It looks like the problem that caused us to originally warn instead of
error was that of a symbol being assigned to the same version twice.
Now we don't warn if a symbol is assigned to the same version twice,
but error if it is assigned to multiple.
This fixes pr28342.
llvm-svn: 326813
We do not expand memory region correctly for following scripts:
.foo.1 :
{
*(.foo.1)
. += 0x1000;
} > ram
Patch generalizes expanding of output sections and memory
regions in one place and fixes the issue.
Differential revision: https://reviews.llvm.org/D43999
llvm-svn: 326688
"division by zero" or "modulo by zero" are not
very informative errors and even probably confusing
as does not let to know that error is coming from linker script.
Patch adds location reporting.
Differential revision: https://reviews.llvm.org/D43934
llvm-svn: 326686
This avoids the Writer unnecessarily having a member to retain ownership
of the function body.
Differential Revision: https://reviews.llvm.org/D43933
llvm-svn: 326580
LLD can not catch a memory area overflow when using a data command.
If we have the script below:
.foo :
{
*(.foo)
BYTE(0x1)
} > ram
where BYTE overflows the ram region, we do not report it currently.
Patch fixes that.
Differential revision: https://reviews.llvm.org/D43948
llvm-svn: 326545
Also remove the space between --lto-O and value, as the option with a
space is not accepted at present.
Leave --opt-remarks-filename as it does not currently accept the = form.
llvm.org/pr36563
llvm-svn: 326536
GNU linkers by convention supports both `--foo bar` and `--foo=bar` styles
for all long options that take arguments.
Differential Revision: https://reviews.llvm.org/D43972
llvm-svn: 326506
Summary:
If an executable needs text relocations, it should be marked as such so
that the loader can prepare for text relocations. We currently create a
dummy segment with DT_TEXTREL for that purpose.
Generic ABI as of 2000 [1] mentioned that "Its [DT_TEXTREL's] use
has been superseded by the DF_TEXTREL flag". However, it's actually not
superseded even after 18 years. OpenBSD and musl recognize only DT_TEXTREL.
So we still need to set both.
[1] http://www.sco.com/developers/gabi/2000-07-17/ch5.dynamic.html
Reviewers: rafael
Subscribers: emaste, llvm-commits, arichardson
Differential Revision: https://reviews.llvm.org/D43920
llvm-svn: 326503
The argument is section=address, not just address.
(For compatibility with GNU linkers we need to, but do not yet, accept
--section-start=section=address.)
llvm-svn: 326463
We intend to maintain compatibility with GNU ld, and in the GNU world
long options are conventionally specified as --long-option=value. For
whatever reason GNU ld.bfd accepts both --long-option value and
--long-option=value, but documents the former. Follow suit.
Some lld long options do not accept the = form; this is probably a bug
to be fixed (along with a man page update).
Reported by Ingo Schwarze, for --entry.
llvm-svn: 326459
For now this is NFC, but this small refactor opens the door to
letting us embed a hash of the PDB in the build id field of the
PDB.
Differential Revision: https://reviews.llvm.org/D43913
llvm-svn: 326453
With the current code if the script has a PHDRS we always obey and try
to allocate a header. This can cause Min - HeaderSize to underflow.
It looks like bfd actually prints an error for this case. With this
patch we do the same.
Found while looking at pr36515.
llvm-svn: 326441
LLD crashes with broken scripts shown in testcase,
because fails to read memory regon name and accesses
MemoryRegions's element which is nullptr.
Patch fixes it.
Differential revision: https://reviews.llvm.org/D43866
llvm-svn: 326431
This is PR36515.
Currenly if we have a script like .debug_info 0 : { *(.debug_info) },
we would not remove this section and keep it in the output.
That does not work, because it is common case for
debug sections to have a zero address expression.
Patch changes behavior so that we remove only sections
that do not use symbols in its expressions.
Differential revision: https://reviews.llvm.org/D43863
llvm-svn: 326430
Previously, we didn't try to make effort to put .note sections next
to each other in the output file, so two .note sections were likely
to be stored to two separate NOTE segments. That's undesirable because
we should create as few segments as possible in general.
