[PATCH] hivex/python fix for i386 integer size issue
Hi, While working on Debian packages of hivex 1.2.5, I came across a test failure for the Python bindings with Python 2.7 on the i386 architecture. (The tests ran fine on amd64.) ,---- | $ make -C python check | make[1]: Entering directory `/home/bengen/src/deb/hivex/hivex.git/python' | 010-import.py | 020-open.py | 021-close.py | 200-write.py | python: hivex-py.c:52: get_handle: Assertion `obj' failed. `---- I narrowed this down to hivex-py.c:py_hivex_node_add_child(): The call ,---- | PyArg_ParseTuple (args, (char *) "OLs:hivex_node_add_child", | &py_h, &parent, &name) `---- results in `py_h' set to NULL, though Python's documentation claims that this cannot happen. I think this happens because `parent' is declared a `long int', but "L" in the format string corresponds to a `long long'. On amd64, they have the same size, but on i386 they don't, so the PyObject pointer is written to the wrong address. Please consider applying the patch below which just changes the format string. After regenerating hivex-py.c, I have successfully tested the 1.2.5 code base on both architectures. Cheers, -Hilko diff --git a/generator/generator.ml b/generator/generator.ml index 7e706d1..9722312 100755 --- a/generator/generator.ml +++ b/generator/generator.ml @@ -2875,7 +2875,7 @@ put_val_type (char *val, size_t len, hive_type t) pr "O" | ANode n | AValue n -> - pr "L" + pr "l" | AString n -> pr "s" | AStringNullable n ->
On Thu, Apr 28, 2011 at 12:20:08AM +0200, Hilko Bengen wrote:
Hi,
While working on Debian packages of hivex 1.2.5, I came across a test failure for the Python bindings with Python 2.7 on the i386 architecture. (The tests ran fine on amd64.)
,---- | $ make -C python check | make[1]: Entering directory `/home/bengen/src/deb/hivex/hivex.git/python' | 010-import.py | 020-open.py | 021-close.py | 200-write.py | python: hivex-py.c:52: get_handle: Assertion `obj' failed. `----
I narrowed this down to hivex-py.c:py_hivex_node_add_child():
The call
,---- | PyArg_ParseTuple (args, (char *) "OLs:hivex_node_add_child", | &py_h, &parent, &name) `----
results in `py_h' set to NULL, though Python's documentation claims that this cannot happen. I think this happens because `parent' is declared a `long int', but "L" in the format string corresponds to a `long long'. On amd64, they have the same size, but on i386 they don't, so the PyObject pointer is written to the wrong address.
Please consider applying the patch below which just changes the format string. After regenerating hivex-py.c, I have successfully tested the 1.2.5 code base on both architectures.
Cheers, -Hilko
diff --git a/generator/generator.ml b/generator/generator.ml index 7e706d1..9722312 100755 --- a/generator/generator.ml +++ b/generator/generator.ml @@ -2875,7 +2875,7 @@ put_val_type (char *val, size_t len, hive_type t) pr "O" | ANode n | AValue n -> - pr "L" + pr "l" | AString n -> pr "s" | AStringNullable n ->
Elizabeth, can you try this to see if it fixes your problem too? Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones New in Fedora 11: Fedora Windows cross-compiler. Compile Windows programs, test, and build Windows installers. Over 70 libraries supprt'd http://fedoraproject.org/wiki/MinGW http://www.annexia.org/fedora_mingw
I am going to have to wait until tomorrow to try it -- the work firewall is blocking git.annexia.org. I'll try download it at home tonight. But it makes sense that it would solve it -- I am running on a i386 and after a little more testing found that at one point py_h was set, and then the next call it was null. Fingers crossed for tomorrow! Elizabeth On Thu, Apr 28, 2011 at 3:49 AM, Richard W.M. Jones <rjones@redhat.com> wrote:
On Thu, Apr 28, 2011 at 12:20:08AM +0200, Hilko Bengen wrote:
Hi,
While working on Debian packages of hivex 1.2.5, I came across a test failure for the Python bindings with Python 2.7 on the i386 architecture. (The tests ran fine on amd64.)
,---- | $ make -C python check | make[1]: Entering directory `/home/bengen/src/deb/hivex/hivex.git/python' | 010-import.py | 020-open.py | 021-close.py | 200-write.py | python: hivex-py.c:52: get_handle: Assertion `obj' failed. `----
I narrowed this down to hivex-py.c:py_hivex_node_add_child():
The call
,---- | PyArg_ParseTuple (args, (char *) "OLs:hivex_node_add_child", | &py_h, &parent, &name) `----
results in `py_h' set to NULL, though Python's documentation claims that this cannot happen. I think this happens because `parent' is declared a `long int', but "L" in the format string corresponds to a `long long'. On amd64, they have the same size, but on i386 they don't, so the PyObject pointer is written to the wrong address.
Please consider applying the patch below which just changes the format string. After regenerating hivex-py.c, I have successfully tested the 1.2.5 code base on both architectures.
Cheers, -Hilko
diff --git a/generator/generator.ml b/generator/generator.ml index 7e706d1..9722312 100755 --- a/generator/generator.ml +++ b/generator/generator.ml @@ -2875,7 +2875,7 @@ put_val_type (char *val, size_t len, hive_type t) pr "O" | ANode n | AValue n -> - pr "L" + pr "l" | AString n -> pr "s" | AStringNullable n ->
Elizabeth, can you try this to see if it fixes your problem too?
Rich.
-- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones New in Fedora 11: Fedora Windows cross-compiler. Compile Windows programs, test, and build Windows installers. Over 70 libraries supprt'd http://fedoraproject.org/wiki/MinGW http://www.annexia.org/fedora_mingw
It worked! I'm ready to rock and roll with hivex for python now! This is going to mean big things for the digital forensics community -- up until now our only options have been C or Perl for parsing registry hives, but a lot of the other tools are written in Python. This will streamline our code base. Thanks to Hilko for finding the bug and Richard for writing hivex in the first place. Elizabeth On Thu, Apr 28, 2011 at 12:04 PM, Elizabeth Schweinsberg <bethlogic@gmail.com> wrote:
I am going to have to wait until tomorrow to try it -- the work firewall is blocking git.annexia.org. I'll try download it at home tonight. But it makes sense that it would solve it -- I am running on a i386 and after a little more testing found that at one point py_h was set, and then the next call it was null.
Fingers crossed for tomorrow! Elizabeth
On Thu, Apr 28, 2011 at 3:49 AM, Richard W.M. Jones <rjones@redhat.com> wrote:
On Thu, Apr 28, 2011 at 12:20:08AM +0200, Hilko Bengen wrote:
Hi,
While working on Debian packages of hivex 1.2.5, I came across a test failure for the Python bindings with Python 2.7 on the i386 architecture. (The tests ran fine on amd64.)
,---- | $ make -C python check | make[1]: Entering directory `/home/bengen/src/deb/hivex/hivex.git/python' | 010-import.py | 020-open.py | 021-close.py | 200-write.py | python: hivex-py.c:52: get_handle: Assertion `obj' failed. `----
I narrowed this down to hivex-py.c:py_hivex_node_add_child():
The call
,---- | PyArg_ParseTuple (args, (char *) "OLs:hivex_node_add_child", | &py_h, &parent, &name) `----
results in `py_h' set to NULL, though Python's documentation claims that this cannot happen. I think this happens because `parent' is declared a `long int', but "L" in the format string corresponds to a `long long'. On amd64, they have the same size, but on i386 they don't, so the PyObject pointer is written to the wrong address.
Please consider applying the patch below which just changes the format string. After regenerating hivex-py.c, I have successfully tested the 1.2.5 code base on both architectures.
Cheers, -Hilko
diff --git a/generator/generator.ml b/generator/generator.ml index 7e706d1..9722312 100755 --- a/generator/generator.ml +++ b/generator/generator.ml @@ -2875,7 +2875,7 @@ put_val_type (char *val, size_t len, hive_type t) pr "O" | ANode n | AValue n -> - pr "L" + pr "l" | AString n -> pr "s" | AStringNullable n ->
Elizabeth, can you try this to see if it fixes your problem too?
Rich.
-- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones New in Fedora 11: Fedora Windows cross-compiler. Compile Windows programs, test, and build Windows installers. Over 70 libraries supprt'd http://fedoraproject.org/wiki/MinGW http://www.annexia.org/fedora_mingw
On Fri, Apr 29, 2011 at 03:15:03PM -0400, Elizabeth Schweinsberg wrote:
It worked! I'm ready to rock and roll with hivex for python now!
Thanks for testing it.
This is going to mean big things for the digital forensics community -- up until now our only options have been C or Perl for parsing registry hives, but a lot of the other tools are written in Python. This will streamline our code base.
If there are any things missing, then bring them up for discussion. At the moment I'm also interested in the Windows event logs, although for Win >= Vista there is at least a reasonably good set of GPL tools for it (http://computer.forensikblog.de/files/evtx/). Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones virt-p2v converts physical machines to virtual machines. Boot with a live CD or over the network (PXE) and turn machines into Xen guests. http://et.redhat.com/~rjones/virt-p2v
On Thu, Apr 28, 2011 at 12:20:08AM +0200, Hilko Bengen wrote:
Hi,
While working on Debian packages of hivex 1.2.5, I came across a test failure for the Python bindings with Python 2.7 on the i386 architecture. (The tests ran fine on amd64.)
,---- | $ make -C python check | make[1]: Entering directory `/home/bengen/src/deb/hivex/hivex.git/python' | 010-import.py | 020-open.py | 021-close.py | 200-write.py | python: hivex-py.c:52: get_handle: Assertion `obj' failed. `----
I narrowed this down to hivex-py.c:py_hivex_node_add_child():
The call
,---- | PyArg_ParseTuple (args, (char *) "OLs:hivex_node_add_child", | &py_h, &parent, &name) `----
results in `py_h' set to NULL, though Python's documentation claims that this cannot happen. I think this happens because `parent' is declared a `long int', but "L" in the format string corresponds to a `long long'. On amd64, they have the same size, but on i386 they don't, so the PyObject pointer is written to the wrong address.
Please consider applying the patch below which just changes the format string. After regenerating hivex-py.c, I have successfully tested the 1.2.5 code base on both architectures.
Cheers, -Hilko
diff --git a/generator/generator.ml b/generator/generator.ml index 7e706d1..9722312 100755 --- a/generator/generator.ml +++ b/generator/generator.ml @@ -2875,7 +2875,7 @@ put_val_type (char *val, size_t len, hive_type t) pr "O" | ANode n | AValue n -> - pr "L" + pr "l" | AString n -> pr "s" | AStringNullable n ->
Looking at the code, I think this was clearly a mistake and I've pushed your fix. Thanks for your contribution! Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones virt-top is 'top' for virtual machines. Tiny program with many powerful monitoring features, net stats, disk stats, logging, etc. http://et.redhat.com/~rjones/virt-top
participants (3)
-
Elizabeth Schweinsberg -
Hilko Bengen -
Richard W.M. Jones