From: Paul Pluzhnikov <ppluzhnikov@google.com>
To: Ulrich Weigand <uweigand@de.ibm.com>
Cc: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>,
gdb-patches ml <gdb-patches@sourceware.org>,
Tom Tromey <tromey@redhat.com>,
Jan Kratochvil <jan.kratochvil@redhat.com>
Subject: Re: [patch] Speed up find_pc_section
Date: Wed, 26 Aug 2009 07:21:00 -0000 [thread overview]
Message-ID: <8ac60eac0908260020l4200cf84v2686a76b5858d13@mail.gmail.com> (raw)
In-Reply-To: <8ac60eac0908231548x135edf2doa04fa59a49455bcd@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 1622 bytes --]
On Sun, Aug 23, 2009 at 3:48 PM, Paul Pluzhnikov<ppluzhnikov@google.com> wrote:
> On Fri, Aug 21, 2009 at 4:30 AM, Ulrich Weigand <uweigand@de.ibm.com> wrote:
...
>> I'm still not completely happy about the assertions you added. An
>> assertion failure is supposed to be an indication of a bug in GDB
>> -- it should never be possible to trigger the assertion just by
>> providing particular user input (this includes the binary file).
>
> I am still working on fixing that part...
Here is the patch. Tested on Linux/x86_64 and on Fedora11/i686 with
prelinking. No regressions.
It also passes Jan's new gdb.base/solib-overlap.exp test case for RH
Bug 515434, issuing these warnings:
warning: Unexpected overlap between section `.note.gnu.build-id' from
`/home/paul/gdb-cvs/build/gdb/testsuite/gdb.base/solib-overlap-lib1-0x50000000.so'
[0xf4, 0x118) and section `.note.gnu.build-id' from
`/home/paul/gdb-cvs/build/gdb/testsuite/gdb.base/solib-overlap-lib2-0x50000000.so'
[0xf4, 0x118)
warning: Unexpected overlap between section `.gnu.hash' from
`/home/paul/gdb-cvs/build/gdb/testsuite/gdb.base/solib-overlap-lib1-0x50000000.so'
[0x118, 0x158) and section `.gnu.hash' from
`/home/paul/gdb-cvs/build/gdb/testsuite/gdb.base/solib-overlap-lib2-0x50000000.so'
[0x118, 0x158)
... etc ... etc ...
Thanks,
--
Paul Pluzhnikov
2009-08-25 Paul Pluzhnikov <ppluzhnikov@google.com>
* objfiles.c (qsort_cmp): Remove asserts.
(insert_section_p, filter_debuginfo_sections): New function.
(filter_overlapping_sections): Likewise.
(update_section_map): Adjust.
[-- Attachment #2: gdb-find_pc_section-20090825.txt --]
[-- Type: text/plain, Size: 6857 bytes --]
Index: objfiles.c
===================================================================
RCS file: /cvs/src/src/gdb/objfiles.c,v
retrieving revision 1.93
diff -p -u -r1.93 objfiles.c
--- objfiles.c 21 Aug 2009 17:57:17 -0000 1.93
+++ objfiles.c 26 Aug 2009 07:07:04 -0000
@@ -790,15 +790,9 @@ qsort_cmp (const void *a, const void *b)
const CORE_ADDR sect2_addr = obj_section_addr (sect2);
if (sect1_addr < sect2_addr)
- {
- gdb_assert (obj_section_endaddr (sect1) <= sect2_addr);
- return -1;
- }
+ return -1;
else if (sect1_addr > sect2_addr)
- {
- gdb_assert (sect1_addr >= obj_section_endaddr (sect2));
- return 1;
- }
+ return 1;
return 0;
}
@@ -823,12 +817,133 @@ preferred_obj_section (struct obj_sectio
return b;
}
+/* Return 1 if SECTION should be inserted into the section map.