Differential Revision: https://reviews.llvm.org/D43858
llvm-svn: 326410
It now includes both linkerscript and non linkerscript variants.
Extracted from a patch by Rui while I was trying to figure out what
exactly was changing.
llvm-svn: 326409
Summary:
This change removes large "echo" commands from the test by writing
tests themselves as linker scripts.
Reviewers: rafael
Subscribers: emaste, javed.absar, llvm-commits, arichardson
Differential Revision: https://reviews.llvm.org/D43900
llvm-svn: 326403
Some linker script test cases contain only a few lines of assembly
and a long linker script. Such tests are easier to maintain if we
write the main test file as a linkier script instead of assembly.
Differential Revision: https://reviews.llvm.org/D43887
llvm-svn: 326363
When merging in types from a type server PDB, we would use a
pointer into the type server PDB's mapped file buffer directly
to avoid copying data. However, we would close the type server
PDB after we finished merging in its types, which would unmap
all of its memory. This would lead to a use after free.
We fix this by making a strong reference in the PDBLinker class
to all referenced type server PDBs, thereby making it safe to
hold pointers into its memory mapped contents.
This fixes llvm.org/pr36455
Differential Revision: https://reviews.llvm.org/D43834
llvm-svn: 326345
This should resolve the issue that lld build fails in some hosts
that uses case-insensitive file system.
Differential Revision: https://reviews.llvm.org/D43788
llvm-svn: 326339
Let X and Y be types. Previously, functions F(X, Y) and G(Y, X) had
the same hash value because their hash values are computed as follows:
hash(F) = hash(X) + hash(Y)
hash(G) = hash(Y) + hash(X)
This patch fixes the issue by using hash_combine.
Differential Revision: https://reviews.llvm.org/D43856
llvm-svn: 326336
Update LLD test expectations for new symbol ordering introduced by
Differential D43685.
Differential Revision: https://reviews.llvm.org/D43875
llvm-svn: 326335
We should process symbols inside output section declarations the same way as top-level ones.
Differential Revision: https://reviews.llvm.org/D43008
llvm-svn: 326305
This patch simplifies initializeSymbols. Since that function is called
at the tail context of ObjFile::parse, and the function is called only
once from that function, that's effectively just a continuation of
ObjFile::parse. So this patch merge it with ObjFile::parse.
Differential Revision: https://reviews.llvm.org/D43848
llvm-svn: 326296
The problem I want to address now is that chunks have too many data
members for "offsets", and their origins are not well defined.
For example, InputSegment has OutputSegmentOffset, but it's base class
also has OutputOffset. That's very confusing.
Differential Revision: https://reviews.llvm.org/D43726
llvm-svn: 326291
These output section names are ELF-specific. We shouldn't have this rule
for WebAssembly.
Differential Revision: https://reviews.llvm.org/D43712
llvm-svn: 326289
SubSection inherited from SyntheticSection, and SyntheticSection inherits
from OutputSection, so SubSection was an OutputSection. But that's wrong
because SubSection is not actually a WebAssembly output section.
It shares some functionalities with OutputSection, but overall it's very
different.
This patch removes that inheritance.
Differential Revision: https://reviews.llvm.org/D43719
llvm-svn: 326286
FileOutputBuffer automatically removes an existing file, so we don't
need to do that. Actually doing that is discouraged because when the
linker fails to create an output for some reason after instantiating
FileOutputBufffer, FileOutputBuffer removes a temporary file and don't
touch an existing file. That's an desired behavior from the user's
point of view.
(Internally, FileOutputBuffer writes its contents to a temporary file
and then rename it over to an existing file on commit()).
Differential Revision: https://reviews.llvm.org/D43728
llvm-svn: 326281
Looks like these accessor functions are a bit overly defensive, and
due to the amount of code, that part isn't easy to read. We have
code to log translation results in writeTo and writeRelocations, so
I don't think we need to log it again in these functions.
I think the new function is much easier to understand.
Differential Revision: https://reviews.llvm.org/D43713
llvm-svn: 326277
Instead of {Function,Global,Data}Symbol, use Defined{Function,Global,Data}
because undefined symbol should never reach this function.
