[PATCH] D97661: [llvm-objcopy/strip] Fix off-by-one error in SYMTAB_SHNDX need check

James Henderson via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Mar 1 01:04:21 PST 2021


jhenderson created this revision.
jhenderson added reviewers: rupprecht, MaskRay, grimar, alexshap.
Herald added subscribers: abrachet, emaste.
jhenderson requested review of this revision.
Herald added a project: LLVM.

The check for whether an extended symbol index table was required dropped the first SHN_LORESERVE sections from the sections array before checking whether the remaining sections had symbols. Unfortunately, the null section header is not present in this list, so the check was skipping the first section that might be important. If that section contained a symbol, and no subsequent ones did, the .symtab_shndx section would not be emitted, leading to a corrupt object.

Also consolidate and expand test coverage in the area to cover this bug and other aspects of the SYMTAB_SHNDX section.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D97661

Files:
  llvm/test/tools/llvm-objcopy/ELF/auto-remove-add-symtab-shndx.test
  llvm/test/tools/llvm-objcopy/ELF/auto-remove-shndx.test
  llvm/test/tools/llvm-objcopy/ELF/strict-no-add.test
  llvm/tools/llvm-objcopy/ELF/Object.cpp


Index: llvm/tools/llvm-objcopy/ELF/Object.cpp
===================================================================
--- llvm/tools/llvm-objcopy/ELF/Object.cpp
+++ llvm/tools/llvm-objcopy/ELF/Object.cpp
@@ -2450,8 +2450,10 @@
   bool NeedsLargeIndexes = false;
   if (Obj.sections().size() >= SHN_LORESERVE) {
     SectionTableRef Sections = Obj.sections();
+    // Sections doesn't include the null section header, so account for this
+    // when skipping the first N sections.
     NeedsLargeIndexes =
-        any_of(drop_begin(Sections, SHN_LORESERVE),
+        any_of(drop_begin(Sections, SHN_LORESERVE - 1),
                [](const SectionBase &Sec) { return Sec.HasSymbol; });
     // TODO: handle case where only one section needs the large index table but
     // only needs it because the large index table hasn't been removed yet.
Index: llvm/test/tools/llvm-objcopy/ELF/strict-no-add.test
===================================================================
--- llvm/test/tools/llvm-objcopy/ELF/strict-no-add.test
+++ /dev/null
@@ -1,10 +0,0 @@
-# This test makes sure that sections added at the end that don't have symbols
-# defined in them don't trigger the creation of a large index table.
-
-RUN: %python %p/../Inputs/ungzip.py %p/Inputs/many-sections.o.gz > %t.0
-RUN: echo 'foo' > %t
-RUN: llvm-objcopy -R 's65[3-9][0-9][0-9]' -R 's6529[0-9]' -R 's652[89][0-9]' %t.0 %t2
-RUN: llvm-objcopy --add-section=.s0=%t --add-section=.s1=%t --add-section=.s2=%t %t2 %t2
-RUN: llvm-readobj --sections %t2 | FileCheck --check-prefix=SECS %s
-
-SECS-NOT: Name: .symtab_shndx
Index: llvm/test/tools/llvm-objcopy/ELF/auto-remove-shndx.test
===================================================================
--- llvm/test/tools/llvm-objcopy/ELF/auto-remove-shndx.test
+++ /dev/null
@@ -1,5 +0,0 @@
-# RUN: %python %p/../Inputs/ungzip.py %p/Inputs/many-sections.o.gz > %t
-# RUN: llvm-objcopy -R 's65[3-9][0-9][0-9]' -R 's6529[0-9]' -R 's6528[1-9]' %t %t2
-# RUN: llvm-readobj --sections %t2 | FileCheck --check-prefix=SECS %s
-
-# SECS-NOT: Name: .symtab_shndx
Index: llvm/test/tools/llvm-objcopy/ELF/auto-remove-add-symtab-shndx.test
===================================================================
--- /dev/null
+++ llvm/test/tools/llvm-objcopy/ELF/auto-remove-add-symtab-shndx.test
@@ -0,0 +1,47 @@
+# This test makes sure that the .symtab_shndx section is automatically removed
+# when it is not needed and added when it is.
+
+RUN: %python %p/../Inputs/ungzip.py %p/Inputs/many-sections.o.gz > %t.0
+RUN: echo 'foo' > %t
+# many-sections.o contains sections from s1 to s65535, each with a symbol.
+# Remove enough so that the last section with a symbol has index
+# SHN_LORESERVE (i.e. s65280), and show the section hasn't been removed.
+RUN: llvm-objcopy -R 's65[3-9][0-9][0-9]' -R 's6529[0-9]' -R 's6528[1-9]' %t.0 %t2
+RUN: llvm-readobj --sections %t2 | FileCheck --check-prefix=SHNDX %s
+# Remove one more to cause removal of the section.
+RUN: llvm-objcopy -R s65280 %t2 %t3
+RUN: llvm-readobj --sections %t3 | FileCheck --check-prefix=NOSHNDX %s
+
+# Add additional sections. No .symtab_shndx should be added, as they contain no
+# symbols.
+RUN: llvm-objcopy --add-section=newsec1=%t --add-section=newsec2=%t --add-section=newsec3=%t %t3 %t4
+RUN: llvm-readobj --sections %t4 | FileCheck --check-prefix=NOSHNDX %s
+
+# Add a symbol to one of these sections and show that the .symtab_shndx
+# section has been created and that the new symbol has the right section index.
+RUN: llvm-objcopy --add-symbol=newsym=newsec1:0 %t4 %t5
+RUN: llvm-readobj --symbols --sections %t5 | \
+RUN:   FileCheck --check-prefixes=SHNDX,SHNDX-SYM -DSECTION=newsec1 %s
+
+# Show that removing and adding symbols, but not their sections, has the same
+# effect regarding removal/addition of the .symtab_shndx section.
+# FIXME: llvm-objcopy does not currently remove this section in these
+# circumstances. See https://bugs.llvm.org/show_bug.cgi?id=49364
+# Test is left here to document the current behaviour.
+RUN: llvm-objcopy -N 'sym65[3-9][0-9][0-9]' -N 'sym652[89][0-9]' -w %t.0 %t6
+RUN: llvm-readobj --sections %t6 | not FileCheck --check-prefix=NOSHNDX %s
+# FIXME: Remove these test lines once the bug has been fixed. They simply show
+# that the section is removed by a run without any further operations.
+RUN: llvm-objcopy %t6 %t6b
+RUN: llvm-readobj --sections %t6b | FileCheck --check-prefix=NOSHNDX %s
+
+# Add a symbol to ensure the section is reinstated.
+RUN: llvm-objcopy --add-symbol=newsym=s65280:0 -w %t6 %t7
+RUN: llvm-readobj --sections --symbols %t7 | \
+RUN:   FileCheck --check-prefixes=SHNDX,SHNDX-SYM -DSECTION=s65280 %s
+
+NOSHNDX-NOT:    Name: .symtab_shndx
+SHNDX:          Name: .symtab_shndx
+SHNDX-SYM:      Name: newsym
+SHNDX-SYM:      Section:
+SHNDX-SYM-SAME:           [[SECTION]]


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D97661.327040.patch
Type: text/x-patch
Size: 4817 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210301/1e3eed12/attachment.bin>


More information about the llvm-commits mailing list