Re: [RFC] [PATCH 2.6.37-rc5-tip 8/20] 8: uprobes: mmap and fork hooks.

[Srikar Dronamraju <srikar at linux dot vnet dot ibm dot com>]

* Peter Zijlstra <peterz@infradead.org> [2011-01-26 11:20:39]:

> On Wed, 2011-01-26 at 14:33 +0530, Srikar Dronamraju wrote:
> > 
> > 
> > I actually dont like to release the write_lock and then reacquire it.
> > write_opcode, which is called thro install_uprobe, i.e to insert the
> > actual breakpoint instruction takes a read lock on the mmap_sem.
> > Hence uprobe_mmap gets called in context with write lock on mmap_sem
> > held, I had to release it before calling install_uprobe. 
> 
> Ah, right, so that's going to give you a head-ache ;-)
> 
> The moment you release this mmap_sem, the map you're going to install
> the probe point in can go away.
> 
> The only way to make this work seems to start by holding the mmap_sem
> for writing and make a breakpoint install function that assumes its
> taken and doesn't try to acquire it again.
> 


Yes, this can be done.
I would have to do something like this in register_uprobe().

list_for_each_entry_safe(mm, tmpmm, &tmp_list, uprobes_list) {
		down_read(&mm->map_sem);
                if (!install_uprobe(mm, uprobe))
                        ret = 0;
		up_read(&mm->map_sem);
                list_del(&mm->uprobes_list);
                mmput(mm);
}

Agree that this is much better than what we have now.