Differential Revision: https://reviews.llvm.org/D43710
llvm-svn: 326275
I think calling reserve() for each object file is too many and isn't useful.
We can add reserve() later. By default, we shouldn't add reserve() to a lot
of places.
Differential Revision: https://reviews.llvm.org/D43722
llvm-svn: 326273
Previously, one function adds all types of undefined symbols. That
doesn't fit to the wasm's undefined symbol semantics well because
different types of undefined symbols are very different in wasm.
As a result, separate control flows merge in this addUndefined function
and then separate again for each type. That wasn't easy to read.
This patch separates the function into three functions. Now it is pretty
clear what we are doing for each undefined symbol type.
Differential Revision: https://reviews.llvm.org/D43697
llvm-svn: 326271
This means we don't need to write the linking metadata section
at all for executable (non-relocatable) output.
Differential Revision: https://reviews.llvm.org/D42869
llvm-svn: 326268
Summary:
Some of the tests invoke llvm-readelf. This currently appears to work
probably because the LLVM binary directory is included in $PATH. However,
this is quite fragile so let's just make lit expand the full path.
Reviewers: ruiu, espindola
Reviewed By: ruiu
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43827
llvm-svn: 326257
Imagine that we have sections A, B, C, where A == C and
symbol ordering file containing symbols: symC, symB, symA
Previously because of ICF it was possible that final order would be
B, A or B, C. That violates order specified in ordering file.
Patch changes that.
Differential revision: https://reviews.llvm.org/D43234
llvm-svn: 326179
It should be possible to resolve undefined symbols in dynamic libraries
using symbols defined in a linker script.
Differential Revision: https://reviews.llvm.org/D43011
llvm-svn: 326176
These are required for handling circular dependencies between
static libraries, which is something that lld-link always does without
any extra parameters.
Differential Revision: https://reviews.llvm.org/D43786
llvm-svn: 326172
This fixes pr36475.
I think this code can be simplified a bit, but I would like to check
in the more direct fix if we are in agreement on the direction and
then refactor.
This is not something that bfd does. The issue is not noticed in bfd
because it keeps fewer sections from the linkerscript in the output.
The reasons why it seems reasonable to do this:
- As George noticed, we would still keep the flags if the output
section had both an empty synthetic section and a regular section
- We need an heuristic to find the flags of output sections. Using the
flags of a synthetic section that would have been there seems a
reasonable heuristic.
llvm-svn: 326137
This patch removes the HashString function from StringExtraces and
replaces its uses with calls to djbHash from DJB.h.
This change is *almost* NFC. While the algorithm is identical, the
djbHash implementation in StringExtras used 0 as its default seed while
the implementation in DJB uses 5381. The latter has been shown to result
in less collisions and improved avalanching and is used by the DWARF
accelerator tables.
Because some test were implicitly relying on the hash order, I've
reverted to using zero as a seed for the following two files:
lld/include/lld/Core/SymbolTable.h
llvm/lib/Support/StringMap.cpp
Differential revision: https://reviews.llvm.org/D43615
llvm-svn: 326091
It looks like some of our tests depend on the ordering of hashed values.
I'm reverting my changes while I try to reproduce and fix this locally.
Failing builds:
lab.llvm.org:8011/builders/lld-x86_64-darwin13/builds/18388
lab.llvm.org:8011/builders/clang-cmake-x86_64-sde-avx512-linux/builds/6743
lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-windows10pro-fast/builds/15607
llvm-svn: 326082
This removes the HashString function from StringExtraces and replaces
its uses with calls to djbHash from DJB.h
This is *almost* NFC. While the algorithm is identical, the djbHash
implementation in StringExtras used 0 as its seed while the
implementation in DJB uses 5381. The latter has been shown to result in
less collisions and improved avalanching.
https://reviews.llvm.org/D43615
(cherry picked from commit 77f7f965bc9499a9ae768a296ca5a1f7347d1d2c)
llvm-svn: 326081
MIPS ABIs require that if an executable file uses non-PIC model, the
EI_ABIVERSION entry in the ELF header should be incremented from 0 to 1.
That allows obsoleted / limited dynamic linkers refuse to link them.
llvm-svn: 325890
This continues direction started in D43069.
