libdwelf: Make dwelf_elf_begin return NULL only when there is an error.

dwelf_elf_begin was slightly different from elf_begin in case the file
turned out to not be an ELF file. elf_begin would return an Elf handle
with ELF_K_NONE. But dwelf_elf_begin would return NULL. This made it
impossible to tell the difference between a file or decompression error
and a (decompressed) file not being an ELF file.

Since dwelf_elf_begin could still return different kinds of ELF files
(ELF_K_ELF or ELF_K_AR - and theoretically ELF_K_COFF) this was not
really useful anyway. So make it so that dwelf_elf_begin always returns
an Elf handle unless there was a real error reading or decompressing
the file. Otherwise return NULL to make clear there was a real error.

Make sure that the decompression function returns DWFL_E_BADELF only
when the file isn't compressed. In which case the Elf handle won't
be replaced and can be returned (as ELF_K_NONE).

Add a new version to dwelf_elf_begin so programs can rely on it
returning NULL only for real errors.

Signed-off-by: Mark Wielaard <mark@klomp.org>

8 files changed, 45 insertions, 13 deletions

diff --git a/libdw/ChangeLog b/libdw/ChangeLog
index 6b779e77..bf1f4857 100644
--- a/libdw/ChangeLog
+++ b/libdw/ChangeLog
@@ -1,3 +1,7 @@
+2019-08-12  Mark Wielaard  <mark@klomp.org>
+
+	* libdw.map (ELFUTILS_0.177): Add new version of dwelf_elf_begin.
+
 2019-06-28  Mark Wielaard  <mark@klomp.org>
 
 	* libdw.map (ELFUTILS_0.177): New section. Add
diff --git a/libdw/libdw.map b/libdw/libdw.map
index 2e1c0e9e..decac05c 100644
--- a/libdw/libdw.map
+++ b/libdw/libdw.map
@@ -365,4 +365,8 @@ ELFUTILS_0.175 {
 ELFUTILS_0.177 {
   global:
     dwelf_elf_e_machine_string;
+    # Replaced ELFUTILS_0.175 versions.  Both versions point to the
+    # same implementation, but users of the new symbol version can
+    # presume that NULL is only returned on error (otherwise ELF_K_NONE).
+    dwelf_elf_begin;
 } ELFUTILS_0.175;
diff --git a/libdwelf/ChangeLog b/libdwelf/ChangeLog
index 29f9a509..5b48ed8f 100644
--- a/libdwelf/ChangeLog
+++ b/libdwelf/ChangeLog
@@ -1,3 +1,9 @@
+2019-08-12  Mark Wielaard  <mark@klomp.org>
+
+	* libdwelf.h (dwelf_elf_begin): Update documentation.
+	* dwelf_elf_begin.c (dwelf_elf_begin): Don't suppress ELF_K_NONE.
+	Mark old and new version.
+
 2019-06-28  Mark Wielaard  <mark@klomp.org>
 
 	* Makefile.am (libdwelf_a_SOURCES): Add dwelf_elf_e_machine_string.c.
diff --git a/libdwelf/dwelf_elf_begin.c b/libdwelf/dwelf_elf_begin.c
index 79825338..c7d63a1c 100644
--- a/libdwelf/dwelf_elf_begin.c
+++ b/libdwelf/dwelf_elf_begin.c
@@ -41,13 +41,13 @@ dwelf_elf_begin (int fd)
 {
   Elf *elf = NULL;
   Dwfl_Error e = __libdw_open_elf (fd, &elf);
-  if (elf != NULL && elf_kind (elf) != ELF_K_NONE)
+  if (e == DWFL_E_NOERROR)
     return elf;
 
-  /* Elf wasn't usable.  Make sure there is a proper elf error message.  */
-
-  if (elf != NULL)
-    elf_end (elf);
+  /* Elf wasn't usable.  Make sure there is a proper elf error
+     message.  This is probably not the real error, because there is
+     no good way to propagate errnos or decompression errors, but
+     better than nothing.  */
 
   if (e != DWFL_E_LIBELF)
     {
@@ -60,3 +60,5 @@ dwelf_elf_begin (int fd)
 
