Re: inventing the wheel?

[Carl-Daniel Hailfinger <c-d.hailfinger.kernel.2003@gmx.net>]

Carl Spalletta wrote:
> I was interested in finding a tool that would tell me all the paths
> through the kernel leading to some particular function, for example in
> the case of do_mmap_pgoff:
> 
> do_mmap_pgoff do_mmap2 old_mmap old_mmap_i386
> do_mmap_pgoff do_mmap2 sys_mmap2
> do_mmap_pgoff do_mmap aio_setup_ring ioctx_alloc sys_io_setup
> do_mmap_pgoff do_mmap elf_map load_elf_binary
> ...
> 
> I submitted a tool ('fscope') to do this but no one has picked up
> on the discussion. So I am wondering if there isn't already some
> existing and better way to accomplish the same thing.
> 
> Could somebody tell me please, what is that way?
> 
> I know you can do a backtrace w/ gdb but that begs the question
> how are you going to sure you have found every path?

It seems everybody is busy trying Linus' sparse, so maybe it was
overlooked. Right now, it seems we have a few new tools to play with,
which were not available at the time of 2.4-test.

(Alphabetic order)
-Checker (Stanford people)
-Fscope (Carl Spalletta)
-Smatch (Dan Carpenter)
-Sparse (Linus Torvalds)

Now we only need one additional tool to *prove* correctness of the
kernel ;-)

-an automatic race finder

Liberal use of these tools should result in the most stable kernel ever.


Regards,
Carl-Daniel