Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Sergio Durigan Junior <sergiodj@redhat.com>
To: GDB Patches <gdb-patches@sourceware.org>
Cc: Sergio Durigan Junior <sergiodj@redhat.com>
Subject: [PATCH] Use "get_current_arch" instead of "get_objfile_arch" on SystemTap SDT code (and fix ARM bug)
Date: Thu, 05 Dec 2013 06:34:00 -0000	[thread overview]
Message-ID: <1386225226-18549-1-git-send-email-sergiodj@redhat.com> (raw)

While implementing another patch and testing things on ARM, I found a
strange bug (at least initially).  There was an expression on some
probe's argument that did a displacement of the "fp" register (i.e.,
"[fp, #-8]").  However, during the evaluation of the parsed expression
GDB got confused and could not access the "fp" register due to an
invalid address being returned.

After a good deal of time spent on the investigation of this, I found
that the culprit was the gdbarch being used; it did not provide the
correct number of pseudo-registers for the architecture, and thus the
internal number of the "fp" register ended up being completely wrong.
Then, I began to read the code, and found this on gdb/objfiles.h:struct
objfile_per_bfd_storage:

  /* The gdbarch associated with the BFD.  Note that this gdbarch is
     determined solely from BFD information, without looking at target
     information.  The gdbarch determined from a running target may
     differ from this e.g. with respect to register types and names.  */

  struct gdbarch *gdbarch;

Well, then the fix became obvious.  I replaced all the occurrences of
"get_objfile_arch" on gdb/stap-probe.c to "get_current_arch", and voilà.

This patch has been tested on both x86_64 Fedora 17 and ARM Fedora 19.

2013-12-05  Sergio Durigan Junior  <sergiodj@redhat.com>

	* stap-probe.c (stap_parse_probe_arguments): Use "get_current_arch"
	instead of "get_objfile_arch".
	(stap_can_evaluate_probe_arguments): Likewise.
	(handle_stap_probe): Likewise.
	(stap_gen_info_probes_table_values): Likewise.
---
 gdb/ChangeLog    | 8 ++++++++
 gdb/stap-probe.c | 8 ++++----
 2 files changed, 12 insertions(+), 4 deletions(-)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index f5ba7d3..ec56b57 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,11 @@
+2013-12-05  Sergio Durigan Junior  <sergiodj@redhat.com>
+
+	* stap-probe.c (stap_parse_probe_arguments): Use "get_current_arch"
+	instead of "get_objfile_arch".
+	(stap_can_evaluate_probe_arguments): Likewise.
+	(handle_stap_probe): Likewise.
+	(stap_gen_info_probes_table_values): Likewise.
+
 2013-12-05  Joel Brobecker  <brobecker@adacore.com>
 	    Tristan Gingold  <gingold@adacore.com>
 
diff --git a/gdb/stap-probe.c b/gdb/stap-probe.c
index a734793..3ab54b0 100644
--- a/gdb/stap-probe.c
+++ b/gdb/stap-probe.c
@@ -917,7 +917,7 @@ static void
 stap_parse_probe_arguments (struct stap_probe *probe)
 {
   const char *cur;
-  struct gdbarch *gdbarch = get_objfile_arch (probe->p.objfile);
+  struct gdbarch *gdbarch = get_current_arch ();
 
   gdb_assert (!probe->args_parsed);
   cur = probe->args_u.text;
@@ -1086,7 +1086,7 @@ static int
 stap_can_evaluate_probe_arguments (struct probe *probe_generic)
 {
   struct stap_probe *stap_probe = (struct stap_probe *) probe_generic;
-  struct gdbarch *gdbarch = get_objfile_arch (stap_probe->p.objfile);
+  struct gdbarch *gdbarch = get_current_arch ();
 
   /* For SystemTap probes, we have to guarantee that the method
      stap_is_single_operand is defined on gdbarch.  If it is not, then it
@@ -1337,7 +1337,7 @@ handle_stap_probe (struct objfile *objfile, struct sdt_note *el,
 {
   bfd *abfd = objfile->obfd;
   int size = bfd_get_arch_size (abfd) / 8;
-  struct gdbarch *gdbarch = get_objfile_arch (objfile);
+  struct gdbarch *gdbarch = get_current_arch ();
   struct type *ptr_type = builtin_type (gdbarch)->builtin_data_ptr;
   CORE_ADDR base_ref;
   const char *probe_args = NULL;
@@ -1540,7 +1540,7 @@ stap_gen_info_probes_table_values (struct probe *probe_generic,
 
   gdb_assert (probe_generic->pops == &stap_probe_ops);
 
-  gdbarch = get_objfile_arch (probe->p.objfile);
+  gdbarch = get_current_arch ();
 
   if (probe->sem_addr)
     val = print_core_address (gdbarch, probe->sem_addr);
-- 
1.7.11.7


             reply	other threads:[~2013-12-05  6:34 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-12-05  6:34 Sergio Durigan Junior [this message]
2013-12-05 13:04 ` Pedro Alves
2013-12-05 22:12   ` [PATCH] Sanitize gdbarch access on probe/SDT API (was: Re: [PATCH] Use "get_current_arch" instead of...) Sergio Durigan Junior
2013-12-05 23:06     ` [PATCH] Sanitize gdbarch access on probe/SDT API Sergio Durigan Junior
2013-12-05 23:30       ` Tom Tromey
2013-12-06  2:22         ` Sergio Durigan Junior
2013-12-06 13:50     ` Pedro Alves
2013-12-06 15:49       ` Sergio Durigan Junior
2013-12-06 16:00         ` Pedro Alves
2013-12-06 16:04           ` Sergio Durigan Junior

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=1386225226-18549-1-git-send-email-sergiodj@redhat.com \
    --to=sergiodj@redhat.com \
    --cc=gdb-patches@sourceware.org \
    /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