+ We want to insert only non-overlay and non-TLS section. */
+
+static int
+insert_section_p (const struct bfd *abfd,
+ const struct bfd_section *section)
+{
+ const bfd_vma lma = bfd_section_lma (abfd, section);
+
+ if (lma != 0 && lma != bfd_section_vma (abfd, section)
+ && (bfd_get_file_flags (abfd) & BFD_IN_MEMORY) == 0)
+ /* This is an overlay section. IN_MEMORY check is needed to avoid
+ discarding sections from the "system supplied DSO" (aka vdso)
+ on Linux. */
+ return 0;
+ if ((bfd_get_section_flags (abfd, section) & SEC_THREAD_LOCAL) != 0)
+ /* This is a TLS section. */
+ return 0;
+
+ return 1;
+}
+
+/* Filter out overlapping sections where one section came from the real
+ objfile, and the other from a separate debuginfo file.
+ Return the size of table after redundant sections have been eliminated. */
+
+static int
+filter_debuginfo_sections (struct obj_section **map, int map_size)
+{
+ int i, j;
+
+ for (i = 0, j = 0; i < map_size - 1; i++)
+ {
+ struct obj_section *const sect1 = map[i];
+ struct obj_section *const sect2 = map[i + 1];
+ const struct objfile *const objfile1 = sect1->objfile;
+ const struct objfile *const objfile2 = sect2->objfile;
+ const CORE_ADDR sect1_addr = obj_section_addr (sect1);
+ const CORE_ADDR sect2_addr = obj_section_addr (sect2);
+
+ if (sect1_addr == sect2_addr
+ && (objfile1->separate_debug_objfile == objfile2
+ || objfile2->separate_debug_objfile == objfile1))
+ {
+ map[j++] = preferred_obj_section (sect1, sect2);
+ ++i;
+ }
+ else
+ map[j++] = sect1;
+ }
+
+ if (i < map_size)
+ map[j++] = map[i];
+
+ /* The map should not have shrunk to less than half the original size. */
+ gdb_assert (map_size / 2 <= j);
+
+ return j;
+}
+
+/* Filter out overlapping sections, issuing a warning if any are found.
+ Overlapping sections could really be overlay sections which we didn't
+ classify as such in insert_section_p, or we could be dealing with a
+ corrupt binary. */
+
+static int
+filter_overlapping_sections (struct obj_section **map, int map_size)
+{
+ int i, j;
+
+ for (i = 0, j = 0; i < map_size - 1; )
+ {
+ int k;
+
+ map[j++] = map[i];
+ for (k = i + 1; k < map_size; k++)
+ {
+ struct obj_section *const sect1 = map[i];
+ struct obj_section *const sect2 = map[k];
+ const CORE_ADDR sect1_addr = obj_section_addr (sect1);
+ const CORE_ADDR sect2_addr = obj_section_addr (sect2);
+ const CORE_ADDR sect1_endaddr = obj_section_endaddr (sect1);
+
+ gdb_assert (sect1_addr <= sect2_addr);
+
+ if (sect1_endaddr <= sect2_addr)
+ break;
+ else
+ {
+ /* We have an overlap. Report it. */
+
+ struct objfile *const objf1 = sect1->objfile;
+ struct objfile *const objf2 = sect2->objfile;
+
+ const struct bfd *const abfd1 = objf1->obfd;
+ const struct bfd *const abfd2 = objf2->obfd;
+
+ const struct bfd_section *const bfds1 = sect1->the_bfd_section;
+ const struct bfd_section *const bfds2 = sect2->the_bfd_section;
+
+ const CORE_ADDR sect2_endaddr = obj_section_endaddr (sect2);
+
+ struct gdbarch *const gdbarch = get_objfile_arch (objf1);
+
+ warning (_("Unexpected overlap between "
+ "section `%s' from `%s' [%s, %s) and "
+ "section `%s' from `%s' [%s, %s)"),
+ bfd_section_name (abfd1, bfds1), objf1->name,
+ paddress (gdbarch, sect1_addr),
+ paddress (gdbarch, sect1_endaddr),
+ bfd_section_name (abfd2, bfds2), objf2->name,
+ paddress (gdbarch, sect2_addr),
+ paddress (gdbarch, sect2_endaddr));
+ }
+ }
+ i = k;
+ }
+ return map_size;
+}
+
+
/* Update PMAP, PMAP_SIZE with non-TLS sections from all objfiles. */
static void
update_section_map (struct obj_section ***pmap, int *pmap_size)
{
- int map_size, i, j;
+ int alloc_size, map_size, i;
struct obj_section *s, **map;
struct objfile *objfile;
@@ -837,55 +952,27 @@ update_section_map (struct obj_section *
map = *pmap;
xfree (map);
-#define insert_p(objf, sec) \
- ((bfd_get_section_flags ((objf)->obfd, (sec)->the_bfd_section) \
- & SEC_THREAD_LOCAL) == 0)
-
- map_size = 0;
+ alloc_size = 0;
ALL_OBJSECTIONS (objfile, s)
- if (insert_p (objfile, s))
- map_size += 1;
+ if (insert_section_p (objfile->obfd, s->the_bfd_section))
+ alloc_size += 1;
- map = xmalloc (map_size * sizeof (*map));
+ map = xmalloc (alloc_size * sizeof (*map));
i = 0;
ALL_OBJSECTIONS (objfile, s)
- if (insert_p (objfile, s))
+ if (insert_section_p (objfile->obfd, s->the_bfd_section))
map[i++] = s;
-#undef insert_p
-
- qsort (map, map_size, sizeof (*map), qsort_cmp);
-
- /* With separate debuginfo files, we may have up to two (almost)
- identical copies of some obj_sections in the map.
