Port aglib.symbol + aglib.kernel + aglib.godbg #2605
Conversation
patryk4815
force-pushed
the
port-kernel2
branch
from
9829eed to
988fffa
Compare
patryk4815
force-pushed
the
port-kernel2
branch
from
13a8170 to
91f6b1d
Compare
patryk4815
changed the title
| @@ -99,8 +97,6 @@ def update() -> None: | |||
| else: | |||
| raise Exception("Pointer size not supported") | |||
|
|
|||
| module.null = pwndbg.dbg.selected_inferior().evaluate_expression("0").cast(void) | |||
There was a problem hiding this comment.
this is not used anywhere, but it cause a lot of trouble in lldb - soo I removed it
pwndbg/arguments.py
Outdated
| try: | ||
| n_args_default = len(sym[0].type.fields()) | ||
| n_args_default = len(sym.type.fields()) |
There was a problem hiding this comment.
len(sym.type.fields()) returns 0 if functions is found, but no debug info...
How should we fallback? Maybe we should create another method for that?
sym.type.func_arguments() -> List[Type] | None and detect no symbols and raise exception? or return None
Only for gdb:
In [26]: sym.type
Out[26]: <gdb.Type code=TYPE_CODE_FUNC name=<text variable, no debug info>>
In [28]: sym.type.name
Out[28]: '<text variable, no debug info>'
There was a problem hiding this comment.
in previous version gdb.lookup_symbol returns None when function don't have debug info.
And in new version with aglib.symbol.lookup_frame_symbol we returns *function address, but some data is missing like function argument (due to missing debug info - when debugging binary without debug info)
There was a problem hiding this comment.
patryk4815
force-pushed
the
port-kernel2
branch
from
4c04601 to
4b4fd97
Compare
# Conflicts: # pwndbg/aglib/kernel/rbtree.py # pwndbg/aglib/tls.py # pwndbg/commands/start.py # pwndbg/dbg/lldb/__init__.py
pwndbg/aglib/kernel/rbtree.py
Outdated
| rb_root_type = gdb.lookup_type("struct rb_root") | ||
| rb_node_type = gdb.lookup_type("struct rb_node") | ||
| rb_root_type = pwndbg.aglib.typeinfo.load("struct rb_root") | ||
| rb_node_type = pwndbg.aglib.typeinfo.load("struct rb_node") |
There was a problem hiding this comment.
After this change, there is no exception here that can be thrown
In [1]: gdb.lookup_type('asdf')
---------------------------------------------------------------------------
error Traceback (most recent call last)
Cell In[1], line 1
----> 1 gdb.lookup_type('asdf')
error: No type named asdf.
In [2]: pwndbg.aglib.typeinfo.load('asdf')
In [3]:so the try: ... except: .. can be removed?
There was a problem hiding this comment.
Also this was wrong before as failure to fetch struct, should set the types to None while it could keep an old/previous value I guess.
There was a problem hiding this comment.
I also wonder if NEW_MODULE is ever valid in the context of kernel debugging?
There was a problem hiding this comment.
probably NEW_MODULE is invalid in kernel context
| pwndbg.dbg_mod.SymbolLookupType.ANY: Domain.ANY, | ||
| pwndbg.dbg_mod.SymbolLookupType.VARIABLE: Domain.VARIABLE, | ||
| pwndbg.dbg_mod.SymbolLookupType.FUNCTION: Domain.FUNCTION, | ||
| }[type] |
There was a problem hiding this comment.
Why do we need this mapping?
If it is really needed, can we do if/elif/elif/else
Trying to address issues:
__libc_start_main#2540Changes:
Todo: