Fix inconsistencies in ELF anonymous mappings

* A segment with filesz == 0, memsz > 0 should be an anonymous only mapping. We were failing to load such an ELF. * Anonymous pages are always mapped RW, regardless of the segment protections. PiperOrigin-RevId: 249355239 Change-Id: I251e5c0ce8848cf8420c3aadf337b0d77b1ad991
2026-05-22 17:12:49 -07:00 · 2019-05-21 17:06:05 -07:00
parent 2ac0aeeb42
commit c8857f7269
2 changed files with 197 additions and 56 deletions
@@ -223,13 +223,8 @@ func parseHeader(ctx context.Context, f *fs.File) (elfInfo, error) {
 // mapSegment maps a phdr into the Task. offset is the offset to apply to
 // phdr.Vaddr.
 func mapSegment(ctx context.Context, m *mm.MemoryManager, f *fs.File, phdr *elf.ProgHeader, offset usermem.Addr) error {
-	// Alignment of vaddr and offset must match. We'll need to map on the
-	// page boundary.
+	// We must make a page-aligned mapping.
 	adjust := usermem.Addr(phdr.Vaddr).PageOffset()
-	if adjust != usermem.Addr(phdr.Off).PageOffset() {
-		ctx.Infof("Alignment of vaddr %#x != off %#x", phdr.Vaddr, phdr.Off)
-		return syserror.ENOEXEC
-	}

 	addr, ok := offset.AddLength(phdr.Vaddr)
 	if !ok {
@@ -239,17 +234,11 @@ func mapSegment(ctx context.Context, m *mm.MemoryManager, f *fs.File, phdr *elf.
 	}
 	addr -= usermem.Addr(adjust)

-	fileOffset := phdr.Off - adjust
 	fileSize := phdr.Filesz + adjust
 	if fileSize < phdr.Filesz {
 		ctx.Infof("Computed segment file size overflows: %#x + %#x", phdr.Filesz, adjust)
 		return syserror.ENOEXEC
 	}
-	memSize := phdr.Memsz + adjust
-	if memSize < phdr.Memsz {
-		ctx.Infof("Computed segment mem size overflows: %#x + %#x", phdr.Memsz, adjust)
-		return syserror.ENOEXEC
-	}
 	ms, ok := usermem.Addr(fileSize).RoundUp()
 	if !ok {
 		ctx.Infof("fileSize %#x too large", fileSize)
@@ -257,51 +246,64 @@ func mapSegment(ctx context.Context, m *mm.MemoryManager, f *fs.File, phdr *elf.
 	}
 	mapSize := uint64(ms)

-	prot := progFlagsAsPerms(phdr.Flags)
-	mopts := memmap.MMapOpts{
-		Length: mapSize,
-		Offset: fileOffset,
-		Addr:   addr,
-		Fixed:  true,
-		// Linux will happily allow conflicting segments to map over
-		// one another.
-		Unmap:    true,
-		Private:  true,
-		Perms:    prot,
-		MaxPerms: usermem.AnyAccess,
-	}
-	defer func() {
-		if mopts.MappingIdentity != nil {
-			mopts.MappingIdentity.DecRef()
-		}
-	}()
-	if err := f.ConfigureMMap(ctx, &mopts); err != nil {
-		ctx.Infof("File is not memory-mappable: %v", err)
-		return err
-	}
-	if _, err := m.MMap(ctx, mopts); err != nil {
-		ctx.Infof("Error mapping PT_LOAD segment %+v at %#x: %v", phdr, addr, err)
-		return err
-	}
+	if mapSize > 0 {
+		// This must result in a page-aligned offset. i.e., the original
+		// phdr.Off must have the same alignment as phdr.Vaddr. If that is not
+		// true, MMap will reject the mapping.
+		fileOffset := phdr.Off - adjust