We can keep sections that are explicitly assigned to segment in script.
It helps to simplify code.
Differential revision: https://reviews.llvm.org/D43571
llvm-svn: 325887
Latest patch version now.
Original commit message:
[ELF] - Do not crash with --emit-relocs and --icf=all together.
Previously we would crash because did not mark .rel[a] sections
as dead and they tried to access parent which was not live
after ICF and therefore was null.
Differential revision: https://reviews.llvm.org/D43241
llvm-svn: 325879
Previously we would crash because did not mark .rel[a] sections
as dead and they tried to access parent which was not live
after ICF and therefore was null.
Differential revision: https://reviews.llvm.org/D43241
llvm-svn: 325877
This is for fixing PR36297.
Issue itself is that if we have SECTIONS { .bar (a+b) : { *(.stub) } };
script and no section .stub, when LLD will remove .bar, but
produce output with undefined symbols a and b.
Differential revision: https://reviews.llvm.org/D43069
llvm-svn: 325875
This is relative to "Bug 36166 - tools/gold/X86/comdat.ll is failing only on Debian Unstable" (PR36166).
Something changed in newer versions of gold and now together with gold-plugin
there is an issue shown in PR, symbol may get wrong visibility.
LLD works fine, but we have no testcase for the same use case, patch adds it.
Differential revision: https://reviews.llvm.org/D43193
llvm-svn: 325874
This responds to PR36475,
r325763 led to unexprected layout change, though
new behavior seems to be more correct.
Previously we could have following script:
.foo : { *(.foo) }
.bar : { *(.synthetic_empty) BYTE(0x11) }}
where synthetic_empty is a synthetic section which is empty and
hence removed by linker.
Before r325763 .bar would receive section flags from .synthetic_empty,
but after this revision it receives flags the same as .foo section has.
It is the same as if there would not be any synthetic_empty section in a script,
so looks reasonable and consistent behavior:
.foo : { *(.foo) }
.bar : { BYTE(0x11) }}
Patch adds testcase to document it.
Differential revision: https://reviews.llvm.org/D43632
llvm-svn: 325873
We have an internal program that does't link without this patch. I don't
know of any open-source program that needs this, but there might be.
Since this patch improves compatibility with GNU linkers with a few lines
of code, I think it's worth to be committed.
The problem is about undefined symbols in DSOs. Some programs depend on
the GNU linkers' behavior that they pull out object files from archive
files to resolve undefined symbols in DSOs. We already allow that kind of
"reverse" dependency (from DSOs to the main executable) for regular
symbols, in particular, for "__progname" symbol (which is usually in
crt0.o), but that doesn't work if the symbol is in an archive file.
This patch is to make it work.
Differential Revision: https://reviews.llvm.org/D43658
llvm-svn: 325849
We sometimes need to iterate over input sections for a given
output section. It is not very convinent because we have to iterate
over section descriptions.
Patch introduces getInputSections helper, it simplifies things.
Differential revision: https://reviews.llvm.org/D43574
llvm-svn: 325763
This reverts commit r325679 that was committed without discussion.
Actually, in the discussion thread, most people opposed to have this
option in lld. Reverting that change doesn't mean that this is a
final decision, but that needs to be discussed first.
llvm-svn: 325714
Address @ruiu's post commit review comment about a value which is intended
to be a unsigned 32 bit integer as using uint32_t rather than unsigned.
llvm-svn: 325713
lld uses an arena allocator to one of allocations
like these can just use make<>.
Differential Revision: https://reviews.llvm.org/D43587
llvm-svn: 325706
In r324043, --nopie was renamed to --no-pie to presumably fix a typo.
As it turns out, "nopie" wasn't a typo but the spelling used by
OpenBSD's binutils ld. Gold on the other hand spells the flag "no-pie".
(Vanilla binutils doesn't have a flag like this at all.)
Since they do the same thing, let's support both spellings.
llvm-svn: 325679
This removes script input file and inlines script into
testcase body. That is consistent with othet LS tests
and makes testcase easier to read.
llvm-svn: 325673
Summary:
This fixes two failing tests on Windows with an installed version of python that has spaces in the path.