   return NULL;
 }
+OLD_VERSION (dwelf_elf_begin, ELFUTILS_0.175)
+NEW_VERSION (dwelf_elf_begin, ELFUTILS_0.177)
diff --git a/libdwelf/libdwelf.h b/libdwelf/libdwelf.h
index cb7ea091..dbb8f08c 100644
--- a/libdwelf/libdwelf.h
+++ b/libdwelf/libdwelf.h
@@ -128,9 +128,12 @@ extern void dwelf_strtab_free (Dwelf_Strtab *st)
 /* Creates a read-only Elf handle from the given file handle.  The
    file may be compressed and/or contain a linux kernel image header,
    in which case it is eagerly decompressed in full and the Elf handle
-   is created as if created with elf_memory ().  On error NULL is
-   returned.  The Elf handle should be closed with elf_end ().  The
-   file handle will not be closed.  Does not return ELF_K_NONE handles.  */
+   is created as if created with elf_memory ().  On decompression or
+   file errors NULL is returned (and elf_errno will be set).  If there
+   was no error, but the file is not an ELF file, then an ELF_K_NONE
+   Elf handle is returned (just like with elf_begin).  The Elf handle
+   should be closed with elf_end ().  The file handle will not be
+   closed.  */
 extern Elf *dwelf_elf_begin (int fd);
 
 /* Returns a human readable string for the given ELF header e_machine
diff --git a/libdwfl/ChangeLog b/libdwfl/ChangeLog
index 8cbe90c9..04a39637 100644
--- a/libdwfl/ChangeLog
+++ b/libdwfl/ChangeLog
@@ -1,3 +1,11 @@
+2019-08-12  Mark Wielaard  <mark@klomp.org>
+
+	* gzip.c (open_stream): Return DWFL_E_ERRNO on bad file operation.
+	* open.c (libdw_open_elf): New argument bad_elf_ok. Check it and
+	return DWFL_E_NOERROR in case it is set and error was DWFL_E_BADELF.
+	(__libdw_open_file): Call libdw_open_elf with bad_elf_ok false.
+	(__libdw_open_elf): Call libdw_open_elf with bad_elf_ok true.
+
 2019-08-05  Omar Sandoval  <osandov@fb.com>
 
 	* dwfl_segment_report_module.c (dwfl_segment_report_module): Assign
diff --git a/libdwfl/gzip.c b/libdwfl/gzip.c
index c2c13baf..043d0b6e 100644
--- a/libdwfl/gzip.c
+++ b/libdwfl/gzip.c
@@ -139,14 +139,14 @@ open_stream (int fd, off_t start_offset, struct unzip_state *state)
 {
     int d = dup (fd);
     if (unlikely (d < 0))
-      return DWFL_E_BADELF;
+      return DWFL_E_ERRNO;
     if (start_offset != 0)
       {
 	off_t off = lseek (d, start_offset, SEEK_SET);
 	if (off != start_offset)
 	  {
 	    close (d);
-	    return DWFL_E_BADELF;
+	    return DWFL_E_ERRNO;
 	  }
       }
     state->zf = gzdopen (d, "r");
@@ -288,6 +288,7 @@ unzip (int fd, off_t start_offset,
   if (result == DWFL_E_NOERROR && gzdirect (state.zf))
     {
       gzclose (state.zf);
+      /* Not a compressed stream after all.  */
       return fail (&state, DWFL_E_BADELF);
     }
 
diff --git a/libdwfl/open.c b/libdwfl/open.c
index 74367359..35fc5283 100644
--- a/libdwfl/open.c
+++ b/libdwfl/open.c
@@ -118,7 +118,7 @@ what_kind (int fd, Elf **elfp, Elf_Kind *kind, bool *may_close_fd)
 
 static Dwfl_Error
 libdw_open_elf (int *fdp, Elf **elfp, bool close_on_fail, bool archive_ok,
-		bool never_close_fd)
+		bool never_close_fd, bool bad_elf_ok)
 {
   bool may_close_fd = false;
 
@@ -164,6 +164,10 @@ libdw_open_elf (int *fdp, Elf **elfp, bool close_on_fail, bool archive_ok,
       && !(archive_ok && kind == ELF_K_AR))
     error = DWFL_E_BADELF;
 
+  /* This basically means, we keep a ELF_K_NONE Elf handle and return it.  */
+  if (bad_elf_ok && error == DWFL_E_BADELF)
+    error = DWFL_E_NOERROR;
+
   if (error != DWFL_E_NOERROR)
     {
       elf_end (elf);
@@ -184,11 +188,11 @@ libdw_open_elf (int *fdp, Elf **elfp, bool close_on_fail, bool archive_ok,
 Dwfl_Error internal_function
 __libdw_open_file (int *fdp, Elf **elfp, bool close_on_fail, bool archive_ok)
 {
-  return libdw_open_elf (fdp, elfp, close_on_fail, archive_ok, false);
+  return libdw_open_elf (fdp, elfp, close_on_fail, archive_ok, false, false);
 }
 
 Dwfl_Error internal_function
 __libdw_open_elf (int fd, Elf **elfp)
 {
-  return libdw_open_elf (&fd, elfp, false, true, true);
+  return libdw_open_elf (&fd, elfp, false, true, true, true);
 }