-	// We need to clear the end of the last page that exceeds fileSize so
-	// we don't map part of the file beyond fileSize.
-	//
-	// Note that Linux *does not* clear the portion of the first page
-	// before phdr.Off.
-	if mapSize > fileSize {
-		zeroAddr, ok := addr.AddLength(fileSize)
-		if !ok {
-			panic(fmt.Sprintf("successfully mmaped address overflows? %#x + %#x", addr, fileSize))
+		prot := progFlagsAsPerms(phdr.Flags)
+		mopts := memmap.MMapOpts{
+			Length: mapSize,
+			Offset: fileOffset,
+			Addr:   addr,
+			Fixed:  true,
+			// Linux will happily allow conflicting segments to map over
+			// one another.
+			Unmap:    true,
+			Private:  true,
+			Perms:    prot,
+			MaxPerms: usermem.AnyAccess,
 		}
-		zeroSize := int64(mapSize - fileSize)
-		if zeroSize < 0 {
-			panic(fmt.Sprintf("zeroSize too big? %#x", uint64(zeroSize)))
-		}
-		if _, err := m.ZeroOut(ctx, zeroAddr, zeroSize, usermem.IOOpts{IgnorePermissions: true}); err != nil {
-			ctx.Warningf("Failed to zero end of page [%#x, %#x): %v", zeroAddr, zeroAddr+usermem.Addr(zeroSize), err)
+		defer func() {
+			if mopts.MappingIdentity != nil {
+				mopts.MappingIdentity.DecRef()
+			}
+		}()
+		if err := f.ConfigureMMap(ctx, &mopts); err != nil {
+			ctx.Infof("File is not memory-mappable: %v", err)
 			return err
 		}
+		if _, err := m.MMap(ctx, mopts); err != nil {
+			ctx.Infof("Error mapping PT_LOAD segment %+v at %#x: %v", phdr, addr, err)
+			return err
+		}
+
+		// We need to clear the end of the last page that exceeds fileSize so
+		// we don't map part of the file beyond fileSize.
+		//
+		// Note that Linux *does not* clear the portion of the first page
+		// before phdr.Off.
+		if mapSize > fileSize {
+			zeroAddr, ok := addr.AddLength(fileSize)
+			if !ok {
+				panic(fmt.Sprintf("successfully mmaped address overflows? %#x + %#x", addr, fileSize))
+			}
+			zeroSize := int64(mapSize - fileSize)
+			if zeroSize < 0 {
+				panic(fmt.Sprintf("zeroSize too big? %#x", uint64(zeroSize)))
+			}
+			if _, err := m.ZeroOut(ctx, zeroAddr, zeroSize, usermem.IOOpts{IgnorePermissions: true}); err != nil {
+				ctx.Warningf("Failed to zero end of page [%#x, %#x): %v", zeroAddr, zeroAddr+usermem.Addr(zeroSize), err)
+				return err
+			}
+		}
+	}
+
+	memSize := phdr.Memsz + adjust
+	if memSize < phdr.Memsz {
+		ctx.Infof("Computed segment mem size overflows: %#x + %#x", phdr.Memsz, adjust)
+		return syserror.ENOEXEC
 	}

 	// Allocate more anonymous pages if necessary.
@@ -321,9 +323,13 @@ func mapSegment(ctx context.Context, m *mm.MemoryManager, f *fs.File, phdr *elf.
 			Addr:   anonAddr,
 			// Fixed without Unmap will fail the mmap if something is
 			// already at addr.
-			Fixed:    true,
-			Private:  true,
-			Perms:    progFlagsAsPerms(phdr.Flags),
+			Fixed:   true,
+			Private: true,
+			// N.B. Linux uses vm_brk to map these pages, ignoring
+			// the segment protections, instead always mapping RW.
+			// These pages are not included in the final brk
+			// region.
+			Perms:    usermem.ReadWrite,
 			MaxPerms: usermem.AnyAccess,
 		}); err != nil {
 			ctx.Infof("Error mapping PT_LOAD segment %v anonymous memory: %v", phdr, err)
@@ -401,6 +401,141 @@ TEST(ElfTest, DataSegment) {
             })));
 }

