[llvm-objdump] Emit warning if --start-address/--stop-address specify range outside file's address range.

[ychen]

Summary

NB: the warning is about the input file itself regardless of the options used
such as -r, -s etc..

Only sections with the SHF_ALLOC flag are considered.

https://bugs.llvm.org/show_bug.cgi?id=41911

[MaskRay]

llvm/tools/llvm-objdump/llvm-objdump.cpp

2016 ↗

(On Diff #209984)

Why don't ET_CORE and ET_REL need the warning?

2043 ↗

(On Diff #209984)

covers

Start < BaseAddr + Size) && Stop > BaseAddr checks whether two ranges overlap.

2061 ↗

(On Diff #209984)

I suggest prefixing the function with maybeWarn (or check) because it doesn't always warn.

[ychen]

• update

[ychen]

llvm/tools/llvm-objdump/llvm-objdump.cpp

2016 ↗

(On Diff #209984)

Section addresses in ET_REL files are not VMA, probably may not be very useful. ET_CORE could be here, I'm not sure its potential usefulness either. What's your opinion?

[MaskRay]

llvm/tools/llvm-objdump/llvm-objdump.cpp

2016 ↗

(On Diff #209984)

Ah, I agree that --start-address should not on ET_REL. ET_CORE doesn't have a section header table (it is not covered by a PT_LOAD).

I think either the current condition or return Elf->getEType() != ELF::ET_REL; is fine.

I have a slight preference for !=ET_REL because it (1) makes code a bit simpler (2) there is no fundamental problem with ET_CORE (3) will work if someone attaches the section header to the core (though we don't need to test ET_CORE for now).

[ychen]

• Use !=ET_REL

[ychen]

• update

[ychen]

ping?

[MaskRay]

LGMT but please wait for other opinions

llvm/tools/llvm-objdump/llvm-objdump.cpp

2016 ↗

(On Diff #209984)

You can write this as return Elf->getEType() != ELF::ET_REL; and delete the surrounding {}.

[jhenderson]

LGTM.

llvm/trunk/test/tools/llvm-objdump/warn-on-out-of-range-start-stop-address.test

+## This test checks warning messages if --start-address/--stop-address
+## do not intersect with address ranges of sections that have the SHF_ALLOC
+## flag.
+
+# RUN: yaml2obj --docnum=1 %s > %t
+# RUN: yaml2obj --docnum=2 %s > %t.2
+# RUN: yaml2obj --docnum=3 %s > %t.o
+# RUN: yaml2obj --docnum=4 %s > %t.3
+
+## Warn if no section covers any part of the specified range.
+
+## - Section ends at start of range:
+##           | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1004 --stop-address=0x1006 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## - Range is between two sections:
+##             | range |
+## | section |           | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1005 --stop-address=0x1006 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## - Range appears after any section:
+##                  | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1405 --stop-address=0x1406 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## - Range starts at 0. (--start-address defaults to 0).
+# RUN: llvm-objdump --file-headers --stop-address=0x1000 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN-STOP-ONLY
+
+## - Range ends at UINT64_MAX. (--stop-address defaults to UINT64_MAX)
+# RUN: llvm-objdump --file-headers --start-address=0x1500 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN-START-ONLY
+
+## No warning if a section covers at least part of the specified range.
+
+## - Ranges are identical:
+## | range   |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1000 --stop-address=0x1004 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Range is entirely within section:
+##  | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1001 --stop-address=0x1003 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section is entirely within range:
+## |   range   |
+##  | section |
+# RUN: llvm-objdump --file-headers --start-address=0xfff --stop-address=0x1005 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range share same start, section larger:
+## | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1000 --stop-address=0x1003 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range share same start, range larger:
+## | range        |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1000 --stop-address=0x1005 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range share same end, section larger:
+##   | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1001 --stop-address=0x1004 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range share same end, range larger:
+## | range    |
+##  | section |
+# RUN: llvm-objdump --file-headers --start-address=0xfff --stop-address=0x1004 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range partially overlap, range first:
+## | range |
+##   | section |
+# RUN: llvm-objdump --file-headers --start-address=0xfff --stop-address=0x1003 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Section and range partially overlap, section first:
+##       | range |
+## | section |
+# RUN: llvm-objdump --file-headers --start-address=0x1001 --stop-address=0x1005 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## - Range starts before first section and ends after second:
+## |            range               |
+##   | section |    | section |
+# RUN: llvm-objdump --file-headers --start-address=0xfff --stop-address=0x1405 %t 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## Warn only for the input file that does not have the specified range.
+# RUN: llvm-objdump --file-headers --start-address=0x2001 --stop-address=0x2005 %t %t.2 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=MULTI-INPUT
+
+## Warn if the specified range is in a segment but not in any section.
+# RUN: llvm-objdump --file-headers --start-address=0x1008 --stop-address=0x1009 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## Warning for --start-address/--stop-address works regardless of the other options used including --section.
+# RUN: llvm-objdump --syms --section=.text2 --start-address=0x1004 --stop-address=0x1005 %t 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## Sections without the SHF_ALLOC flag are ignored in address range calculation.
+# RUN: llvm-objdump --file-headers --start-address=0x1 --stop-address=0x3 %t.3 2>&1 \
+# RUN:   | FileCheck %s --check-prefix=WARN
+
+## No warning for relocatable objects.
+# RUN: llvm-objdump --file-headers --start-address=0x1004 --stop-address=0x1005 %t.o 2>&1 \
+# RUN:   | FileCheck %s --implicit-check-not=warning:
+
+## No warning if neither --start-address nor --stop-address are specified.
+# RUN: llvm-objdump --file-headers %t 2>&1 | FileCheck %s --implicit-check-not=warning:
+
+# WARN: warning: no section overlaps the range {{.*}} specified by --start-address/--stop-address
+# WARN-STOP-ONLY: warning: no section has address less than 0x1000 specified by --stop-address
+# WARN-START-ONLY: warning: no section has address greater than or equal to 0x1500 specified by --start-address
+
+# MULTI-INPUT: file format
+# MULTI-INPUT: warning: no section overlaps the range [0x2001,0x2005) specified by --start-address/--stop-address
+# MULTI-INPUT: file format
+# MULTI-INPUT-NOT: warning:
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name:    .text
+    Type:    SHT_PROGBITS
+    Flags:   [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address: 0x1000
+    Size:    4
+  - Name:    .text2
+    Type:    SHT_PROGBITS
+    Flags:   [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address: 0x1400
+    Size:    4
+ProgramHeaders:
+  - Type:     PT_LOAD
+    Flags:    [ PF_X, PF_R ]
+    VAddr:    0x1000
+    FileSize: 0x500
+    Sections:
+      - Section: .text
+      - Section: .text2
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_DYN
+  Machine: EM_X86_64
+Sections:
+  - Name:    .text
+    Type:    SHT_PROGBITS
+    Flags:   [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address: 0x2000
+    Size:    4
+ProgramHeaders:
+  - Type:     PT_LOAD
+    Flags:    [ PF_X, PF_R ]
+    VAddr:    0x1000
+    FileSize: 0x4
+    Sections:
+      - Section: .text
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_REL
+  Machine: EM_X86_64
+Sections:
+  - Name:    .text
+    Type:    SHT_PROGBITS
+    Flags:   [ SHF_ALLOC, SHF_EXECINSTR ]
+    Address: 0x1000
+    Size:    4
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_DYN
+  Machine: EM_X86_64
+Sections:
+  - Name:    .non.alloc
+    Type:    SHT_PROGBITS
+    Size:    2

