[Lldb-commits] [lldb] 18c5ad5 - [lldb] Fix block address resolution for functions in multiple sections (#137955)
via lldb-commits
lldb-commits at lists.llvm.org
Wed May 7 02:18:06 PDT 2025
Author: Pavel Labath
Date: 2025-05-07T11:18:03+02:00
New Revision: 18c5ad5c6c178365d270439742863e14c8981ea3
URL: https://github.com/llvm/llvm-project/commit/18c5ad5c6c178365d270439742863e14c8981ea3
DIFF: https://github.com/llvm/llvm-project/commit/18c5ad5c6c178365d270439742863e14c8981ea3.diff
LOG: [lldb] Fix block address resolution for functions in multiple sections (#137955)
Continuing the theme from #116777 and #124931, this patch ensures we
compute the correct address when a functions is spread across multiple
sections. Due to this, it's not sufficient to adjust the offset in the
section+offset pair (Address::Slide). We must actually slide the file
offset and then recompute the section using the result.
I found this out due to a failure to disassemble some parts of the
function, so I'm testing with that, although it's likely there are other
things that were broken due to this.
Added:
lldb/test/Shell/Commands/command-disassemble-sections.s
Modified:
lldb/include/lldb/Symbol/Block.h
lldb/source/Symbol/Block.cpp
Removed:
################################################################################
diff --git a/lldb/include/lldb/Symbol/Block.h b/lldb/include/lldb/Symbol/Block.h
index 501c912b674ac..601895631e148 100644
--- a/lldb/include/lldb/Symbol/Block.h
+++ b/lldb/include/lldb/Symbol/Block.h
@@ -354,7 +354,12 @@ class Block : public UserID, public SymbolContextScope {
// Member variables.
SymbolContextScope &m_parent_scope;
collection m_children;
+
+ /// Address ranges of this block. They are relative to the function entry
+ /// point so one must add/subtract GetFunction().GetAddress().GetFileAddress()
+ /// when converting from/to to the AddressRange representation.
RangeList m_ranges;
+
lldb::InlineFunctionInfoSP m_inlineInfoSP; ///< Inlined function information.
lldb::VariableListSP m_variable_list_sp; ///< The variable list for all local,
///static and parameter variables
diff --git a/lldb/source/Symbol/Block.cpp b/lldb/source/Symbol/Block.cpp
index 9d01293ea64e0..3de3e5eecbf35 100644
--- a/lldb/source/Symbol/Block.cpp
+++ b/lldb/source/Symbol/Block.cpp
@@ -283,39 +283,43 @@ uint32_t Block::GetRangeIndexContainingAddress(const Address &addr) {
return m_ranges.FindEntryIndexThatContains(file_addr - func_file_addr);
}
+static AddressRange ToAddressRange(const Address &func_addr,
+ const Block::Range &block_range) {
+ assert(func_addr.GetModule());
+ return AddressRange(func_addr.GetFileAddress() + block_range.base,
+ block_range.size,
+ func_addr.GetModule()->GetSectionList());
+}
+
bool Block::GetRangeAtIndex(uint32_t range_idx, AddressRange &range) {
if (range_idx >= m_ranges.GetSize())
return false;
- Function &function = GetFunction();
- const Range &vm_range = m_ranges.GetEntryRef(range_idx);
- range.GetBaseAddress() = function.GetAddress();
- range.GetBaseAddress().Slide(vm_range.GetRangeBase());
- range.SetByteSize(vm_range.GetByteSize());
+ Address addr = GetFunction().GetAddress();
+ if (!addr.GetModule())
+ return false;
+
+ range = ToAddressRange(addr, m_ranges.GetEntryRef(range_idx));
return true;
}
AddressRanges Block::GetRanges() {
+ Address addr = GetFunction().GetAddress();
+ if (!addr.GetModule())
+ return {};
+
AddressRanges ranges;
- Function &function = GetFunction();
- for (size_t i = 0, e = m_ranges.GetSize(); i < e; ++i) {
- ranges.emplace_back();
- auto &range = ranges.back();
- const Range &vm_range = m_ranges.GetEntryRef(i);
- range.GetBaseAddress() = function.GetAddress();
- range.GetBaseAddress().Slide(vm_range.GetRangeBase());
- range.SetByteSize(vm_range.GetByteSize());
- }
+ for (size_t i = 0, e = m_ranges.GetSize(); i < e; ++i)
+ ranges.push_back(ToAddressRange(addr, m_ranges.GetEntryRef(i)));
return ranges;
}
bool Block::GetStartAddress(Address &addr) {
- if (m_ranges.IsEmpty())
+ Address func_addr = GetFunction().GetAddress();
+ if (!func_addr.GetModule() || m_ranges.IsEmpty())
return false;
- Function &function = GetFunction();
- addr = function.GetAddress();
- addr.Slide(m_ranges.GetEntryRef(0).GetRangeBase());
+ addr = ToAddressRange(func_addr, m_ranges.GetEntryRef(0)).GetBaseAddress();
return true;
}
diff --git a/lldb/test/Shell/Commands/command-disassemble-sections.s b/lldb/test/Shell/Commands/command-disassemble-sections.s
new file mode 100644
index 0000000000000..d278527b898ce
--- /dev/null
+++ b/lldb/test/Shell/Commands/command-disassemble-sections.s
@@ -0,0 +1,110 @@
+## Test disassembling of functions which are spread over multiple sections (ELF
+## segments are modelled as LLDB sections).