+// Additonal pages beyond filesz are always RW.
+//
+// N.B. Linux uses set_brk -> vm_brk to additional pages beyond filesz (even
+// though start_brk itself will always be beyond memsz). As a result, the
+// segment permissions don't apply; the mapping is always RW.
+TEST(ElfTest, ExtraMemPages) {
+  ElfBinary<64> elf = StandardElf();
+
+  // Create a standard ELF, but extend to 1.5 pages. The second page will be the
+  // beginning of a multi-page data + bss segment.
+  elf.data.resize(kPageSize + kPageSize / 2);
+
+  decltype(elf)::ElfPhdr phdr = {};
+  phdr.p_type = PT_LOAD;
+  // RWX segment. The extra anon page will be RW anyways.
+  //
+  // N.B. Linux uses clear_user to clear the end of the file-mapped page, which
+  // respects the mapping protections. Thus if we map this RO with memsz >
+  // (unaligned) filesz, then execve will fail with EFAULT. See padzero(elf_bss)
+  // in fs/binfmt_elf.c:load_elf_binary.
+  //
+  // N.N.B.B. The above only applies to the last segment. For earlier segments,
+  // the clear_user error is ignored.
+  phdr.p_flags = PF_R | PF_W | PF_X;
+  phdr.p_offset = kPageSize;
+  phdr.p_vaddr = 0x41000;
+  phdr.p_filesz = kPageSize / 2;
+  // The header is going to push vaddr up by a few hundred bytes. Keep p_memsz a
+  // bit less than 2 pages so this mapping doesn't extend beyond 0x43000.
+  phdr.p_memsz = 2 * kPageSize - kPageSize / 2;
+  elf.phdrs.push_back(phdr);
+
+  elf.UpdateOffsets();
+
+  TempPath file = ASSERT_NO_ERRNO_AND_VALUE(CreateElfWith(elf));
+
+  pid_t child;
+  int execve_errno;
+  auto cleanup = ASSERT_NO_ERRNO_AND_VALUE(
+      ForkAndExec(file.path(), {file.path()}, {}, &child, &execve_errno));
+  ASSERT_EQ(execve_errno, 0);
+
+  ASSERT_NO_ERRNO(WaitStopped(child));
+
+  EXPECT_THAT(child,
+              ContainsMappings(std::vector<ProcMapsEntry>({
+                  // text page.
+                  {0x40000, 0x41000, true, false, true, true, 0, 0, 0, 0,
+                   file.path().c_str()},
+                  // data + bss page from file.
+                  {0x41000, 0x42000, true, true, true, true, kPageSize, 0, 0, 0,
+                   file.path().c_str()},
+                  // extra page from anon.
+                  {0x42000, 0x43000, true, true, false, true, 0, 0, 0, 0, ""},
+              })));
+}
+
+// An aligned segment with filesz == 0, memsz > 0 is anon-only.
+TEST(ElfTest, AnonOnlySegment) {
+  ElfBinary<64> elf = StandardElf();
+
+  decltype(elf)::ElfPhdr phdr = {};
+  phdr.p_type = PT_LOAD;
+  // RO segment. The extra anon page will be RW anyways.
+  phdr.p_flags = PF_R;
+  phdr.p_offset = 0;
+  phdr.p_vaddr = 0x41000;
+  phdr.p_filesz = 0;
+  phdr.p_memsz = kPageSize - 0xe8;
+  elf.phdrs.push_back(phdr);
+
+  elf.UpdateOffsets();
+
+  // UpdateOffsets adjusts p_vaddr and p_offset by the header size, but we need
+  // a page-aligned p_vaddr to get a truly anon-only page.
+  elf.phdrs[2].p_vaddr = 0x41000;
+  // N.B. p_offset is now unaligned, but Linux doesn't care since this is
+  // anon-only.
+
+  TempPath file = ASSERT_NO_ERRNO_AND_VALUE(CreateElfWith(elf));
+
+  pid_t child;
+  int execve_errno;
+  auto cleanup = ASSERT_NO_ERRNO_AND_VALUE(
+      ForkAndExec(file.path(), {file.path()}, {}, &child, &execve_errno));
+  ASSERT_EQ(execve_errno, 0);
+
+  ASSERT_NO_ERRNO(WaitStopped(child));
+
+  EXPECT_THAT(child,
+              ContainsMappings(std::vector<ProcMapsEntry>({
+                  // text page.
+                  {0x40000, 0x41000, true, false, true, true, 0, 0, 0, 0,
+                   file.path().c_str()},
+                  // anon page.
+                  {0x41000, 0x42000, true, true, false, true, 0, 0, 0, 0, ""},
+              })));
+}
+
+// p_offset must have the same alignment as p_vaddr.
+TEST(ElfTest, UnalignedOffset) {
+  ElfBinary<64> elf = StandardElf();
+
+  // Unaligned offset.
+  elf.phdrs[1].p_offset += 1;
+
+  elf.UpdateOffsets();
+
+  TempPath file = ASSERT_NO_ERRNO_AND_VALUE(CreateElfWith(elf));
+
+  pid_t child;
+  int execve_errno;
+  auto cleanup = ASSERT_NO_ERRNO_AND_VALUE(
+      ForkAndExec(file.path(), {file.path()}, {}, &child, &execve_errno));
+
+  // execve(2) return EINVAL, but behavior varies between Linux and gVisor.
+  //
+  // On Linux, the new mm is committed before attempting to map into it. By the
+  // time we hit EINVAL in the segment mmap, the old mm is gone. Linux returns
+  // to an empty mm, which immediately segfaults.
+  //
+  // OTOH, gVisor maps into the new mm before committing it. Thus when it hits
+  // failure, the caller is still intact to receive the error.
+  if (IsRunningOnGvisor()) {
+    ASSERT_EQ(execve_errno, EINVAL);
+  } else {
+    ASSERT_EQ(execve_errno, 0);
+
+    int status;
+    ASSERT_THAT(RetryEINTR(waitpid)(child, &status, 0),
+                SyscallSucceedsWithValue(child));
+    EXPECT_TRUE(WIFSIGNALED(status) && WTERMSIG(status) == SIGSEGV) << status;
+  }
+}
+
 // Linux will allow PT_LOAD segments to overlap.
 TEST(ElfTest, DirectlyOverlappingSegments) {
  // NOTE(b/37289926): see PIEOutOfOrderSegments.