ASSERT(1.5.0 r242 gcc on Ubuntu 9.10 x86 and x64) is_dynamo_address((app_pc) xsp)) core/linux/signal.c:986

[derekbruening]

From derek.br...@gmail.com on November 23, 2009 15:29:57

I suspect a problem with the feature added in issue #149 .
In addition to the assert reported below, issue #149 also has two other
issues discovered post-commit. That issue can be re-opened instead if this
assert is unrelated to the first issue:

• to investigate whether the arg to the initial thread func is being passed
  properly since we're not copying whatever the kernel puts on the
  dstack onto the app stack
• the vfork handling is incorrect. it was added post-review. it was
  incorrectly
  pasted from the clone handling and uses syscall args that do not exist
  for vfork.
  the vfork test must only pass for coincidental reasons.

The assert, along with a subsequent hang which is not surprising if the
child died due to the assert:

Date: Mon, 23 Nov 2009 13:51:28 -0500
From: Peter Feiner

I also reran gcc (version 4.4) under DynamoRIO debug builds using no client
at all. Both my 32-bit and 64-bit system fail the same assertion on line 986
of core/linux/signal.c:

ASSERT(is_dynamo_address((app_pc) xsp));

Here's the 32-bit output

<Application gcc (9744). Internal Error Internal DynamoRIO Error:
/home/peter/dynamorio-read-only/core/linux/signal.c:986
version 1.5.0, build 18

0xbfcfbeec 0x0035e791
0xbfcfc04c 0x0044ca5a
0xbfcfc06c 0x0042f849
0xbfcfc0bc 0x1f5528e4
0xbfcfc198 0x0805fcfe
0xbfcfc218 0x0805ff03
0xbfcfc248 0x0804d8b0
0xbfcfc2f8 0x0804df29
0xbfcfc3d8 0x0804feb5
0xbfcfc448 0x0804eafd
0xbfcfc528 0x0804edbd>
<Stopping application gcc (9744)>

Here's the 64-bit output:

<Application gcc (21734). Internal Error Internal DynamoRIO Error:
/home/peter/ece1724/dynamorio-read-only/core/linux/signal.c:986
version 1.5.0, build 18

0x00007fff498d54b0 0x000000007109ed93
0x00007fff498d5600 0x00000000711801f1
0x00007fff498d5620 0x000000007116645b
0x00007fff498d5660 0x0000000071f0fa44>
<Stopping application gcc (21734)>

In both cases, the application hangs and is unresponsive to signals issued
via ctrl-| and ctrl-c. To kill the hung processes, I have to issue kill -9.
I generated stack dumps for each hung process before killing it. Here's the
32-bit dump:

$ gdb program 9744
(gdb) bt
#0 0x0043398e in syscall_0args () from
/home/peter/dynamorio-read-only/exports-debug/lib32/debug/libdynamorio.so
#1 0x1f5950a8 in ?? ()
#2 0x0044ded3 in signal_thread_exit (dcontext=0x1f53d800) at
/home/peter/dynamorio-read-only/core/linux/signal.c:1248
#3 0x00444098 in os_thread_exit (dcontext=0x1f53d800) at
/home/peter/dynamorio-read-only/core/linux/os.c:1397
#4 0x002f94dc in dynamo_thread_exit_common (dcontext=0x1f53d800, id=, other_thread=0) at
/home/peter/dynamorio-read-only/core/dynamo.c:2067
#5 0x002f9a99 in dynamo_thread_exit () at
/home/peter/dynamorio-read-only/core/dynamo.c:2157
#6 0x002faaf6 in dynamo_process_exit_cleanup () at
/home/peter/dynamorio-read-only/core/dynamo.c:1091
#7 dynamo_process_exit () at
/home/peter/dynamorio-read-only/core/dynamo.c:1160
#8 0x004338d8 in cat_done_saving_dstack () from
/home/peter/dynamorio-read-only/exports-debug/lib32/debug/libdynamorio.so
#9 0x00433937 in cat_have_lock () from
/home/peter/dynamorio-read-only/exports-debug/lib32/debug/libdynamorio.so
#10 0x1f590000 in ?? ()
#11 0x1f53d800 in ?? ()
Backtrace stopped: previous frame inner to this frame (corrupt stack?)

Here's the 64-bit dump:

$ gdb program 21734
(gdb) bt
#0 0x000000007116a2a0 in syscall_ready () from
/home/peter/ece1724/dynamorio-read-only/exports-gcc4.4-64bit-debug/lib64/debug/libdynamorio.so
#1 0x0000000071f27700 in ?? ()
#2 0x0000000071174d69 in thread_yield () at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:1930
#3 0x0000000071182328 in signal_thread_exit (dcontext=0x18) at
/home/peter/ece1724/dynamorio-read-only/core/linux/signal.c:1248
#4 0x0000000071178fab in os_thread_exit (dcontext=0x71f27700) at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:1397
#5 0x0000000071041499 in dynamo_thread_exit_common (dcontext=0x71f27700,
id=21734, other_thread=0) at
/home/peter/ece1724/dynamorio-read-only/core/dynamo.c:2067
#6 0x0000000071041972 in dynamo_thread_exit () at
/home/peter/ece1724/dynamorio-read-only/core/dynamo.c:2157
#7 0x00000000710428aa in dynamo_process_exit_cleanup () at
/home/peter/ece1724/dynamorio-read-only/core/dynamo.c:1091
#8 dynamo_process_exit () at
/home/peter/ece1724/dynamorio-read-only/core/dynamo.c:1160
#9 0x000000007116a1ea in cat_done_saving_dstack () from
/home/peter/ece1724/dynamorio-read-only/exports-gcc4.4-64bit-debug/lib64/debug/libdynamorio.so
#10 0x000000007116a24b in global_do_syscall_sysenter () from
/home/peter/ece1724/dynamorio-read-only/exports-gcc4.4-64bit-debug/lib64/debug/libdynamorio.so
#11 0x0000000071f76000 in ?? ()
#12 0x0000000071f27700 in ?? ()
#13 0x00000000000000e7 in ?? ()
#14 0x0000000000000001 in ?? ()
#15 0x000000000000003c in ?? ()
#16 0x0000000000000001 in ?? ()
#17 0x000000007117b5ad in handle_exit (dcontext=0x71f27700) at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:3349
#18 pre_system_call (dcontext=0x71f27700) at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:3427
#19 0x0000000071f27700 in ?? ()
#20 0x00000000000000e7 in ?? ()
#21 0x0000000000000001 in ?? ()
#22 0x000000000000003c in ?? ()
#23 0x0000000000000001 in ?? ()
#24 0x000000007117b5ad in handle_exit (dcontext=0x71f76000) at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:3349
#25 pre_system_call (dcontext=0x71f76000) at
/home/peter/ece1724/dynamorio-read-only/core/linux/os.c:3427
#26 0x0000000000000000 in ?? ()

These dumps do not change over time. That is, I requested a backtrace from
gdb, waited 10 seconds, and requested another.

Original issue: http://code.google.com/p/dynamorio/issues/detail?id=229

[derekbruening]

From derek.br...@gmail.com on November 23, 2009 12:34:33

Summary: ASSERT(1.5.0 r242 gcc on Ubuntu 9.10 x86 and x64) is_dynamo_address((app_pc) xsp)) core/linux/signal.c:986

[derekbruening]

From derek.br...@gmail.com on December 01, 2009 09:45:50

fixed in r249

Status: Verified