[lldb] Clear cached unwind plans when adding symbols #125839
Merged
+125
−26
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
PR llvm#86603 broke unwinding in for unwind info added via "target symbols add". llvm#86770 attempted to fix this, but the fix was only partial -- it accepted new sources of unwind information, but didn't take into account that the symbol file can alter what lldb percieves as function boundaries. A stripped file will not contain information about private (non-exported) symbols, which will make the public symbols appear very large. If lldb tries to unwind from such a function before symbols are added, then the cached unwind plan will prevent new (correct) unwind plans from being created. target-symbols-add-unwind.test might have caught this, were it not for the fact that the "image show-unwind" command does *not* use cached unwind information (it recomputes it from scratch). The changes in this patch come in three pieces: - Clear cached unwind plans when adding symbols. Since the symbol boundaries can change, we cannot trust anything we've computed previously. - Add a flag to "image show-unwind" to display the cached unwind information (mainly for the use in the test, but I think it's also generally useful). - Rewrite the test to better and more reliably simulate the real-world scenario: I've swapped the running process for a core (minidump) file so it can run anywhere; used the caching version of the show-unwind command; and swapped C for assembly to better control the placement of symbols
|
@llvm/pr-subscribers-lldb Author: Pavel Labath (labath) ChangesPR #86603 broke unwinding in for unwind info added via "target symbols add". #86770 attempted to fix this, but the fix was only partial -- it accepted new sources of unwind information, but didn't take into account that the symbol file can alter what lldb percieves as function boundaries. A stripped file will not contain information about private (non-exported) symbols, which will make the public symbols appear very large. If lldb tries to unwind from such a function before symbols are added, then the cached unwind plan will prevent new (correct) unwind plans from being created. target-symbols-add-unwind.test might have caught this, were it not for the fact that the "image show-unwind" command does not use cached unwind information (it recomputes it from scratch). The changes in this patch come in three pieces:
Full diff: https://github.com/llvm/llvm-project/pull/125839.diff 6 Files Affected:
diff --git a/lldb/include/lldb/Symbol/UnwindTable.h b/lldb/include/lldb/Symbol/UnwindTable.h
index 29b7fa61b4849e0..3166fdec6ebaa77 100644
--- a/lldb/include/lldb/Symbol/UnwindTable.h
+++ b/lldb/include/lldb/Symbol/UnwindTable.h
@@ -38,8 +38,9 @@ class UnwindTable {
ArmUnwindInfo *GetArmUnwindInfo();
SymbolFile *GetSymbolFile();
- lldb::FuncUnwindersSP GetFuncUnwindersContainingAddress(const Address &addr,
- SymbolContext &sc);
+ lldb::FuncUnwindersSP
+ GetFuncUnwindersContainingAddress(const Address &addr,
+ const SymbolContext &sc);
bool GetAllowAssemblyEmulationUnwindPlans();
diff --git a/lldb/source/Commands/CommandObjectTarget.cpp b/lldb/source/Commands/CommandObjectTarget.cpp
index d8265e41a7384eb..db142e8f0eaee77 100644
--- a/lldb/source/Commands/CommandObjectTarget.cpp
+++ b/lldb/source/Commands/CommandObjectTarget.cpp
@@ -3471,6 +3471,17 @@ class CommandObjectTargetModulesShowUnwind : public CommandObjectParsed {
m_type = eLookupTypeFunctionOrSymbol;
break;
+ case 'c':
+ bool value, success;
+ value = OptionArgParser::ToBoolean(option_arg, false, &success);