- Filter out duplicates. */
- for (i = 0, j = 0; i < map_size; ++i)
- {
- struct obj_section *sect1 = map[i];
- struct obj_section *sect2 = (i + 1 < map_size) ? map[i + 1] : NULL;
-
- if (sect2 == NULL
- || obj_section_addr (sect1) != obj_section_addr (sect2))
- map[j++] = sect1;
- else
- {
- map[j++] = preferred_obj_section (sect1, sect2);
- ++i;
- }
- }
-
- if (j < map_size)
- {
- /* Some duplicates were eliminated.
- The new size shouldn't be less than half of the original. */
- gdb_assert (map_size / 2 <= j);
- map_size = j;
-
- map = xrealloc (map, map_size * sizeof (*map)); /* Trim excess space. */
- }
+ qsort (map, alloc_size, sizeof (*map), qsort_cmp);
+ map_size = filter_debuginfo_sections(map, alloc_size);
+ map_size = filter_overlapping_sections(map, map_size);
+
+ if (map_size < alloc_size)
+ /* Some sections were eliminated. Trim excess space. */
+ map = xrealloc (map, map_size * sizeof (*map));
else
- gdb_assert (j == map_size);
+ gdb_assert (alloc_size == map_size);
*pmap = map;
*pmap_size = map_size;
next prev parent reply other threads:[~2009-08-26 7:20 UTC|newest]
Thread overview: 82+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-07-17 7:34 Paul Pluzhnikov
2009-07-17 15:59 ` Paul Pluzhnikov
2009-07-17 16:27 ` Tom Tromey
2009-07-17 17:19 ` Paul Pluzhnikov
2009-07-21 17:57 ` Tom Tromey
2009-07-21 20:51 ` Paul Pluzhnikov
2009-07-21 21:03 ` Tom Tromey
2009-07-22 15:23 ` Pedro Alves
2009-07-22 16:33 ` Tom Tromey
2009-07-22 17:02 ` Paul Pluzhnikov
2009-07-22 17:02 ` Pedro Alves
2009-07-22 17:16 ` Paul Pluzhnikov
2009-07-22 18:08 ` Paul Pluzhnikov
2009-07-22 18:10 ` Pedro Alves
2009-07-22 18:12 ` Paul Pluzhnikov
2009-07-22 19:19 ` Pedro Alves
2009-07-22 19:34 ` Paul Pluzhnikov
2009-07-22 19:54 ` Pedro Alves
2009-08-17 21:15 ` [commit] Fix reread_symbols crash (Re: [patch] Speed up find_pc_section) Ulrich Weigand
2009-08-04 14:22 ` [patch] Speed up find_pc_section Tom Tromey
2009-08-04 15:06 ` Paul Pluzhnikov
2009-08-04 15:38 ` Tom Tromey
[not found] ` <8ac60eac0908042358q4d2061d2md3c49cf4aab26398@mail.gmail.com>
[not found] ` <m34osmi5jx.fsf@fleche.redhat.com>
[not found] ` <8ac60eac0908050940we3dc478rd182f4367a650f1b@mail.gmail.com>
[not found] ` <8ac60eac0908052259l7b1c21d1t212991886a5f8b18@mail.gmail.com>
[not found] ` <m3eirofxwh.fsf@fleche.redhat.com>
[not found] ` <8ac60eac0908070030g7500a5ack3fcc81862e2a5b0a@mail.gmail.com>
2009-08-07 23:30 ` Paul Pluzhnikov