llvm/trunk/tools/llvm-objdump/llvm-objdump.cpp

if (!NameOrErr) {
     consumeError(NameOrErr.takeError());
     Name = unwrapOrError(C.getRawName(), Filename);
   } else {
     Name = NameOrErr.get();
   }
   outs() << Name << "\n";
 }
 
+// For ELF only now.
+static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) {
+  if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) {
+    if (Elf->getEType() != ELF::ET_REL)
+      return true;
+  }
+  return false;
+}
+
+static void checkForInvalidStartStopAddress(ObjectFile *Obj,
+                                            uint64_t Start, uint64_t Stop) {
+  if (!shouldWarnForInvalidStartStopAddress(Obj))
+    return;
+
+  for (const SectionRef &Section : Obj->sections())
+    if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) {
+      uint64_t BaseAddr = Section.getAddress();
+      uint64_t Size = Section.getSize();
+      if ((Start < BaseAddr + Size) && Stop > BaseAddr)
+        return;
+    }
+
+  if (StartAddress.getNumOccurrences() == 0)
+    warn("no section has address less than 0x" +
+         Twine::utohexstr(Stop) + " specified by --stop-address");
+  else if (StopAddress.getNumOccurrences() == 0)
+    warn("no section has address greater than or equal to 0x" +
+         Twine::utohexstr(Start) + " specified by --start-address");
+  else
+    warn("no section overlaps the range [0x" +
+         Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) +
+         ") specified by --start-address/--stop-address");
+}
+
 static void dumpObject(ObjectFile *O, const Archive *A = nullptr,
                        const Archive::Child *C = nullptr) {
   // Avoid other output when using a raw option.
   if (!RawClangAST) {
     outs() << '\n';
     if (A)
       outs() << A->getFileName() << "(" << O->getFileName() << ")";
     else
       outs() << O->getFileName();
     outs() << ":\tfile format " << O->getFileFormatName() << "\n\n";
   }
 
+  if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences())
+    checkForInvalidStartStopAddress(O, StartAddress, StopAddress);
+
   StringRef ArchiveName = A ? A->getFileName() : "";
   if (FileHeaders)
     printFileHeaders(O);
   if (ArchiveHeaders && !MachOOpt && C)
     printArchiveChild(ArchiveName, *C);
   if (Disassemble)
     disassembleObject(O, Relocations);
   if (Relocations && !Disassemble)