crash-python initialization fails #50

Activity

1. dassanjib

   changed the title [-]python-crash initialization fails[/-] [+]crash-python initialization fails[/+] on Aug 13, 2024
2. Werkov commented on Aug 13, 2024

   

   Werkov

   on Aug 13, 2024

   Contributor

   More actions

   What's your
   python3 --version?
   (type hinting syntax is only supported from 3.6)
   But this might be actually the python-in-gdb, do you have also this gdb installed?
3. dassanjib commented on Aug 14, 2024

   

   dassanjib

   on Aug 14, 2024

   Author

   More actions

   Thank you for you comments. Please find python & gdb versions' information, below.

   [crash-python]python --version
   Python 3.8.19
   [crash-python]
   [crash-python]./pycrash vmlinux vmcore 
   warning: help text documentation is out-of-date
   To update it, run 'make doc-help'
   Traceback (most recent call last):
     File "<string>", line 4, in <module>
     File "./build/lib/crash/__init__.py", line 6
       def archname() -> str:
                      ^
   SyntaxError: invalid syntax
   ./test-gdb-compatibility.gdbinit:16: Error in sourced command file:
   Error while executing Python code.
   fatal: crash-python cannot initialize
   

   Yes I installed python-in-gdb (followed this page https://github.com/crash-python/crash-python).
   Although the package shows gdb version 12.1 but, post installation I found a different version (8.3). Is there anything am I missing?

   [gdb-python]./gdb/gdb --version
   GNU gdb (GDB) 8.3.50.20190416-git
   
   
4. Werkov commented on Aug 14, 2024

   

   Werkov

   on Aug 14, 2024

   Contributor

   More actions

   Have you checked out the right branch when building from git? I.e. gdb-12.1-target, not master.
5. dassanjib commented on Aug 19, 2024

   

   dassanjib

   on Aug 19, 2024

   Author

   More actions

   I am trying with gdb-12.1-target now. I shall come back with updates.
6. dassanjib commented on Aug 19, 2024

   

   dassanjib

   on Aug 19, 2024

   Last edited by dassanjib

   Author

   More actions

   I have been facing multiple compilation errors for gdb-12.1-target.

   python/py-target.c: In member function ??virtual target_xfer_status python_target::xfer_partial(target_object, const char*, gdb_b
   yte*, const gdb_byte*, ULONGEST, ULONGEST, ULONGEST*)??:
   python/py-target.c:218:8: error: ??PyMemoryView_FromMemory?? was not declared in this scope
     obj = PyMemoryView_FromMemory(buf, len, PyBUF_WRITE);
           ^~~~~~~~~~~~~~~~~~~~~~~
   python/py-target.c:218:8: note: suggested alternative: ??PyMemoryView_FromBuffer??
     obj = PyMemoryView_FromMemory(buf, len, PyBUF_WRITE);
           ^~~~~~~~~~~~~~~~~~~~~~~
           PyMemoryView_FromBuffer
   python/py-target.c:234:8: error: ??PyMemoryView_FromMemory?? was not declared in this scope
     obj = PyMemoryView_FromMemory(buf, len, PyBUF_READ);
           ^~~~~~~~~~~~~~~~~~~~~~~
   python/py-target.c:234:8: note: suggested alternative: ??PyMemoryView_FromBuffer??
     obj = PyMemoryView_FromMemory(buf, len, PyBUF_READ);
           ^~~~~~~~~~~~~~~~~~~~~~~
           PyMemoryView_FromBuffer
   python/py-target.c: In function ??void decode_and_write_register(thread_info*, regcache*, int, PyObject*)??:
   python/py-target.c:520:24: warning: format ??%s?? expects argument of type ??char*??, but argument 8 has type ??const std::shared_ptr
   <std::__cxx11::basic_string<char> >?? [-Wformat=]
       name, except.message);
                           ^
   Makefile:1662: recipe for target 'python/py-target.o' failed
   
7. dassanjib commented on Aug 23, 2024

   

   dassanjib

   on Aug 23, 2024

   Last edited by dassanjib

   Author

   More actions

   Meanwhile I fix this build issue, may I share some findings with gdb 15.1.

   GDB backtrace fails to access kernel memory in vmalloc range. This is with a linux kernel crash coredump on x86_64 platform.

   In my understanding 0xffffc90000000000 is the vmalloc start range with 48-bit virtual addressing system.
   https://www.kernel.org/doc/Documentation/x86/x86_64/mm.txt

   ffffc90000000000 |  -55    TB | ffffe8ffffffffff |   32 TB | vmalloc/ioremap space (vmalloc_base)
   

   And I have CONFIG_VMAP_STACK config enabled in my kernel configuration. So kernel call stack addresses
   are expected to be from that range.
   https://lwn.net/Articles/692208/
   https://lwn.net/Articles/691631/
   https://www.kernel.org/doc/Documentation/mm/vmalloced-kernel-stacks.rst

   But gdb backtrace fails, I probably dont have the answer why.

   (gdb) bt
    #0  0xffffffff817807ab in neigh_get_next (n=0xffff888129ff0a00, pos=0x0 <fixed_percpu_data>, 
       seq=<optimized out>, seq=<optimized out>) at net/core/neighbour.c:3220
    Backtrace stopped: Cannot access memory at address 0xffffc90000c77dc0
   

   However we dont face this issue with linux "crash" utility tool.

   crash>bt
   #6 [ffffc90000c77dc0] neigh_seq_next at ffffffff8178096d
   #7 [ffffc90000c77de0] seq_read_iter at ffffffff812302c4
   #8 [ffffc90000c77e38] seq_read at ffffffff8123054e
   #9 [ffffc90000c77eb8] proc_reg_read at ffffffff812806b6
   #10 [ffffc90000c77ed0] vfs_read at ffffffff812046c0
   #11 [ffffc90000c77f08] ksys_read at ffffffff81204a7a
   #12 [ffffc90000c77f40] do_syscall_64 at ffffffff81a17bbd
   #13 [ffffc90000c77f50] entry_SYSCALL_64_after_hwframe at ffffffff81c00099
   

   My query is, does crash-pthon throw similar backtrace failure errors with linux kernel coredump from x86_64?

   Thanks in advance.
   Sanjib