Previously we would merge relocation sections by name.
That did not work in some cases, like testcase shows.
Patch implements logic to merge relocation sections if their target
sections were merged into the same output section.
Differential revision: https://reviews.llvm.org/D33824
llvm-svn: 304886
This reduces how many times we have to map from OutputSection to
OutputSectionCommand. It is a required step to moving
clearOutputSections earlier.
In order to always use writeTo in OutputSectionCommand we have to call
fabricateDefaultCommands for -r links and move section compression
after it.
llvm-svn: 303784
We used to place orphans by just using compareSectionsNonScript.
Then we noticed that since linker scripts can use another order, we
should first try match the section to a given PT_LOAD. But there is
nothing special about PT_LOAD. The same issue can show up for
PT_GNU_RELRO for example.
In general, we have to search for the most similar section and put the
orphan next to it. Most similar being defined as how long they follow
the same code path in compareSecitonsNonScript.
That is what this patch does. We now compute a rank for each output
section, with a bit for each branch in what was
compareSectionsNonScript.
With this findOrphanPos is now fully general and orphan placement can
be optimized by placing every section with the same rank at once.
The included testcase is a variation of many-sections.s that uses
allocatable sections to avoid the fast path in the existing
code. Without threads it goes form 46 seconds to 0.9 seconds.
llvm-svn: 302903
We were already pretty close, the one exception was when a name was
reused in another SECTIONS directive:
SECTIONS {
.text : { *(.text) }
.data : { *(.data) }
}
SECTIONS {
.data : { *(other) }
}
In this case we would create a single .data and magically output
"other" while looking at the first OutputSectionCommand.
We now create two .data sections. This matches what gold does. If we
really want to create a single one, we should change the parser so that
the above is parsed as if the user had written
SECTIONS {
.text : { *(.text) }
.data : { *(.data) *(other)}
}
That is, there should be only one OutputSectionCommand for .data and
it would have two InputSectionDescriptions.
By itself this patch makes the code a bit more complicated, but is an
important step in allowing assignAddresses to operate just on the
linker script.
llvm-svn: 301484
Patch implements --compress-debug-sections=zlib.
In compare with D20211 (a year old patch, abandoned), it implementation
uses streaming and fully reimplemented, does not support zlib-gnu for
simplification.
This is PR32308.
Differential revision: https://reviews.llvm.org/D31941
llvm-svn: 300444
Executable sections should not be padded with zero by default. On some
architectures, 0x00 is the start of a valid instruction sequence, so can confuse
disassembly between InputSections (and indeed the start of the next InputSection
in some situations). Further, in the case of misjumps into padding, padding may
start to be executed silently.
On x86, the "0xcc" byte represents the int3 trap instruction. It is a single
byte long so can serve well as padding. This change switches x86 (and x86_64) to
use this value for padding in executable sections, if no linker script directive
overrides it. It also puts the behaviour into place making it easy to change the
behaviour of other targets when desired. I do not know the relevant instruction
sequences for trap instructions on other targets however, so somebody should add
this separately.
Because the old behaviour simply wrote padding in the whole section before
overwriting most of it, this change also modifies the padding algorithm to write
padding only where needed. This in turn has caused a small behaviour change with
regards to what values are written via Fill commands in linker scripts, bringing
it into line with ld.bfd. The fill value is now written starting from the end of
the previous block, which means that it always starts from the first byte of the
fill, whereas the old behaviour meant that the padding sometimes started mid-way
through the fill value. See the test changes for more details.
Reviewed by: ruiu
Differential Revision: https://reviews.llvm.org/D30886
Bugzilla: http://bugs.llvm.org/show_bug.cgi?id=32227
llvm-svn: 299635
Was fixed, details on review page.
Original commit message:
That removes CopyRelSection class completely, making
Bss/BssRelRo to be just regular synthetics.
This is splitted from D30541 and polished.
Difference from D30541 that all logic of SharedSymbol
converting to DefinedRegular was removed for now and
probably will be posted as separate patch.
Differential revision: https://reviews.llvm.org/D30892
llvm-svn: 298062
With this we have a single section hierarchy. It is a bit less code,
but the main advantage will be in a future patch being able to handle
foo = symbol_in_obj;
in a linker script. Currently that fails since we try to find the
output section of symbol_in_obj. With this we should be able to just
return an InputSection from the expression.
llvm-svn: 297313
In compare with D30458, this makes Bss/BssRelRo to be pure
synthetic sections.
That removes CopyRelSection class completely, making
Bss/BssRelRo to be just regular synthetics.
SharedSymbols involved in creating copy relocations are
converted to DefinedRegular, what also simplifies things.
