[PATCH 2/5] gdb: value_fetch_lazy_register: fix getting frame id while it is computed

Simon Marchi simon.marchi@polymtl.ca
Mon Jun 28 17:44:26 GMT 2021 

This gives an internal error:

    $ ./gdb -q -nx --data-directory=data-directory  -q testsuite/outputs/gdb.python/py-unwind/py-unwind -x /home/simark/src/binutils-gdb/gdb/testsuite/gdb.python/py-unwind.py -ex "set debug frame 1" -ex r
    ...
    /home/simark/src/binutils-gdb/gdb/frame.c:623: internal-error: frame_id get_frame_id(frame_info*): Assertion `fi->this_id.p != frame_id_status::COMPUTING' failed.

The backtrace is:

    #9  0x000055d0b48e294a in internal_error (file=0x55d0b4c71380 "/home/simark/src/binutils-gdb/gdb/frame.c", line=623, fmt=0x55d0b4c712a0 "%s: Assertion `%s' failed.") at /home/simark/src/binutils-gdb/gdbsupport/errors.cc:55
    #10 0x000055d0b28aade3 in get_frame_id (fi=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/frame.c:623
    #11 0x000055d0b28b5350 in get_prev_frame_id_by_id (id=<error reading variable: Cannot access memory at address 0xfffffffffffffd9c>) at /home/simark/src/binutils-gdb/gdb/frame.c:2601
    #12 0x000055d0b375a1ab in value_fetch_lazy_register (val=0x6110000b3500) at /home/simark/src/binutils-gdb/gdb/value.c:3955
    #13 0x000055d0b375a9e7 in value_fetch_lazy (val=0x6110000b3500) at /home/simark/src/binutils-gdb/gdb/value.c:4024
    #14 0x000055d0b28a0548 in value_of_register (regnum=6, frame=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/findvar.c:274
    #15 0x000055d0b301b05a in pending_framepy_read_register (self=0x7f79056bb7f0, args=0x7f790581aa90) at /home/simark/src/binutils-gdb/gdb/python/py-unwind.c:398
    <some Python frames>
    #27 0x000055d0b301c6b4 in pyuw_sniffer (self=0x62100018e250, this_frame=0x6210001ff5e0, cache_ptr=0x6210001ff5f8) at /home/simark/src/binutils-gdb/gdb/python/py-unwind.c:568
    #28 0x000055d0b28a6587 in frame_unwind_try_unwinder (this_frame=0x6210001ff5e0, this_cache=0x6210001ff5f8, unwinder=0x62100018e250) at /home/simark/src/binutils-gdb/gdb/frame-unwind.c:130
    #29 0x000055d0b28a67b5 in frame_unwind_find_by_frame (this_frame=0x6210001ff5e0, this_cache=0x6210001ff5f8) at /home/simark/src/binutils-gdb/gdb/frame-unwind.c:193
    #30 0x000055d0b28aa80f in compute_frame_id (fi=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/frame.c:585
    #31 0x000055d0b28aaeae in get_frame_id (fi=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/frame.c:635
    #32 0x000055d0b28b5350 in get_prev_frame_id_by_id (id=<error reading variable: Cannot access memory at address 0xfffffffffffffd9c>) at /home/simark/src/binutils-gdb/gdb/frame.c:2601
    #33 0x000055d0b375a1ab in value_fetch_lazy_register (val=0x6110000b3280) at /home/simark/src/binutils-gdb/gdb/value.c:3955
    #34 0x000055d0b375a9e7 in value_fetch_lazy (val=0x6110000b3280) at /home/simark/src/binutils-gdb/gdb/value.c:4024
    #35 0x000055d0b28a0548 in value_of_register (regnum=1, frame=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/findvar.c:274
    #36 0x000055d0b2826b1c in eval_op_register (expect_type=0x6210001949f0, exp=0x6030000d6840, noside=EVAL_NORMAL, name=0x60400003e2e8 "rbx") at /home/simark/src/binutils-gdb/gdb/eval.c:1074
    #37 0x000055d0b218139f in expr::register_operation::evaluate (this=0x60400003e2d0, expect_type=0x6210001949f0, exp=0x6030000d6840, noside=EVAL_NORMAL) at /home/simark/src/binutils-gdb/gdb/expop.h:830
    #38 0x000055d0b282f5d5 in expr::operation::evaluate_for_cast (this=0x60400003e2d0, expect_type=0x6210001949f0, exp=0x6030000d6840, noside=EVAL_NORMAL) at /home/simark/src/binutils-gdb/gdb/eval.c:2506
    #39 0x000055d0b2181c64 in expr::unop_cast_operation::evaluate (this=0x6030000d6870, expect_type=0x6210001949f0, exp=0x6030000d6840, noside=EVAL_NORMAL) at /home/simark/src/binutils-gdb/gdb/expop.h:1975
    #40 0x000055d0b2820b04 in expression::evaluate (this=0x6030000d6840, expect_type=0x6210001949f0, noside=EVAL_NORMAL) at /home/simark/src/binutils-gdb/gdb/eval.c:101
    #41 0x000055d0b2820c46 in evaluate_expression (exp=0x6030000d6840, expect_type=0x6210001949f0) at /home/simark/src/binutils-gdb/gdb/eval.c:115
    #42 0x000055d0b337b0fd in stap_probe::evaluate_argument (this=0x60f0000053b0, n=1, frame=0x6210001ff5e0) at /home/simark/src/binutils-gdb/gdb/stap-probe.c:1435
    #43 0x000055d0b32d86e8 in svr4_handle_solib_event () at /home/simark/src/binutils-gdb/gdb/solib-svr4.c:1932
    #44 0x000055d0b32f1f39 in handle_solib_event () at /home/simark/src/binutils-gdb/gdb/solib.c:1258
    #45 0x000055d0b21e6c62 in bpstat_stop_status (aspace=0x6030000554e0, bp_addr=0x7ffff7fd18d8, thread=0x61700003db80, ws=0x7ffcff02c708, stop_chain=0x0) at /home/simark/src/binutils-gdb/gdb/breakpoint.c:5470
    #46 0x000055d0b2aa44c2 in handle_signal_stop (ecs=0x7ffcff02c6e0) at /home/simark/src/binutils-gdb/gdb/infrun.c:6254
    #47 0x000055d0b2aa0962 in handle_inferior_event (ecs=0x7ffcff02c6e0) at /home/simark/src/binutils-gdb/gdb/infrun.c:5740
    #48 0x000055d0b2a94a0f in fetch_inferior_event () at /home/simark/src/binutils-gdb/gdb/infrun.c:4111

When the get_frame_id call is done at frame 31, on frame #0, the
id is not computed yet, and therefore compute_frame_id is called.
compute_frame_id looks for an unwinder, which goes through the Python
unwinders.  When value_fetch_lazy_register is called at frame 12,
get_frame_id is called again for frame #0.  But we are in the process of
computing that frame id, so it can't work.  The assert in get_frame_id
works as intended.

The second call to get_frame_id happens in

    VALUE_FRAME_ID (val)

at value.c:3955.  If we dig down this, we see that:

 1. VALUE_NEXT_FRAME_ID is called first, returning the next frame's id,
    stored in struct value (it happens to be the sentinel's frame id, in
    this case)
 2. it is passed to get_prev_frame_id_by_id
 3. frame_find_by_id is called on that, that returns the frame_info for
    the sentinel frame
 4. get_prev_frame is called on that, that returns a frame_info for our
    frame #0
 5. get_frame_id is called on that, this is where the gdb_assert fails.
 6. had the previous step worked, value_fetch_lazy_register would have
    called frame_find_by_id on it, to get the frame_info for frame #0

In the end, all value_fetch_lazy_register wants is a frame_info for the
value's frame.  In step 4, we have that.  In step 5, we fetch the
frame's id only to go back to a frame_info in step 6.  That seems
unnecessary, and causes the internal error.

Change value_fetch_lazy_register to do the more straightforward thing:

 - fetch the value's next frame id
 - fetch the frame info from that
 - fetch the prev frame from that (the value's frame)

With this change, the command shown above no longer asserts and the
produced message makes sense:

    { value_fetch_lazy (frame=0,regnum=1(rbx),...) -> register=1 bytes=[c0e0fff7ff7f0000] }

gdb/ChangeLog:

	* value.c (value_fetch_lazy_register): Don't use VALUE_FRAME_ID
	to fetch value's frame for debug message.

Change-Id: I294f06f8b32206d442da5355f1b897159414e5b6
---
 gdb/value.c | 21 +++++++++------------
 1 file changed, 9 insertions(+), 12 deletions(-)

diff --git a/gdb/value.c b/gdb/value.c
index 9df035a50b3b..6c969a911376 100644
--- a/gdb/value.c
+++ b/gdb/value.c
@@ -3883,8 +3883,6 @@ value_fetch_lazy_memory (struct value *val)
 static void
 value_fetch_lazy_register (struct value *val)
 {
-  struct frame_info *next_frame;
-  int regnum;
   struct type *type = check_typedef (value_type (val));
   struct value *new_val = val, *mark = value_mark ();
 
@@ -3896,8 +3894,8 @@ value_fetch_lazy_register (struct value *val)
     {
       struct frame_id next_frame_id = VALUE_NEXT_FRAME_ID (new_val);
 
-      next_frame = frame_find_by_id (next_frame_id);
-      regnum = VALUE_REGNUM (new_val);
+      frame_info *next_frame = frame_find_by_id (next_frame_id);
+      int regnum = VALUE_REGNUM (new_val);
 
       gdb_assert (next_frame != NULL);
 
@@ -3948,18 +3946,17 @@ value_fetch_lazy_register (struct value *val)
 
   if (frame_debug)
     {
-      struct gdbarch *gdbarch;
-      struct frame_info *frame;
-      /* VALUE_FRAME_ID is used here, instead of VALUE_NEXT_FRAME_ID,
-	 so that the frame level will be shown correctly.  */
-      frame = frame_find_by_id (VALUE_FRAME_ID (val));
-      regnum = VALUE_REGNUM (val);
-      gdbarch = get_frame_arch (frame);
+      frame_id next_frame_id = VALUE_NEXT_FRAME_ID (val);
+      frame_info *next_frame = frame_find_by_id (next_frame_id);
+      frame_info *value_frame = get_prev_frame (next_frame);
+
+      int regnum = VALUE_REGNUM (val);
+      gdbarch *gdbarch = get_frame_arch (value_frame);
 
       fprintf_unfiltered (gdb_stdlog,
 			  "{ value_fetch_lazy "
 			  "(frame=%d,regnum=%d(%s),...) ",
-			  frame_relative_level (frame), regnum,
+			  frame_relative_level (value_frame), regnum,
 			  user_reg_map_regnum_to_name (gdbarch, regnum));
 
       fprintf_unfiltered (gdb_stdlog, "->");
-- 
2.32.0

More information about the Gdb-patches mailing list