fixup1! [llvm-objcopy] Add change-section-lma *+/-offset

llvm-objcopy did not support change-section-lma argument.

This patch adds support for a use case of change-section-lma, that is
shifting load address of all sections by the same offset. This seems to
be the only practical use case of change-section-lma, found in other
software such as Zephyr RTOS's build system.

This is an option that could possibly be supported in some other than
ELF formats, however this change only implements it for ELF. When used
with other formats an error message is raised.

In comparison, the behavior of GNU objcopy is inconsistent. For some ELF
files it behaves the same as described above. For others, it copies the
file without modifying the p_paddr fields when it would be expected. In
some experiments it modifies arbitrary fields in section or program
headers. It is unclear what exactly determines this.
The executable file generated by yaml2obj in this test is not parsable
by GNU objcopy. With Machine set to EM_AARCH64, the file can be parsed
and the first test in the test file completes with 0 exit code. However,
the result is rather arbitrary. AArch64 GNU objcopy subtracts 0x1000
from p_filesz and p_memsz of the first LOAD section and 0x1000 from
p_offset of the second LOAD section. It does not look meaningful.

Author: eleanor-arm

llvm/docs/CommandGuide/llvm-objcopy.rst

@@ -528,9 +528,9 @@ them.
 
  Mark all defined global symbols as weak in the output.
 
-.. option:: --change-section-lma \*{+-}<offset>
+.. option:: --change-section-lma \*{+-}<val>
 
- Currently only supports changing LMA for all sections by the same offset.
+ Shift LMA of non-zero-sized sections in the program header by ``<val>``
 
 MACH-O-SPECIFIC OPTIONS
 -----------------------

llvm/lib/ObjCopy/ConfigManager.cpp

@@ -25,7 +25,8 @@ Expected<const COFFConfig &> ConfigManager::getCOFFConfig() const {
       Common.StripNonAlloc || Common.StripSections || Common.Weaken ||
       Common.DecompressDebugSections ||
       Common.DiscardMode == DiscardType::Locals ||
-      !Common.SymbolsToAdd.empty() || Common.GapFill != 0 || Common.PadTo != 0)
+      !Common.SymbolsToAdd.empty() || Common.GapFill != 0 ||
+      Common.PadTo != 0 || Common.ChangeSectionLMAValAll != 0)
     return createStringError(llvm::errc::invalid_argument,
                              "option is not supported for COFF");
 
@@ -46,7 +47,8 @@ Expected<const MachOConfig &> ConfigManager::getMachOConfig() const {
       Common.StripNonAlloc || Common.StripSections ||
       Common.DecompressDebugSections || Common.StripUnneeded ||
       Common.DiscardMode == DiscardType::Locals ||
-      !Common.SymbolsToAdd.empty() || Common.GapFill != 0 || Common.PadTo != 0)
+      !Common.SymbolsToAdd.empty() || Common.GapFill != 0 ||
+      Common.PadTo != 0 || Common.ChangeSectionLMAValAll != 0)
     return createStringError(llvm::errc::invalid_argument,
                              "option is not supported for MachO");
 
@@ -66,7 +68,7 @@ Expected<const WasmConfig &> ConfigManager::getWasmConfig() const {
       !Common.SectionsToRename.empty() || !Common.SetSectionAlignment.empty() ||
       !Common.SetSectionFlags.empty() || !Common.SetSectionType.empty() ||
       !Common.SymbolsToRename.empty() || Common.GapFill != 0 ||
-      Common.PadTo != 0)
+      Common.PadTo != 0 || Common.ChangeSectionLMAValAll != 0)
     return createStringError(llvm::errc::invalid_argument,
                              "only flags for section dumping, removal, and "
                              "addition are supported");
@@ -94,7 +96,8 @@ Expected<const XCOFFConfig &> ConfigManager::getXCOFFConfig() const {
       Common.PreserveDates || Common.StripAllGNU || Common.StripDWO ||
       Common.StripDebug || Common.StripNonAlloc || Common.StripSections ||
       Common.Weaken || Common.StripUnneeded || Common.DecompressDebugSections ||
-      Common.GapFill != 0 || Common.PadTo != 0) {
+      Common.GapFill != 0 || Common.PadTo != 0 ||
+      Common.ChangeSectionLMAValAll != 0) {
     return createStringError(
         llvm::errc::invalid_argument,
         "no flags are supported yet, only basic copying is allowed");

llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp

@@ -671,7 +671,7 @@ static Error handleArgs(const CommonConfig &Config, const ELFConfig &ELFConfig,
   }
 
   if (Config.ChangeSectionLMAValAll != 0) {
-    for (auto &Seg : Obj.segments()) {
+    for (Segment &Seg : Obj.segments()) {
       if (Seg.FileSize > 0)
         Seg.PAddr += Config.ChangeSectionLMAValAll;
     }

llvm/test/tools/llvm-objcopy/ELF/change-section-lma-multiple-load.test

@@ -0,0 +1,55 @@
+# RUN: yaml2obj %s -o %t
+# RUN: llvm-objcopy --change-section-lma *+0x20 %t %t2
+# RUN: llvm-readelf -l %t2 | FileCheck %s --check-prefix=CHECK-PLUS
+# RUN: llvm-objcopy --change-section-lma *-0x10 %t2 %t3
+# RUN: llvm-readelf -l %t3 | FileCheck %s --check-prefix=CHECK-MINUS
+# RUN: not llvm-objcopy --change-section-lma .text3=0x5000 %t  2>&1 | FileCheck %s --check-prefix=ERR-SET-ADDRESS
+# RUN: not llvm-objcopy --change-section-lma .text3+0x30 %t  2>&1 | FileCheck %s --check-prefix=ERR-SPECIFIC-SEC
+# RUN: not llvm-objcopy --change-section-lma *+0c50 %t  2>&1 | FileCheck %s --check-prefix=ERR-INVALID-VAL
+# RUN: not llvm-objcopy --change-section-lma 0 %t  2>&1 | FileCheck %s --check-prefix=ERR-MISSING-OFFSET
+
+# CHECK-PLUS:  Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz
+# CHECK-PLUS:  PHDR           0x000002 0x0000000000001002 0x0000000000001022 0x000038 0x000000
+# CHECK-PLUS:  LOAD           0x000000 0x0000000000001000 0x0000000000001020 0x001120 0x001120
+# CHECK-PLUS:  LOAD           0x001120 0x0000000000005000 0x0000000000006020 0x001000 0x001000
+# CHECK-PLUS:  NOTE           0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000
+
+# CHECK-MINUS:  PHDR           0x000002 0x0000000000001002 0x0000000000001012 0x000038 0x000000
+# CHECK-MINUS:  LOAD           0x000000 0x0000000000001000 0x0000000000001010 0x001120 0x001120
+# CHECK-MINUS:  LOAD           0x001120 0x0000000000005000 0x0000000000006010 0x001000 0x001000
+# CHECK-MINUS:  NOTE           0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000
+
+# ERR-SET-ADDRESS: error: bad format for --change-section-lma: changing LMA to a specific value is not supported. Use *+val or *-val instead
+# ERR-SPECIFIC-SEC: error: bad format for --change-section-lma: changing a specific section LMA is not supported. Use *+val or *-val instead
+# ERR-INVALID-VAL: error: bad format for --change-section-lma: value after *+ is 0c50 when it should be integer
+# ERR-MISSING-OFFSET: error: bad format for --change-section-lma: missing LMA offset
+
+!ELF
+FileHeader:
+  Class:           ELFCLASS64
+  Data:            ELFDATA2LSB
+  Type:            ET_EXEC
+Sections:
+  - Name:          .text1
+    Type:          SHT_PROGBITS
+    Size:          0x1000
+  - Name:          .text2
+    Type:          SHT_PROGBITS
+    Size:          0x1000
+ProgramHeaders:
+  - Type:          PT_PHDR
+    FileSize:      0x38
+    Offset:        0x2
+    VAddr:         0x1002
+  - Type:          PT_LOAD
+    Offset:        0x0
+    VAddr:         0x1000
+    FirstSec:      .text1
+    LastSec:       .text1
+  - Type:          PT_LOAD
+    VAddr:         0x5000
+    PAddr:         0x6000
+    FirstSec:      .text2
+    LastSec:       .text2
+  - Type:          PT_NOTE
+    FileSize:      0x0

llvm/test/tools/llvm-objcopy/ELF/change-section-lma-single-load.test

@@ -552,31 +552,36 @@ static Error loadNewSectionData(StringRef ArgValue, StringRef OptionName,
   return Error::success();
 }
 