2009-08-09 21:37 ` Paul Pluzhnikov
2009-08-10 18:09 ` Tom Tromey
2009-08-10 20:39 ` Paul Pluzhnikov
2009-08-17 19:45 ` Ulrich Weigand
2009-08-17 19:57 ` Ulrich Weigand
2009-08-17 22:55 ` Paul Pluzhnikov
2009-08-18 13:48 ` Ulrich Weigand
2009-08-20 18:03 ` Paul Pluzhnikov
2009-08-20 18:39 ` Ulrich Weigand
2009-08-20 21:06 ` Paul Pluzhnikov
2009-08-20 22:34 ` Daniel Jacobowitz
2009-08-21 12:36 ` Ulrich Weigand
2009-08-23 23:25 ` Paul Pluzhnikov
2009-08-26 7:21 ` Paul Pluzhnikov [this message]
2009-08-26 14:37 ` Jan Kratochvil
2009-08-26 14:38 ` Paul Pluzhnikov
2009-08-26 15:17 ` Paul Pluzhnikov
2009-08-26 23:45 ` Jan Kratochvil
2009-08-27 2:56 ` Paul Pluzhnikov
2009-09-02 17:02 ` Paul Pluzhnikov
2009-09-08 18:37 ` What should we do re: "[patch] Speed up find_pc_section" Joel Brobecker
2009-09-08 20:16 ` Paul Pluzhnikov
2009-09-08 21:17 ` Joel Brobecker
2009-09-09 5:58 ` [patch] Speed up find_pc_section Joel Brobecker
2009-09-09 7:56 ` Tristan Gingold
2009-09-09 15:04 ` Joel Brobecker
2009-09-11 7:44 ` Tristan Gingold
2009-09-10 17:36 ` Paul Pluzhnikov
2009-09-10 18:30 ` Joel Brobecker
2009-09-11 1:30 ` Paul Pluzhnikov
2009-09-11 6:51 ` Pierre Muller
2009-09-11 7:29 ` Paul Pluzhnikov
2009-09-11 7:40 ` Mark Kettenis
2009-09-11 7:51 ` Paul Pluzhnikov
2009-09-11 7:41 ` Pierre Muller
2009-09-11 8:03 ` Paul Pluzhnikov
2009-09-11 8:41 ` Pierre Muller
2009-09-11 17:47 ` Paul Pluzhnikov
2009-09-11 21:15 ` Joel Brobecker
2009-09-13 21:47 ` Paul Pluzhnikov
2009-09-14 16:43 ` Ulrich Weigand
2009-09-14 17:19 ` Paul Pluzhnikov
2009-09-14 17:36 ` Joel Brobecker
2009-09-14 18:10 ` Paul Pluzhnikov
2009-09-14 18:21 ` Joel Brobecker
2009-09-11 20:51 ` Tom Tromey
2009-09-11 21:04 ` Paul Pluzhnikov
2009-09-11 21:14 ` Tom Tromey
2009-09-11 7:53 ` Tristan Gingold
2009-09-11 8:33 ` Paul Pluzhnikov
2009-09-11 8:39 ` Tristan Gingold
2009-09-11 16:23 ` Paul Pluzhnikov
2009-09-09 5:39 ` Joel Brobecker
2009-09-10 16:18 ` Paul Pluzhnikov
2009-09-11 21:06 ` Joel Brobecker
2009-09-14 16:41 ` Ulrich Weigand
2009-08-18 18:18 ` Michael Snyder
2009-07-17 18:56 ` Paul Pluzhnikov
2009-07-21 3:34 ` Paul Pluzhnikov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8ac60eac0908260020l4200cf84v2686a76b5858d13@mail.gmail.com \
--to=ppluzhnikov@google.com \
--cc=Ulrich.Weigand@de.ibm.com \
--cc=gdb-patches@sourceware.org \
--cc=jan.kratochvil@redhat.com \
--cc=tromey@redhat.com \
--cc=uweigand@de.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox