%.

It reduces Clang size from 57,764,984 bytes to 55,296,600 bytes.
That is 4.3% reduction.

http://reviews.llvm.org/D13950

llvm-svn: 251043
2015-10-22 18:49:53 +00:00
Igor Kudrin
1b0d7066ff [ELF2] Add support for GNU Hash section
This patch implements --hash-style command line switch.

* By default, or with "sysv" or "both" parameters, the linker generates
  a standard ELF hash section.
* With "gnu" or "both", it produces a GNU-style hash section.

That section requires the symbols in the dynamic symbol table section, which
are referenced in the GNU hash section, to be placed after not hashed ones and
to be sorted to correspond the order of hash buckets in the GNU Hash section.

The division function, as well as estimations for the section's parameters,
are just the first rough attempt and the subjects for further adjustments.

Differential Revision: http://reviews.llvm.org/D13815

llvm-svn: 251000
2015-10-22 08:21:35 +00:00
Igor Kudrin
ab665fc475 [ELF2] Determine the order of entries of symbol tables in the finalize() phase.
* Move the responsibility to call SymbolBody::setDynamicSymbolTableIndex()
  from the hash table to the dynamic symbol table.
* Hash table is not longer responsible for filling the dynamic symbol table.
* The final order of symbols of both symbol tables is set before writing
  phase starts.
* Remove repeaded scan of the symbol table during writting SymbolTableSection.

Differential Revision: http://reviews.llvm.org/D13911

llvm-svn: 250864
2015-10-20 21:47:58 +00:00
George Rimar
0f5ac9f571 [ELF2] .shstrtab section implemented
The section header table index of the entry that is associated with the section name string table.

Differential Revision: http://reviews.llvm.org/D13904

llvm-svn: 250836
2015-10-20 17:21:35 +00:00
George Rimar
648a2c37fb [ELF2] - Lazy relocation support for x86_64.
Target has supportsLazyRelocations() method which can switch lazy relocations on/off (currently all targets are OFF except x64 which is ON). So no any other targets are affected now.

Differential Revision: http://reviews.llvm.org/D13856?id=37726

llvm-svn: 250808
2015-10-20 08:54:27 +00:00
Rafael Espindola
c159c967f6 Add support for merging the contents of SHF_MERGE sections.
For now SHF_STRINGS are not supported.

llvm-svn: 250737
2015-10-19 21:00:02 +00:00
Rafael Espindola
334c3e11e6 Delay the relocation scan.
We have to scan the relaxations after every symbol is defined.

This fixes pr25218.

llvm-svn: 250702
2015-10-19 15:21:42 +00:00
Michael J. Spencer
1bf73000cf [elf2] Sort allocated TLS sections at the start of the R/W PT_LOAD.
Differential Revision: http://reviews.llvm.org/D13763

llvm-svn: 250572
2015-10-16 23:11:07 +00:00
Rui Ueyama
6455852a28 ELF2: Treat IsMips64EL as a global configuration.
If one file is MIPS64EL, all files are MIPS64EL, and vice versa.
We do not have to look up MIPS-ness for each file. Currently we
do not support 64-bit MIPS, so the config value is always false.

llvm-svn: 250566
2015-10-16 22:51:43 +00:00
Rui Ueyama
c7cc6ecf08 ELF2: Use ELFT to template OutputSections.
This patch is to use ELFT instead of Is64Bits to template OutputSection
and its subclasses. This increases code size slightly because it creates
two identical functions for some classes, but that's only 20 KB out of
33 MB, so it's negligible.

This is as per discussion with Rafael. He's not fan of the idea but OK
with this. We'll revisit later to this topic.

llvm-svn: 250466
2015-10-15 22:27:29 +00:00
Rui Ueyama
c6ef3f28fc ELF2: Do not treat .strtab specially when creating sections.
String table is added to end of the file so that all the other sections
are finalized before string table. But we can just add section names to
the string table before calling finalize() on any section instead.

llvm-svn: 250463
2015-10-15 21:50:30 +00:00
Rui Ueyama
2317d0d4d6 Remove a getter/setter that don't hide anything.
llvm-svn: 250458
2015-10-15 20:55:22 +00:00
Rui Ueyama
96b190428e Fix style.
llvm-svn: 250457
2015-10-15 20:55:20 +00:00
Rafael Espindola
b26d807033 Use OutputSectionBase in a few more places. NFC.
Just making room for having another output section type.

llvm-svn: 250443
2015-10-15 19:31:49 +00:00
Rui Ueyama
d9189ceb23 ELF2: Implement __start_SECNAME and __stop_SECNAME.
If a section name is valid as a C identifier (which is rare because of
the leading '.'), linkers are expected to define __start_<secname> and
__stop_<secname> symbols. They are at beginning and end of the section,
respectively. This is not requested by the ELF standard, but GNU ld and
gold provide this feature.

llvm-svn: 250432
2015-10-15 17:11:03 +00:00
Rafael Espindola
ae81a7bf49 Use OutputSectionBase in a few cases where we don't need a OutputSection.
NFC. This is just preparation for adding a new OutputSection dedicated to
SHF_MERGE input sections.

llvm-svn: 250419
2015-10-15 15:29:53 +00:00
Rafael Espindola
37f0b7d28a Rename CAlloc since it allocates sections, not chunks.
llvm-svn: 250395
2015-10-15 12:33:04 +00:00
Rafael Espindola
bdc7620191 Remove unnecessary llvm:: prefix.
llvm-svn: 250394
2015-10-15 12:30:29 +00:00
Hal Finkel
736c741752 [ELF2] Set a target-specific VAStart for PPC64
After some additional post-commit (post-revert) discussion and research, this
reverts, in part, r250205, so the ABI-recommended starting address can be used
on PPC64 (as is done by other linkers).

Also, this addresses the FIXME in ELF/Writer.cpp by making VAStart a
target-dependent property.

llvm-svn: 250378
2015-10-15 07:49:07 +00:00
Rui Ueyama
2b675074fe ELF2: Support --entry=<addr>.
If an argument for --entry is a number, that's not a symbol name but
an absolute address. If that's the case, the address is directly set
to ELF header's e_entry.

llvm-svn: 250334
2015-10-14 22:20:57 +00:00
Rui Ueyama
4fcadaf5e7 ELF2: Merge .{text,rodata,data,bss}.* sections.
Previously, we used input section names as output section names.
That resulted that we created lots of sections for comdat
or -f{function,data}-section sections.

This patch reduces the number of sections by dropping suffix from
all section names which start with ".text.", ".rodata.", ".data."
or ".bss.". GNU linker does this using the internal linker script,
but for LLD I chose to do that directly.

Interestingly, this makes the linker faster. Time to link Clang
is this.

Before:

  real    0m0.537s
  user    0m0.433s
  sys     0m0.104s

After:

  real    0m0.390s
  user    0m0.268s
  sys     0m0.120s

It make sense because previously we created 57659 sections now only 27.

llvm-svn: 250315
2015-10-14 19:21:25 +00:00