+
+
+# REQUIRES: x86, lld
+
+# RUN: split-file %s %t
+# RUN: llvm-mc -filetype=obj -triple x86_64-pc-linux %t/file.s -o %t/file.o
+# RUN: ld.lld %t/file.o -o %t/file.out -T %t/file.lds
+# RUN: %lldb %t/file.out -o "disassemble --name func1" -o exit | FileCheck %s
+
+# CHECK: (lldb) disassemble --name func1
+# CHECK: file.out`func1:
+# CHECK-NEXT: file.out[0x0] <+0>: int $0x2a
+# CHECK: file.out`func1:
+# CHECK-NEXT: file.out[0x1000] <+4096>: int $0x2f
+
+
+#--- file.lds
+## Linker script placing the parts of the section into
diff erent segments
+## (typically one of these would be for the "hot" code).
+PHDRS {
+ text1 PT_LOAD;
+ text2 PT_LOAD;
+}
+SECTIONS {
+ . = 0;
+ .text.part1 : { *(.text.part1) } :text1
+ .text.part2 : { *(.text.part2) } :text2
+}
+
+#--- file.s
+## A very simple function consisting of two parts and DWARF describing the
+## function.
+ .section .text.part1,"ax", at progbits
+ .p2align 12
+func1:
+ int $42
+.Lfunc1_end:
+
+ .section .text.part2,"ax", at progbits
+ .p2align 12
+func1.__part.1:
+ int $47
+.Lfunc1.__part.1_end:
+
+
+
+ .section .debug_abbrev,"", at progbits
+ .byte 1 # Abbreviation Code
+ .byte 17 # DW_TAG_compile_unit
+ .byte 1 # DW_CHILDREN_yes
+ .byte 37 # DW_AT_producer
+ .byte 8 # DW_FORM_string
+ .byte 19 # DW_AT_language
+ .byte 5 # DW_FORM_data2
+ .byte 17 # DW_AT_low_pc
+ .byte 1 # DW_FORM_addr
+ .byte 85 # DW_AT_ranges
+ .byte 23 # DW_FORM_sec_offset
+ .byte 0 # EOM(1)
+ .byte 0 # EOM(2)
+ .byte 2 # Abbreviation Code
+ .byte 46 # DW_TAG_subprogram
+ .byte 0 # DW_CHILDREN_no
+ .byte 85 # DW_AT_ranges
+ .byte 23 # DW_FORM_sec_offset
+ .byte 3 # DW_AT_name
+ .byte 8 # DW_FORM_string
+ .byte 0 # EOM(1)
+ .byte 0 # EOM(2)
+ .byte 0 # EOM(3)
+
+ .section .debug_info,"", at progbits
+.Lcu_begin0:
+ .long .Ldebug_info_end0-.Ldebug_info_start0 # Length of Unit
+.Ldebug_info_start0:
+ .short 5 # DWARF version number
+ .byte 1 # DWARF Unit Type
+ .byte 8 # Address Size (in bytes)
+ .long .debug_abbrev # Offset Into Abbrev. Section
+ .byte 1 # Abbrev DW_TAG_compile_unit
+ .asciz "Hand-written DWARF" # DW_AT_producer
+ .short 29 # DW_AT_language
+ .quad 0 # DW_AT_low_pc
+ .long .Ldebug_ranges0 # DW_AT_ranges
+ .byte 2 # Abbrev DW_TAG_subprogram
+ .long .Ldebug_ranges0 # DW_AT_ranges
+ .asciz "func1" # DW_AT_name
+ .byte 0 # End Of Children Mark
+.Ldebug_info_end0:
+
+ .section .debug_rnglists,"", at progbits
+ .long .Ldebug_list_header_end0-.Ldebug_list_header_start0 # Length
+.Ldebug_list_header_start0:
+ .short 5 # Version
+ .byte 8 # Address size
+ .byte 0 # Segment selector size
+ .long 1 # Offset entry count
+.Lrnglists_table_base0:
+ .long .Ldebug_ranges0-.Lrnglists_table_base0
+.Ldebug_ranges0:
+ .byte 6 # DW_RLE_start_end
+ .quad func1
+ .quad .Lfunc1_end
+ .byte 6 # DW_RLE_start_end
+ .quad func1.__part.1
+ .quad .Lfunc1.__part.1_end
+ .byte 0 # DW_RLE_end_of_list
+.Ldebug_list_header_end0:
More information about the lldb-commits
mailing list