+ if (success) {
+ m_cached = value;
+ } else {
+ return Status::FromErrorStringWithFormatv(
+ "invalid boolean value '%s' passed for -G option", option_arg);
+ }
+ break;
+
default:
llvm_unreachable("Unimplemented option");
}
@@ -3482,6 +3493,7 @@ class CommandObjectTargetModulesShowUnwind : public CommandObjectParsed {
m_type = eLookupTypeInvalid;
m_str.clear();
m_addr = LLDB_INVALID_ADDRESS;
+ m_cached = false;
}
llvm::ArrayRef<OptionDefinition> GetDefinitions() override {
@@ -3494,6 +3506,7 @@ class CommandObjectTargetModulesShowUnwind : public CommandObjectParsed {
// parsing options
std::string m_str; // Holds name lookup
lldb::addr_t m_addr = LLDB_INVALID_ADDRESS; // Holds the address to lookup
+ bool m_cached = false;
};
CommandObjectTargetModulesShowUnwind(CommandInterpreter &interpreter)
@@ -3583,9 +3596,12 @@ class CommandObjectTargetModulesShowUnwind : public CommandObjectParsed {
if (abi)
start_addr = abi->FixCodeAddress(start_addr);
- FuncUnwindersSP func_unwinders_sp(
- sc.module_sp->GetUnwindTable()
- .GetUncachedFuncUnwindersContainingAddress(start_addr, sc));
+ UnwindTable &uw_table = sc.module_sp->GetUnwindTable();
+ FuncUnwindersSP func_unwinders_sp =
+ m_options.m_cached
+ ? uw_table.GetFuncUnwindersContainingAddress(start_addr, sc)
+ : uw_table.GetUncachedFuncUnwindersContainingAddress(start_addr,
+ sc);
if (!func_unwinders_sp)
continue;
diff --git a/lldb/source/Commands/Options.td b/lldb/source/Commands/Options.td
index 777f8c36c4916ca..8831fed38435b54 100644
--- a/lldb/source/Commands/Options.td
+++ b/lldb/source/Commands/Options.td
@@ -965,6 +965,8 @@ let Command = "target modules show unwind" in {
def target_modules_show_unwind_address : Option<"address", "a">, Group<2>,
Arg<"AddressOrExpression">, Desc<"Show unwind instructions for a function "
"or symbol containing an address">;
+ def target_modules_show_unwind_cached : Option<"cached", "c">,
+ Arg<"Boolean">, Desc<"Show cached unwind information">;
}
let Command = "target modules lookup" in {
diff --git a/lldb/source/Symbol/UnwindTable.cpp b/lldb/source/Symbol/UnwindTable.cpp
index da88b0c9c4ea191..61d51192bf3d1e4 100644
--- a/lldb/source/Symbol/UnwindTable.cpp
+++ b/lldb/source/Symbol/UnwindTable.cpp
@@ -86,6 +86,7 @@ void UnwindTable::Initialize() {
void UnwindTable::ModuleWasUpdated() {
std::lock_guard<std::mutex> guard(m_mutex);
m_scanned_all_unwind_sources = false;
+ m_unwinds.clear();
}
UnwindTable::~UnwindTable() = default;
@@ -118,7 +119,7 @@ UnwindTable::GetAddressRange(const Address &addr, const SymbolContext &sc) {
FuncUnwindersSP
UnwindTable::GetFuncUnwindersContainingAddress(const Address &addr,
- SymbolContext &sc) {
+ const SymbolContext &sc) {
Initialize();
std::lock_guard<std::mutex> guard(m_mutex);
diff --git a/lldb/test/Shell/SymbolFile/Inputs/target-symbols-add-unwind.c b/lldb/test/Shell/SymbolFile/Inputs/target-symbols-add-unwind.c
deleted file mode 100644
index 237c8ce181774d9..000000000000000
--- a/lldb/test/Shell/SymbolFile/Inputs/target-symbols-add-unwind.c
+++ /dev/null
@@ -1 +0,0 @@
-int main() {}
diff --git a/lldb/test/Shell/SymbolFile/target-symbols-add-unwind.test b/lldb/test/Shell/SymbolFile/target-symbols-add-unwind.test
index 5420213d405e868..249d2ef79a891c4 100644
--- a/lldb/test/Shell/SymbolFile/target-symbols-add-unwind.test
+++ b/lldb/test/Shell/SymbolFile/target-symbols-add-unwind.test
@@ -1,27 +1,107 @@
-# TODO: When it's possible to run "image show-unwind" without a running
-# process, we can remove the unsupported line below, and hard-code an ELF
-# triple in the test.
