Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Andrew Burgess <aburgess@broadcom.com>
To: <gdb-patches@sourceware.org>
Cc: Andrew Burgess <aburgess@broadcom.com>
Subject: [PATCH v3 2/4] Remove previous frame if an error occurs when computing frame id during unwind.
Date: Wed, 30 Apr 2014 10:55:00 -0000	[thread overview]
Message-ID: <1398855344-25278-3-git-send-email-aburgess@broadcom.com> (raw)
In-Reply-To: <1398855344-25278-1-git-send-email-aburgess@broadcom.com>

In get_prev_frame_if_no_cycle, if we throw an error during compute_frame_id
then we are left in a state where THIS_FRAME has a PREV_FRAME attached, but
PREV_FRAME has no frame id.  This is an unexpected state that causes
internal errors and assertions to fire.

This patch adds a cleanup that removes the previous frame created by
get_prev_frame_raw if we get an error.

OK to apply?

Thanks,
Andrew



gdb/ChangeLog:

	* frame.c (remove_prev_frame): New function.
	(get_prev_frame_if_no_cycle): Create / discard cleanup using
	remove_prev_frame.

---
 gdb/frame.c | 52 ++++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 38 insertions(+), 14 deletions(-)

diff --git a/gdb/frame.c b/gdb/frame.c
index 97d54e9..5f05968 100644
--- a/gdb/frame.c
+++ b/gdb/frame.c
@@ -1733,6 +1733,22 @@ frame_register_unwind_location (struct frame_info *this_frame, int regnum,
     }
 }
 
+/* Called during frame unwinding to remove a previous frame pointer from a
+   frame passed in ARG.  */
+
+static void
+remove_prev_frame (void *arg)
+{
+  struct frame_info *this_frame, *prev_frame;
+
+  this_frame = (struct frame_info *) arg;
+  prev_frame = this_frame->prev;
+  gdb_assert (prev_frame != NULL);
+
+  prev_frame->next = NULL;
+  this_frame->prev = NULL;
+}
+
 /* Get the previous raw frame, and check that it is not identical to
    same other frame frame already in the chain.  If it is, there is
    most likely a stack cycle, so we discard it, and mark THIS_FRAME as
@@ -1745,28 +1761,36 @@ static struct frame_info *
 get_prev_frame_if_no_cycle (struct frame_info *this_frame)
 {
   struct frame_info *prev_frame;
+  struct cleanup *prev_frame_cleanup;
 
   prev_frame = get_prev_frame_raw (this_frame);
   if (prev_frame == NULL)
     return NULL;
 
-  compute_frame_id (prev_frame);
-  if (frame_stash_add (prev_frame))
-    return prev_frame;
+  /* The cleanup will remove the previous frame that get_prev_frame_raw
+     linked onto THIS_FRAME.  */
+  prev_frame_cleanup = make_cleanup (remove_prev_frame, this_frame);
 
-  /* Another frame with the same id was already in the stash.  We just
-     detected a cycle.  */
-  if (frame_debug)
+  compute_frame_id (prev_frame);
+  if (!frame_stash_add (prev_frame))
     {
-      fprintf_unfiltered (gdb_stdlog, "-> ");
-      fprint_frame (gdb_stdlog, NULL);
-      fprintf_unfiltered (gdb_stdlog, " // this frame has same ID }\n");
+      /* Another frame with the same id was already in the stash.  We just
+	 detected a cycle.  */
+      if (frame_debug)
+	{
+	  fprintf_unfiltered (gdb_stdlog, "-> ");
+	  fprint_frame (gdb_stdlog, NULL);
+	  fprintf_unfiltered (gdb_stdlog, " // this frame has same ID }\n");
+	}
+      this_frame->stop_reason = UNWIND_SAME_ID;
+      /* Unlink.  */
+      prev_frame->next = NULL;
+      this_frame->prev = NULL;
+      prev_frame = NULL;
     }
-  this_frame->stop_reason = UNWIND_SAME_ID;
-  /* Unlink.  */
-  prev_frame->next = NULL;
-  this_frame->prev = NULL;
-  return NULL;
+
+  discard_cleanups (prev_frame_cleanup);
+  return prev_frame;
 }
 
 /* Return a "struct frame_info" corresponding to the frame that called
-- 
1.8.1.3


  parent reply	other threads:[~2014-04-30 10:55 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-04-04 14:46 [RFC 0/4] Catch errors in get_prev_frame Andrew Burgess
2014-04-04 14:47 ` [RFC 1/4] New tests for backtracing with a corrupted stack Andrew Burgess
2014-04-04 14:48 ` [RFC 2/4] Remove previous frame if we error during compute_frame_id Andrew Burgess
2014-04-04 14:53   ` Andrew Burgess
2014-04-15 19:02     ` Pedro Alves
2014-04-04 14:49 ` [RFC 3/4] Deprecate frame_stop_reason_string Andrew Burgess
2014-04-04 14:55   ` Andrew Burgess
2014-04-04 14:50 ` [RFC 4/4] Add TRY_CATCH to get_prev_frame and frame specific strop strings Andrew Burgess
2014-04-15  9:11 ` [RFC 0/4] Catch errors in get_prev_frame Andrew Burgess
2014-04-17 10:15 ` [PATCH v2 1/4] New test for backtrace when the stack pointer is invalid (inaccessible) Andrew Burgess
2014-04-17 10:15 ` [PATCH v2 3/4] Deprecate frame_stop_reason_string Andrew Burgess
2014-04-29 19:56   ` Pedro Alves
2014-04-30 10:46     ` Andrew Burgess
2014-04-17 10:15 ` [PATCH v2 0/4] Catch errors in get_prev_frame Andrew Burgess
2014-04-17 10:15 ` [PATCH v2 4/4] Add a TRY_CATCH to get_prev_frame to better handle errors during unwind Andrew Burgess
2014-04-17 10:15 ` [PATCH v2 2/4] Remove previous frame if an error occurs when computing frame id " Andrew Burgess
2014-04-30 10:55 ` [PATCH v3 0/4] Catch errors in get_prev_frame Andrew Burgess
2014-04-30 10:55   ` [PATCH v3 1/4] New test for backtrace when the stack pointer is invalid (inaccessible) Andrew Burgess
2014-05-28 18:42     ` Pedro Alves
2014-04-30 10:55   ` [PATCH v3 4/4] Add a TRY_CATCH to get_prev_frame to better handle errors during unwind Andrew Burgess
2014-05-28 18:31     ` Pedro Alves
2014-05-28 23:35       ` Andrew Burgess
2014-05-29  9:41         ` Pedro Alves
2014-05-29 23:02           ` Andrew Burgess
2014-05-30 11:46             ` Pedro Alves
2014-04-30 10:55   ` Andrew Burgess [this message]
2014-05-16 15:37     ` [PATCH v3 2/4] Remove previous frame if an error occurs when computing frame id " Pedro Alves
2014-05-28 23:16       ` Andrew Burgess
2014-04-30 10:55   ` [PATCH v3 3/4] Deprecate frame_stop_reason_string Andrew Burgess
2014-05-28 17:26     ` Pedro Alves
2014-05-28 23:26       ` Andrew Burgess
2014-05-29  9:00         ` Pedro Alves
2014-05-29  9:53           ` Andrew Burgess
2014-05-29  9:56             ` Pedro Alves

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=1398855344-25278-3-git-send-email-aburgess@broadcom.com \
    --to=aburgess@broadcom.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