Summary:
This is the third patch to improve module loading in a series that started here (where I explain the motivation and solution): D62499
Add functions to read the r_debug location to know where the linked list of loaded libraries are so I can generate the `xfer:libraries-svr4` packet.
I'm also using this function to implement `GetSharedLibraryInfoAddress` that was "not implemented" for linux.
Most of this code was inspired by the current ds2 implementation here: https://github.com/facebook/ds2/blob/master/Sources/Target/POSIX/ELFProcess.cpp.
Reviewers: clayborg, xiaobai, labath
Reviewed By: clayborg, labath
Subscribers: emaste, krytarowski, mgorny, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D62501
llvm-svn: 363458
Previously it was storing a *pointer*, which left open the possibility
of this pointer being null. We never made use of that possibility (it
does not make sense), and most of the code was already assuming that.
However, there were a couple of null-checks scattered around the code.
This patch replaces the reference with a pointer, making the
non-null-ness explicit, and removes the remaining null-checks.
llvm-svn: 363381
In a dwo/debug_types world, the die offset is not enough to uniquely
idendify a debug info entry. Pass the the entire DIERef object instead.
This is technically NFC, because only AppleIndex implemented this
method (and there, the die offset *is* enough for unique
identification). However, this makes the code simpler, and simplifies
some of the follow-up patches.
llvm-svn: 363373
Summary:
this revision adds Lexing, Parsing and Basic Semantic for the consteval specifier as specified by http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p1073r3.html
with this patch, the consteval specifier is treated as constexpr but can only be applied to function declaration.
Changes:
- add the consteval keyword.
- add parsing of consteval specifier for normal declarations and lambdas expressions.
- add the whether a declaration is constexpr is now represented by and enum everywhere except for variable because they can't be consteval.
- adapt diagnostic about constexpr to print constexpr or consteval depending on the case.
- add tests for basic semantic.
Reviewers: rsmith, martong, shafik
Reviewed By: rsmith
Subscribers: eraman, efriedma, rnkovacs, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D61790
llvm-svn: 363362
Summary:
Type units don't represent actual compilations and a lot of the
operations that we do with lldb compile units (getting their line
tables, variables, etc.) don't make sense for them. There is also a lot
more of them (sometimes over 100x), so making them more lightweight pays
off.
The main change in this patch is that we stop creating lldb CompileUnits
for DWARF type units. The trickiest part here is that the SymbolFile
interface requires that we assign consecutive sequence IDs to the
compile units we create. As DWARF type and compile units can come in any
order (in v5), this means we can no longer use 1-1 mapping between DWARF
and lldb compile units. Instead I build a translation table between the
two indices. To avoid pessimizing the case where there are no type
units, I build the translation table only in case we have at least one
type unit.
Additionaly, I also tried to strenghted type safete by replacing
DWARFUnit with DWARFCompileUnit where applicable. Though that was not
stricly necessary, I found it a good way to ensure that the
transformations I am doing here make sense. In the places where I was
changing the function signatures, and where it was obvious that the
objects being handled were not null, I also replaced pointers with
references.
There shouldn't be any major functional change with this patch. The only
change I observed is that now the types in the type units will not be
parsed when one calls Module::ParseAllDebugSymbols, unless they are
referenced from other compile units. This makes sense, given how
ParseAllDebugSymbols is implemented (it iterates over all compile
units), and it only matters for one hand-writted test where I did not
bother to reference the types from the compile units (which I now do).
Reviewers: clayborg, JDevlieghere, aprantl
Subscribers: jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D63005
llvm-svn: 363250
This replaces the `info` typedef with a nested struct named Info. This
means we now have FooProvider and FooProvider::Info, instead of two
related but separate classes FooProvider and FooInfo. This change is
mostly cosmetic.
llvm-svn: 363211
with a call to snprintf() to find the size of the formatted string,
malloc memory, then snprintf again to format it into the buffer, instead
of calling asprintf.
Orig commit msg:
When reading ObjC class table, use new SPI if it is avail
In the latest OS betas, the objc runtime has a special interface
for the debugger, class_getNameRaw(), instead of the existing
class_getName(), which will return class names in their raw, unmangled
(in the case of swift) form. When lldb can access the unmangled
names of classes, it won't need to fetch them out of the inferior
process after we run our "get the objc class table" expression.