Differential revision: https://reviews.llvm.org/D30541
llvm-svn: 297008
With this we complete the transition out of special output sections,
and with the previous patches it should be possible to merge
OutputSectionBase and OuputSection.
llvm-svn: 296023
With the current design an InputSection is basically anything that
goes directly in a OutputSection. That includes plain input section
but also synthetic sections, so this should probably not be a
template.
llvm-svn: 295993
With a synthetic merge section we can have, for example, a single
.rodata section with stings, fixed sized constants and non merge
constants.
I can be simplified further by not setting Entsize, but that is
probably better done is a followup patch.
This should allow some cleanup in the linker script code now that
every output section command maps to just one output section.
llvm-svn: 294005
[ELF] Fixed formatting. NFC
and
[ELF] Bypass section type check
Differential revision: https://reviews.llvm.org/D28761
They do the opposite of what was asked for in the code review.
llvm-svn: 293320
When reserving copy relocation space for a shared symbol, scan the DSO's
program headers to see if the symbol is in a read-only segment. If so,
reserve space for that symbol in a new synthetic section named .bss.rel.ro
which will be covered by the relro program header.
This fixes the security issue disclosed on the binutils mailing list at:
https://sourceware.org/ml/libc-alpha/2016-12/msg00914.html
Differential Revision: https://reviews.llvm.org/D28272
llvm-svn: 291524
After Mark's patch I was wondering what was the rationale for the ELF
spec requiring us to merge only sections with matching flags and
types. I tried emailing
https://groups.google.com/forum/#!forum/generic-abi, but looks like my
emails are not being posted (the list is probably moderated). I
emailed Cary Coutant instead.
Cary pointed out that the section was a late addition and didn't got
the scrutiny it deserved. Given that and the problems found by
implementing the letter of the standard, I propose changing lld to
merge all sections with the same name and issue errors if the types or
some critical flags are different.
This should allow an unmodified firefox linked with lld to run.
This also merges some code with the linkerscript path.
llvm-svn: 291107
That variable was of type DenseMap<StringRef, unsigned>, but the
unsigned numbers needed to be monotonicly increasing numbers because
the implementation that used the variable depended on that fact.
That was an implementation detail and shouldn't have leaked into Config.
This patch simplifies its type to std::vector<StringRef>.
llvm-svn: 290151
MergeOutputSection class was a bit hard to use because it provdes
a series of finalize functions that have to be called in a right way
at a right time. It also intereacted with MergeInputSection, and the
logic was somewhat entangled between the two classes.
This patch simplifies it by providing only one finalize function.
Now, all you have to do is to call MergeOutputSection::finalize
when you have added all sections to the output section. Then, it
internally merges strings and initliazes StringPiece objects.
I think this is much easier to understand.
This patch also adds comments.
llvm-svn: 287314
This patch introduces the following changes:
- DynamicSection now inherits InputSection<ELFT> and was moved
to SyntheticSections.h/.cpp.
- Link and Entsize fields of DynamicSection are propagated to
its output section
- In<ELFT>::SyntheticSections was removed.
- Finalization of synthetic sections was removed from
OutputSection<ELFT>::finalize. Now finalizeSyntheticSections is
used instead.
Differential revision: https://reviews.llvm.org/D26603
llvm-svn: 286950
Patch allows to pass a symbols file to linker.
LLD will map symbols to sections and sort sections
in output according to symbol ordering file.
That can help to reduce the startup time and/or
amount of pagefaults during startup.
Also, interesting benchmark result was produced by Rafael Espíndola.
After applying the symbols file for clang he timed compiling
X86MCTargetDesc.ii to an object file.
The page faults went from just
56,988 to 56,946 since most faults are not in the binary.
Running time went from 4.403053515 to 4.178112244.
The speedup seems to be because of better cache
locality.
Differential revision: https://reviews.llvm.org/D26130
llvm-svn: 286440
The disadvantage is that we use uint64_t instad of uint32_t for some
value in 32 bit files. The advantage is a substantially simpler code,
faster builds and less code duplication.
llvm-svn: 286414
Previously, we have both input and output section for .MIPS.abiflags.
Now we have only one class for .MIPS.abiflags, which is MipsAbiFlagsSection.
This class is a synthetic input section.
.MIPS.abiflags sections are handled as regular sections until
the control reaches Writer. Writer then aggregates all sections
whose type is SHT_MIPS_ABIFLAGS to create a single synthesized
input section. The synthesized section is then processed normally
as if it came from an input file.
llvm-svn: 286398
Previously, we have both input and output sections for .reginfo and
.MIPS.options. Now for each such sections we have one synthetic input
sections: MipsReginfoSection and MipsOptionsSection respectively.
Both sections are handled as regular sections until the control reaches
Writer. Writer then aggregates all sections whose type is SHT_MIPS_REGINFO
or SHT_MIPS_OPTIONS to create a single synthesized input section. In that
moment Writer also save GP0 value to the MipsGp0 field of the corresponding
ObjectFile. This value required for R_MIPS_GPREL16 and R_MIPS_GPREL32
relocations calculation.
Differential revision: https://reviews.llvm.org/D26444
llvm-svn: 286397
This is similar to what was done for InputSection.
With this the various fields are stored in host order and only
converted to target order when writing.
llvm-svn: 286327
Previously, we added strings from DynamicSection::finalize().
It was a bit tricky because finalize() is supposed to fix the final
size of the section, but adding new strings would change the size of
.dynstr section. So there was a dependency between finalize functions
of .dynamic and .dynstr.
However, I noticed that we can elimiante the dependency by simply
add strings early; we don't have to do that in finalize() but can do
from DynamicSection's ctor.
This patch defines a new function, DynamicSection::addEntries, to
add .dynamic entries that doesn't depend on other sections.
llvm-svn: 285784
We are going to have many more classes for linker-synthesized
input sections, so it's worth to be added to a separate file
than to the file for regular input sections.
llvm-svn: 285740
Some MIPS relocations used to access GOT entries are able to manipulate
16-bit index. The other ones like R_MIPS_CALL_HI16/LO16 can handle
32-bit indexes. 16-bit relocations are generated by default. The 32-bit
relocations are generated by -mxgot flag passed to compiler. Usually
these relocation are not mixed in the same code but files like crt*.o
contain 16-bit relocations so even if all "user's" code compiled with
-mxgot flag a few 16-bit relocations might come to the linking phase.
Now LLD does not differentiate local GOT entries accessed via a 16-bit
and 32-bit indexes. That might lead to relocation's overflow if 16-bit
entries are allocated to far from the beginning of the GOT.
The patch introduces new "part" of MIPS GOT dedicated to the local GOT
entries accessed by 32-bit relocations. That allows to put local GOT
entries accessed via a 16-bit index first and escape relocation's overflow.
Differential revision: https://reviews.llvm.org/D25833
llvm-svn: 284809
In this patch partial gdb_index section is created.
For costructing the .gdb_index section 6 steps should be performed (details are in
SplitDebugInfo.cpp file header), this patch do first 3:
Creates proper section header.
Fills list of compilation units.
Types CU list area is not supposed to be supported, so it is ignored and therefore
can be treated as implemented either.
Differential revision: https://reviews.llvm.org/D24706
llvm-svn: 284708
Use size_t instead of ELFT::uint for the string table offset. If the
linker is built 32-bit, it can't write an output file larger than 2GB.
Other code in this area uses size_t as well.
llvm-svn: 284680
Summary:
Reclaiming the name 'CachedHashString' will let us add a type with that
name that owns its value.
Reviewers: timshen
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D25644
llvm-svn: 284434
This spreads out computing the hash and using it in a hash table. The
speedups are:
firefox
master 6.811232891
patch 6.559280249 1.03841162939x faster
chromium
master 4.369323666
patch 4.33171853 1.00868134338x faster
chromium fast
master 1.856679971
patch 1.850617741 1.00327578725x faster
the gold plugin
master 0.32917962
patch 0.325711944 1.01064645023x faster
clang
master 0.558015452
patch 0.550284165 1.01404962652x faster
llvm-as
master 0.032563515
patch 0.032152077 1.01279662275x faster
the gold plugin fsds
master 0.356221362
patch 0.352772162 1.00977741549x faster
clang fsds
master 0.635096494
patch 0.627249229 1.01251060127x faster
llvm-as fsds
master 0.030183188
patch 0.029889544 1.00982430511x faster
scylla
master 3.071448906
patch 2.938484138 1.04524944215x faster
This seems to be because we don't stall as much. When linking firefox
stalled-cycles-frontend goes from 57.56% to 55.55%.
With -O2 the difference is even more significant since we avoid
recomputing the hash. For firefox we go from 9.990295265 to
9.149627521 seconds (1.09x faster).
llvm-svn: 283367
If we two sections reside in the same PT_LOAD segment,
we compute second section using the following formula:
Off2 = Off1 + VA2 - VA1. This allows OS kernel allocating
sections correctly when loading an image.
Differential revision: https://reviews.llvm.org/D25014
llvm-svn: 282705
Rafael Espindola