* elf/lto/cache.ll
* mach-o/dependency_info.yaml
Reviewers: zturner, llvm-commits, stella.stamenova
Subscribers: emaste, arichardson
Differential Revision: https://reviews.llvm.org/D43265
llvm-svn: 325650
This patch provides migitation for CVE-2017-5715, Spectre variant two,
which affects the P5600 and P6600. It implements the LLD part of
-z hazardplt. Like the Clang part of this patch, I have opted for that
specific option name in case alternative migitation methods are required
in the future.
The mitigation strategy suggested by MIPS for these processors is to use
hazard barrier instructions. 'jalr.hb' and 'jr.hb' are hazard
barrier variants of the 'jalr' and 'jr' instructions respectively.
These instructions impede the execution of instruction stream until
architecturally defined hazards (changes to the instruction stream,
privileged registers which may affect execution) are cleared. These
instructions in MIPS' designs are not speculated past.
These instructions are defined by the MIPS32R2 ISA, so this mitigation
method is not compatible with processors which implement an earlier
revision of the MIPS ISA.
For LLD, this changes PLT stubs to use 'jalr.hb' and 'jr.hb'.
Reviewers: atanasyan, ruiu
Differential Revision: https://reviews.llvm.org/D43488
llvm-svn: 325647
Purely a rename in preparation for adding new global symbol type.
We want to use GlobalSymbol to represent real wasm globals and
DataSymbol for pointers to things in linear memory (what ELF would
call STT_OBJECT).
This reduces the size the patch to add the explicit symbol table
which is coming soon!
Differential Revision: https://reviews.llvm.org/D43476
llvm-svn: 325645
Previously wasm used a separate header to declare markLive
and ELF used to declare ICF. This change makes each backend
consistently declare these in their own headers.
Differential Revision: https://reviews.llvm.org/D43529
llvm-svn: 325631
The profailing style in lld seem to be to not include such empty lines.
Clang-tidy/clang-format seem to handle this just fine.
Differential Revision: https://reviews.llvm.org/D43528
llvm-svn: 325629
I think if statements that end with return is easier to read than
cascaded if-else-if-else-if statements because it makes clear that
execution of a function terminates there. This patch also adds more
blank lines to separate code blocks to meaningful pieces.
Differential Revision: https://reviews.llvm.org/D43517
llvm-svn: 325625
Summary:
With D43396, no clients use the Path parameter anymore.
This is the lld side fix with D43400.
Depends on D43396 and D43400.
Reviewers: pcc
Subscribers: emaste, inglorion, arichardson, llvm-commits
Differential Revision: https://reviews.llvm.org/D43401
llvm-svn: 325620
This header used to be needed here for the `OutRelocation` struct
but that no longer exists.
Differential Revision: https://reviews.llvm.org/D43516
llvm-svn: 325608
Instead include InputFuction and InputSegment directly
in the subclasses that use them (DefinedFunction and
DefinedGlobal).
Differential Revision: https://reviews.llvm.org/D43493
llvm-svn: 325603
This patch removes `OutRelocations` vector from the InputChunk and
directly consume `Relocations` vector instead. This should make the linker
faster because we don't create a temporary data structure, and it matches
the lld's design principle that we don't create temporary data structures
for object files but instead directly consume mmap'ed data whenever possible.
Differential Revision: https://reviews.llvm.org/D43491
llvm-svn: 325549
Summary:
Before the name of the function sounded like it was just a getter for the
private class member Addend. However, it actually calculates the final
value for the r_addend field in Elf_Rela that is used when writing the
.rela.dyn section. I also added a comment to the UseSymVA member to
explain how it interacts with computeAddend().
Differential Revision: https://reviews.llvm.org/D43161
llvm-svn: 325485
We already have isa<> for this, and these methods were simply
duplicating those redundantly.
Differential Revision: https://reviews.llvm.org/D43422
llvm-svn: 325418
Currently, archive file name is missing in this message. In general,
we should avoid constructing strings in an ad-hoc manner and instead
use toString() to get consistent output strings.
Differential Revision: https://reviews.llvm.org/D43420
llvm-svn: 325416
Rafael Espindola