-static Expected<int64_t> parseAdjustSectionLMA(StringRef ArgValue,
+static Expected<int64_t> parseChangeSectionLMA(StringRef ArgValue,
                                                StringRef OptionName) {
   StringRef StringValue;
   if (ArgValue.starts_with("*+")) {
     StringValue = ArgValue.slice(2, StringRef::npos);
   } else if (ArgValue.starts_with("*-")) {
     StringValue = ArgValue.slice(1, StringRef::npos);
-  } else {
+  } else if (ArgValue.contains("=")) {
+    return createStringError(errc::invalid_argument,
+                             "bad format for " + OptionName +
+                                 ": changing LMA to a specific value is not "
+                                 "supported. Use *+val or *-val instead");
+  } else if (ArgValue.contains("+") || ArgValue.contains("-")) {
     return createStringError(errc::invalid_argument,
                              "bad format for " + OptionName +
-                                 ": it is required that all sections "
-                                 "are either incremented, or decremented at "
-                                 "the same time; use *+val, "
-                                 "or *-val");
+                                 ": changing a specific section LMA is not "
+                                 "supported. Use *+val or *-val instead");
   }
   if (StringValue.empty())
     return createStringError(errc::invalid_argument,
                              "bad format for " + OptionName +
-                                 ": missing offset of LMA");
-
-  auto SLMAV = getAsInteger<int64_t>(StringValue);
-  if (!SLMAV)
-    return createStringError(SLMAV.getError(),
-                             "Unable to parse adjustment value");
-  return *SLMAV;
+                                 ": missing LMA offset");
+
+  auto LMAValue = getAsInteger<int64_t>(StringValue);
+  if (!LMAValue)
+    return createStringError(LMAValue.getError(),
+                             "bad format for " + OptionName + ": value after " +
+                                 ArgValue.slice(0, 2) + " is " + StringValue +
+                                 " when it should be integer");
+  return *LMAValue;
 }
 
 // parseObjcopyOptions returns the config and sets the input arguments. If a
@@ -860,12 +865,12 @@ objcopy::parseObjcopyOptions(ArrayRef<const char *> RawArgsArr,
     Config.PadTo = *Addr;
   }
 
-  if (const auto *Arg = InputArgs.getLastArg(OBJCOPY_adjust_section_lma)) {
-    Expected<int64_t> SLMAV =
-        parseAdjustSectionLMA(Arg->getValue(), Arg->getSpelling());
-    if (!SLMAV)
-      return SLMAV.takeError();
-    Config.ChangeSectionLMAValAll = *SLMAV;
+  if (const auto *Arg = InputArgs.getLastArg(OBJCOPY_change_section_lma)) {
+    Expected<int64_t> LMAValue =
+        parseChangeSectionLMA(Arg->getValue(), Arg->getSpelling());
+    if (!LMAValue)
+      return LMAValue.takeError();
+    Config.ChangeSectionLMAValAll = *LMAValue;
   }
 
   for (auto *Arg : InputArgs.filtered(OBJCOPY_redefine_symbol)) {

llvm/test/tools/llvm-objcopy/ELF/change-section-lma.test

@@ -254,12 +254,8 @@ def adjust_start : JoinedOrSeparate<["--"], "adjust-start">,
                    Alias<change_start>,
                    HelpText<"Alias for --change-start">;
 
-defm adjust_section_lma
-    : Eq<"change-section-lma",
-         "Change LMA of all sections by <val>. LMA is the physical address where the section is "
-         "loaded in the memory; as opoosed to VMA, which is the virtual address at which the "
-         "section resides during program execution. Ussually LMA and VMA are the same, but some "
-         "platforms support these to be different resulting in runtime copying">,
+defm change_section_lma
+    : Eq<"change-section-lma", "Shift LMA of non-zero-sized sections in the program header by <val>">,
       MetaVarName<"*{+|-}val">;
 
 defm add_symbol