If the new interface is absent (debugging a process on an older
target), lldb will fall back to class_getName and reading any class
names that it got back in demangled form, at a bit of a performance
cost on the first expression.
<rdar://problem/50688054>
llvm-svn: 363206
Summary:
PersistentStateExpressions (e.g. ClangPersistentVariables) have the
ability to define types using expressions that persist throughout the
debugging session. GetCompilerTypeFromPersistentDecl is a useful
operation to have if you need to use any of those persistently declared types,
like in CommandObjectMemory.
This decouples clang from CommandObjectMemory and decouples Plugins from
Commands in general.
Differential Revision: https://reviews.llvm.org/D62797
llvm-svn: 363183
Summary:
This patch creates a cache of file lists in line tables referenced by
type units. This cache is used to avoid parsing a line table twice
(since a file list will generally be shared by many type units).
It also sets things up in a way that parsing of DW_AT_decl_file
attributes will keep working even when we stop creating lldb compile
units for dwarf type units, but it stops short of actually doing that.
This means that the request for files now go directly to SymbolFileDWARF
instead of being routed there indirectly via the
lldb_private::CompileUnit class.
As a result of this, a number of occurences of SymbolContext variables
in DWARFASTParserClang have become unused, so I remove them.
This patch reduces the number of times a file list is being parsed, but
the situation is still suboptimal, as the parsed list is being copied
multiple times. This will be fixed when we stop creating CompileUnits
for DWARF type units.
Reviewers: clayborg, aprantl, JDevlieghere
Subscribers: jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D62894
llvm-svn: 363143
In the latest OS betas, the objc runtime has a special interface
for the debugger, class_getNameRaw(), instead of the existing
class_getName(), which will return class names in their raw, unmangled
(in the case of swift) form. When lldb can access the unmangled
names of classes, it won't need to fetch them out of the inferior
process after we run our "get the objc class table" expression.
If the new interface is absent (debugging a process on an older
target), lldb will fall back to class_getName and reading any class
names that it got back in demangled form, at a bit of a performance
cost on the first expression.
<rdar://problem/50688054>
llvm-svn: 363103
Summary:
This is the second patch to improve module loading in a series that started here (where I explain the motivation and solution): https://reviews.llvm.org/D62499
I need to read the aux vector to know where the r_debug map with the loaded libraries are.
The AuxVector class was made generic so it could be reused between the POSIX-DYLD plugin and NativeProcess*. The class itself ended up in the ProcessUtility plugin.
Reviewers: clayborg, xiaobai, labath, JDevlieghere
Reviewed By: clayborg, labath, JDevlieghere
Subscribers: emaste, JDevlieghere, mgorny, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D62500
llvm-svn: 363098
This is a followup to rL362981, in which I moved GetObjCLanguageRuntime
from Process to ObjCLanguageRuntime, renaming it to Get along the way.
llvm-svn: 362984
Summary:
This is the first of a few patches I have to improve the performance of dynamic module loading on Android.
In this first diff I'll describe the context of my main motivation and will then link to it in the other diffs to avoid repeating myself.
## Motivation
I have a few scenarios where opening a specific feature on an Android app takes around 40s when lldb is attached to it. The reason for that is because 40 modules are dynamicly loaded at that point in time and each one of them is taking ~1s.
## The problem
To learn about new modules we have a breakpoint on a linker function that is called twice whenever a module is loaded. One time just before it's loaded (so lldb can check which modules are loaded) and another right after it's loaded (so lldb can check again which ones are loaded and calculate the diference).
It's figuring out which modules are loaded that is taking quite some time. This is currently done by traversing the linked list of loaded shared libraries that the linker maintains in memory. Each item in the linked list requires its own `x` packet sent to the gdb server (this is android so the network also plays a part). In my scenario there are 400+ loaded libraries and even though we read 0x800 worth of bytes at a time we still make ~180 requests that end up taking 150-200ms.
We also do this twice, once before the module is loaded (state = eAdd) and another right after (state = eConsistent) which easly adds up to ~400ms per module.
## A solution
**Implement `xfer:libraries-svr4` in lldb-server:**
I noticed in the code that loads the new modules that it had support for the `xfer:libraries-svr4` packet (added ~4 years ago to support the ds2 debug server) but we didn't support it in lldb-server. This single packet returns an xml list of all the loaded modules by the process. The advantage is that there's no more need to make 180 requests to read the linked list. Additionally this new requests takes around 10ms.
**More efficient usage of the `xfer:libraries-svr4` packet in lldb:**
When `xfer:libraries-svr4` is available the Process class has a `LoadModules` function that requests this packet and then loads or unloads modules based on the current list of loaded modules by the process.
This is the function that is used by the DYLDRendezvous class to get the list of loaded modules before and after the module is loaded. However, this is really not needed since the LoadModules function already loaded or unloaded the modules accordingly. I changed this strategy to call LoadModules only once (after the process has loaded the module).
**Bugs**
I found a few issues in lldb while implementing this and have submitted independent patches for them.
I tried to devide this into multiple logical patches to make it easier to review and discuss.
## Tests
I wanted to put these set of diffs up before having all the tests up and running to start having them reviewed from a techical point of view. I'm also having some trouble making the tests running on linux so I need more time to make that happen.
# This diff
The `xfer` packages follow the same protocol, they are requested with `xfer:<object>:<read|write>:<annex>:<offset,length>` and a return that starts with `l` or `m` depending if the offset and length covers the entire data or not. Before implementing the `xfer:libraries-svr4` I refactored the `xfer:auxv` to generically handle xfer packets so we can easly add new ones.
The overall structure of the function ends up being:
* Parse the packet into its components: object, offset etc.
* Depending on the object do its own logic to generate the data.
* Return the data based on its size, the requested offset and length.
Reviewers: clayborg, xiaobai, labath
Reviewed By: labath
Subscribers: mgorny, krytarowski, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D62499
llvm-svn: 362982
Summary:
In an effort to make Process more language agnostic, I removed
GetCPPLanguageRuntime from Process. I'm following up now with an equivalent
change for ObjC.
Differential Revision: https://reviews.llvm.org/D63052
llvm-svn: 362981
Reflow the text for the table to make the table legible. This is purely
cosmetic, but makes understanding the contents of the table easier. NFCI.
llvm-svn: 362961
Ensure that errors are passed through correctly when performing
register read/write operations. Currently, any ptrace() errors are
silently discarded and LLDB behaves as if operation was successful.
Differential Revision: https://reviews.llvm.org/D63054
llvm-svn: 362946
Summary:
Using llvm-style rtti gives us stronger guarantees around casting
LanguageRuntimes.
As discussed in D62755
Differential Revision: https://reviews.llvm.org/D62934
llvm-svn: 362884
This reverts commit 58afc1bdebf9fa8b178d6c9d89af94c5cc091760.
This commit caused the test suite on macOS to fail many tests. It
appears that setting breakpoints is the issue. One example that fails
is the lit test Breakpoint/case-sensitive.test.
llvm-svn: 362862
lldb/lit/SymbolFile/DWARF/debug-types-expressions.test fails with msan.
This change fixes the issue by ensuring FileNameEntry::checksum is
always default-initialized.
llvm-svn: 362843
Summary:
The DWARFCompileUnit is set as the "user data" of the lldb compile unit
directly in the constructor (see ParseCompileUnit).
This means that instead of going through unit indexes, we can just fetch
the DWARF unit directly from there.
Reviewers: clayborg, JDevlieghere
Subscribers: aprantl, jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D62943
llvm-svn: 362783
Summary:
Once we've attached to the process we load all current modules and also set a breakpoint at the rendezvous break address.
However, we don't do this if we already have a load address for the image info address (e.g.: DT_DEBUG on ELF). This code was added 4 years ago when adding support for `$qXfer:Libraries:` packet (https://reviews.llvm.org/D9471) but its intention is not 100% clear to me. It seems to me we're using that check to know if the modules have already been loaded (which they have if `$qXfer:Libraries:` is supported by the gdb server) and skip loading the modules again in the following `if` block. The problem is that we also skip setting the Rendezvous breakpoint so we stop knowing when the process loads new modules.
I fix this by moving the call to set the breakpoint to the end of the function so we always call it as long as we have a valid executable.
Reviewers: ADodds, clayborg, eugene, labath
Reviewed By: eugene, labath
Subscribers: lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D62168
llvm-svn: 362619
Summary:
r362103 exposed a bug, where we could read incorrect data if a skeleton
unit contained more than the single unit DIE. Clang emits these kinds of
units with -fsplit-dwarf-inlining (which is also the default).
Changing lldb to handle these DIEs is nontrivial, as we'd have to change
the UID encoding logic to be able to reference these DIEs, and fix up
various places which are assuming that all DIEs come from the separate
compile unit.
However, it turns out this is not necessary, as the DWO unit contains
all the information that the skeleton unit does. So, this patch just
skips parsing the extra DIEs if we have successfully found the DWO file.
This enforces the invariant that the rest of the code is already
operating under.
This patch fixes a couple of existing tests, but I've also included a
simpler test which does not depend on execution of binaries, and would
have helped us in catching this sooner.
Reviewers: clayborg, JDevlieghere, aprantl
Subscribers: probinson, dblaikie, lldb-commits
Differential Revision: https://reviews.llvm.org/D62852
llvm-svn: 362586
Summary:
I want to remove this method because I think that Process should be
language agnostic, or at least, not have knowledge about specific language
runtimes. There is "GetLanguageRuntime()" which should be used instead. If the
caller a CPPLanguageRuntime, they should cast it as needed. Ideally, this
should only happen in plugins that need C++ specific knowledge.
The next step I would like to do is remove "GetObjCLanguageRuntime()" as well.
There are a lot more instances of that function being used, so I wanted to
upload this one first to get the general reception to this idea.
Reviewers: compnerd, davide, JDevlieghere, jingham, clayborg, labath, aprantl
Subscribers: lldb-commits
Differential Revision: https://reviews.llvm.org/D62755
llvm-svn: 362544
Add a function to flatten the nested aggregate type
Differential Revision: https://reviews.llvm.org/D62702
Patch by Wanyi Ye <kusmour@gmail.com>
llvm-svn: 362543
The x86 assembly inspection engine has code to support detecting a
mid-function epilogue that ends in a RET instruction; add support for
recognizing an epilogue that ends in a JMP, and add a check that the
unwind state has been restored to the original stack setup; reinstate
the post-prologue unwind state after this JMP instruction.
The assembly inspection engine used for other architectures,
UnwindAssemblyInstEmulation, detects mid-function epilogues by
tracking branch instructions within the function and "forwards"
the current unwind state to the targets of the branches. If
an epilogue unwinds the stack and exits, followed by a branch
target, we get back to the correct unwind state. The x86
unwinder should move to this same algorithm, or possibly even
look at implementing an x86 instruction emulation plugin and
get UnwindAssemblyInstEmulation to work for x86 too. I added
a branch instruction recognizier method that will be necessary
if we want to switch the algorithm.
Differential Revision: https://reviews.llvm.org/D62764
<rdar://problem/51074422>
llvm-svn: 362456
Change rL362280 changed CodeView API getRegisterNames() by adding an input
parameter in CPUType. It is called in LLDB and needs to be updated.
Differential Revision: https://reviews.llvm.org/D62772
llvm-svn: 362349
There is likely also an underlying bug in all code that calls
CPlusPlusNameParser with nullptrs, but this patch can also stand for
itself.
rdar://problem/49072829
llvm-svn: 362177
Summary:
My main goal here is to make lldb-server work with Android Studio.
This is currently not the case because lldb-server is started in platform mode listening on a domain socket. When Android Studio connects to it lldb-server crashes because even though it's listening on a domain socket as soon as it gets a connection it asserts that it's a TCP connection, which will obviously fails for any non-tcp connection.
To do this I came up with a new method called GetConnectURI() in Socket that returns the URI needed to connect to the connected portion of the socket.
Reviewers: labath, clayborg, xiaobai
Reviewed By: labath
Subscribers: mgorny, jfb, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D62089
llvm-svn: 362173
Changes:
- update comments to detail the info can come from .debug_info or .debug_types
- Rename "debug_info_data" to "data" now that we can get data from .debug_info or .debug_types.