-# UNSUPPORTED: system-windows, system-darwin
-
-# RUN: cd %T
-# RUN: %clang_host %S/Inputs/target-symbols-add-unwind.c -g \
-# RUN: -fno-unwind-tables -fno-asynchronous-unwind-tables \
-# RUN: -o target-symbols-add-unwind.debug
-# RUN: llvm-objcopy --strip-debug target-symbols-add-unwind.debug \
-# RUN: target-symbols-add-unwind.stripped
-# RUN: %lldb target-symbols-add-unwind.stripped -s %s -o quit | FileCheck %s
-
-process launch --stop-at-entry
-image show-unwind -n main
-# CHECK-LABEL: image show-unwind -n main
+# NB: The minidump core file exists only because "image show-unwind" currently
+# requires a process to exist. If that changes, it can be removed.
+
+# REQUIRES: x86, lld
+
+# RUN: split-file %s %t
+# RUN: yaml2obj %t/a.core.yaml -o %t/a.core
+# RUN: %clang -c --target=x86_64-pc-linux %t/a.s -o %t/a.o
+# RUN: ld.lld --shared %t/a.o -o %t/a.debug --build-id=0xdeadbeef \
+# RUN: --image-base=0x10000
+# RUN: llvm-objcopy --strip-all %t/a.debug %t/a.stripped
+# RUN: cd %t
+# RUN: %lldb -c %t/a.core \
+# RUN: -o "settings set interpreter.stop-command-source-on-error false" \
+# RUN: -s %t/commands -o quit | FileCheck %s
+
+#--- commands
+
+image add a.stripped
+image load --file a.stripped --slide 0
+image list
+# CHECK-LABEL: image list
+# CHECK: [ 0] DEADBEEF 0x0000000000010000 a.stripped
+
+## Due to missing symbol information this (incorrectly) prints the unwind
+## information for public_fn1
+image show-unwind -n public_fn1 --cached true
+# CHECK-LABEL: image show-unwind -n public_fn1
+# CHECK-NEXT: UNWIND PLANS for a.stripped`public_fn1 (start addr 0x12000)
# CHECK-NOT: debug_frame UnwindPlan:
-target symbols add -s target-symbols-add-unwind.stripped target-symbols-add-unwind.debug
+target symbols add -s a.stripped a.debug
# CHECK-LABEL: target symbols add
# CHECK: symbol file {{.*}} has been added to {{.*}}
-image show-unwind -n main
-# CHECK-LABEL: image show-unwind -n main
+image show-unwind -n private_fn --cached true
+# CHECK-LABEL: image show-unwind -n private_fn
+# CHECK-NEXT: UNWIND PLANS for a.stripped`private_fn (start addr 0x12010)
# CHECK: debug_frame UnwindPlan:
# CHECK-NEXT: This UnwindPlan originally sourced from DWARF CFI
# CHECK-NEXT: This UnwindPlan is sourced from the compiler: yes.
+# CHECK-NEXT: This UnwindPlan is valid at all instruction locations: no.
+# CHECK-NEXT: This UnwindPlan is for a trap handler function: no.