- Also call DWARFDebugInfoEntry::GetAbbreviationDeclarationPtr(...) instead of manually grabbing abbreviation.
llvm-svn: 362116
The issue was caused by the error checking code that was added. It was incorrectly adding an extra abbreviation when DWARFEnumState::Complete was received since it would push an extra abbreviation onto the list with the abbreviation code of zero. This cause m_idx_offset in each DWARFAbbreviationDeclarationSet to be set to UINT32_MAX. This valid indicates we must linearly search for attributes, not access them in O(1) time. This caused every DWARFDebugInfoEntry that would try to get its DWARFAbbreviationDeclaration from the CU's DWARFAbbreviationDeclarationSet to always linearly search the abbreviation set for a given abbreviation code. Easy to see why this would cause things to be slow.
This regression was caused by: https://reviews.llvm.org/D59370. I asked to ensure there was no regression is parsing or access speed, but that must not have been done. In my test with 40 DWARF files trying to set a breakpoint by function name and in a header file, I see a 8% speed improvement with this fix.
There was no regression in correctness, just very inefficient access.
Added full unit testing for DWARFAbbreviationDeclarationSet parsing to ensure this doesn't regress.
Differential Revision: https://reviews.llvm.org/D62630
llvm-svn: 362105
When LLDB first started we didn't have our mmap of the DWARF data done correctly and if the backing file would change we would get live changes as the file changed and it would cause problems. We now mmap correctly and do not run into these issues. There was legacy code in DWARFDebugInfoEntry::GetAbbreviationDeclarationPtr(...) that would always extract the abbrev index each time the function was called to verify that DWARF data hadn't changed and a warning was emitted if it did. We no longer need this and the code was removed. The other thing this function did when it parsed the abbrev index was give us the offset of the first attribute bytes by adding the LEB128 size to the offset. This required an extra parameter to DWARFDebugInfoEntry::GetAbbreviationDeclarationPtr(...) which is now removed. I added "lldb::offset_t DWARFDebugInfoEntry::GetFirstAttributeOffset() const" which calculates this when we need it and modified all sites that need the offset to call it.
Now that we aren't decoding and verifying the abbrev index, it speeds up DWARF access by 1% to 2%.
Differential Revision: https://reviews.llvm.org/D62634
llvm-svn: 362103
Summary:
The ParseTypeFromDWARF function consists of a huge switch on the kind of
type being parsed. Each case in this switch starts with parsing the
attributes of the current DIE. A lot of these attributes are specific to
one kind of a type, but a lot of them are common too, leading to code
duplication.
This patch reduces the duplication (and the size of ParseTypeFromDWARF)
by moving the attribute parsing to a separate function. It creates a
struct (ParsedTypeAttributes), which contains a parsed form of all
attributes which are useful for parsing any kind of a type. The parsing
code for a specific type kind can then access the fields which are
relevant for that specific case.
Reviewers: JDevlieghere, clayborg, aprantl
Subscribers: jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D62477
llvm-svn: 362075
The function Extract() is almost a duplicate of FastExtract() but is not used.
Delete it and rename FastExtract() to Extract().
Reviewed By: JDevlieghere
Differential Revision: https://reviews.llvm.org/D62593
llvm-svn: 362049
Summary:
This code is modifying a support file list after it has been created.
This makes it hard to share the file list between type units and
compile units in DWARF. It's not a total showstopper, but supporting
this while also sharing the lists would make things more complicated.
Given that this was added to support a project which never fully
materialised, and that even back then there were some concerns about the
correctness of this approach (according to D11003#200772 the compile
unit name is not guaranteed to be the first one in the support file
list), I think we should just delete this workaround.
Reviewers: clayborg, tberghammer, dsrbecky
Subscribers: aprantl, lldb-commits
Differential Revision: https://reviews.llvm.org/D62517
llvm-svn: 361948
Summary:
The function was not being consistent in how it retrieved the language
of the current compile unit. Sometimes it did so from the lldb CU
object, and sometimes from the DWARF die. This patch unifies the
handling on the latter. The reason for choosing the DWARF method is
because I'd eventually like to stop creating lldb CUs for dwarf type
units (and so this code needs to would need to work without them).
Reviewers: clayborg, JDevlieghere, aprantl
Subscribers: jdoerfert, lldb-commits
Differential Revision: https://reviews.llvm.org/D62481
llvm-svn: 361939
Antonio Afonso