+# CHECK-NEXT: Address range of this UnwindPlan: [a.stripped.PT_LOAD[1]..text + 16-0x0000000000000013)
+
+
+#--- a.s
+
+ .text
+ .cfi_sections .debug_frame
+ .globl public_fn1, public_fn2
+
+ .p2align 12
+public_fn1:
+ .cfi_startproc
+ pushq %rbp
+ .cfi_def_cfa_offset 16
+ .cfi_offset %rbp, -16
+ popq %rbp
+ .cfi_def_cfa %rsp, 8
+ retq
+ .cfi_endproc
+
+ .p2align 4
+private_fn:
+ .cfi_startproc
+ pushq %rbp
+ .cfi_def_cfa_offset 16
+ .cfi_offset %rbp, -16
+ popq %rbp
+ .cfi_def_cfa %rsp, 8
+ retq
+ .cfi_endproc
+
+ .p2align 4
+public_fn2:
+ .cfi_startproc
+ pushq %rbp
+ .cfi_def_cfa_offset 16
+ .cfi_offset %rbp, -16
+ popq %rbp
+ .cfi_def_cfa %rsp, 8
+ retq
+ .cfi_endproc
+
+#--- a.core.yaml
+--- !minidump
+Streams:
+ - Type: SystemInfo
+ Processor Arch: AMD64
+ Platform ID: Linux
+ CPU:
+ Vendor ID: GenuineIntel
+ Version Info: 0x00000000
+ Feature Info: 0x00000000
+ - Type: ThreadList
+ Threads:
+ - Thread Id: 0x000074F3
+ Context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
+ Stack:
+ Start of Memory Range: 0x00007FFD4BC15080
+ Content: 30044000000000000000000000000000
+ - Type: MemoryList
+ Memory Ranges:
+ - Start of Memory Range: 0x00007FFD4BC15080
+ Content: 30044000000000000000000000000000
+...
|
|
Rather than a top-level |
|
If I was adding a new command -- sure, but this is adding only a slightly different way to print something we're already printing (*). I think it would be confusing for this to live in a separate command. Or are you proposing to move the existing command. (*) I conservatively didn't make it the default, though I think it would actually be a better default, as it shows what's actually being used by lldb. Based on this comment the main reason it's printing uncached plans is to log the workings of the instruction emulator. |
|
My bad, I misread "add a flag" as "add a command". I think moving the command and making it an alias would be a nice-to-have (in your copious amounts of spare time ;-), but obviously that's outside the scope of this PR. |
FWIW this behavior was mostly to aid in debugging unwind plan creators - you could attach to an lldb, put a breakpoint in the unwind plan creator methods, and re-do |
jasonmolenda
approved these changes
Feb 5, 2025
There was a problem hiding this comment.
LGTM. I'm fine with changing the default to show the current cached unwind plans, this command is only used by lldb maintainers (mostly, me), normal users won't notice a change here. Jonas' suggestion of moving this command under target modules dump would make sense to me too, but you shouldn't need to do that in this PR unless you want to, I know there's a lot of mechanical code moving to implement a change like that. Again this is not a normal user command, I don't mind changing its behavior or name if anyone thinks that is better.
| m_cached = value; | ||
| } else { | ||
| return Status::FromErrorStringWithFormatv( | ||
| "invalid boolean value '%s' passed for -G option", option_arg); |
There was a problem hiding this comment.
haha, you caught me.
| @@ -3494,6 +3506,7 @@ class CommandObjectTargetModulesShowUnwind : public CommandObjectParsed { | |||
| // parsing options | |||
| std::string m_str; // Holds name lookup | |||
| lldb::addr_t m_addr = LLDB_INVALID_ADDRESS; // Holds the address to lookup | |||
| bool m_cached = false; | |||
There was a problem hiding this comment.
I'm fine with changing the behavior of the existing command if you'd like to. This command is very much only used by lldb maintainers and tests.
There was a problem hiding this comment.
I may be biased because this is the second time that the command has lied to me (the first time being when the propeller thingy caused a huge unwind plan covering the whole file to be inserted into the cache), but I think this would be a better default. I'll flip the flag here. Thanks.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR #86603 broke unwinding in for unwind info added via "target symbols add". #86770 attempted to fix this, but the fix was only partial -- it accepted new sources of unwind information, but didn't take into account that the symbol file can alter what lldb percieves as function boundaries.
A stripped file will not contain information about private (non-exported) symbols, which will make the public symbols appear very large. If lldb tries to unwind from such a function before symbols are added, then the cached unwind plan will prevent new (correct) unwind plans from being created.
target-symbols-add-unwind.test might have caught this, were it not for the fact that the "image show-unwind" command does not use cached unwind information (it recomputes it from scratch).
The changes in this patch come in three pieces: