lkml.org 
[lkml]   [2012]   [Apr]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC 0/6] uprobes: kill uprobes_srcu/uprobe_srcu_id
* Peter Zijlstra <peterz@infradead.org> [2012-04-20 12:14:21]:

> On Mon, 2012-04-16 at 23:47 +0200, Oleg Nesterov wrote:
> > On 04/16, Peter Zijlstra wrote:
> > >
> > > On Mon, 2012-04-16 at 01:44 +0200, Oleg Nesterov wrote:
> > >
> > > > And. I have another reason for down_write() in register/unregister.
> > > > I am still not sure this is possible (I had no time to try to
> > > > implement), but it seems to me we can kill the uprobe counter in
> > > > mm_struct.
> > >
> > > You mean by making register/unregister down_write, you're exclusive with
> > > munmap()
> >
> > .. and with register/unregister.
> >
> > Why do we need mm->uprobes_state.count? It is writeonly, except we
> > check it in the DIE_INT3 notifier before anything else to avoid the
> > unnecessary uprobes overhead.
>
> and uprobe_munmap().

If we can kill mm->uprobs_state.count, we can do away with
uprobe_munmap. Because uprobe_munmap is only around to manage
mm->uprobes_state.count.

>
> > Suppose we kill it, and add the new MMF_HAS_UPROBE flag instead.
> > install_breakpoint() sets it unconditionally,
> > uprobe_pre_sstep_notifier() checks it.
>
> Argh, why are MMF_flags part of sched.h.. one would expect those to be
> in mm.h or mm_types.h.. somewhere near struct mm.
>
> > (And perhaps we can stop right here? I mean how often this can
> > slow down the debugger which installs int3 in the same mm?)
> >
> > Now we need to clear MMF_HAS_UPROBE somehowe, when the last
> > uprobe goes away. Lets ignore uprobe_map/unmap for simplicity.
> >
> > - We add another flag, MMF_UPROBE_RECALC, it is set by
> > remove_breakpoint().
> >
> > - We change handle_swbp(). Ignoring all details it does:
> >
> > if (find_uprobe(vaddr))
> > process_uprobe();
> > else if (test_bit(MMF_HAS_UPROBE) && test_bit(MMF_UPROBE_RECALC))
> > recalc_mmf_uprobe_flag();
> >
> > where recalc_mmf_uprobe_flag() checks all vmas and either
> > clears both flags or MMF_UPROBE_RECALC only.
> >
> > This is the really slow O(n) path, but it can only happen after
> > unregister, and only if we hit another non-uprobe breakpoint
> > in the same mm.
> >
> > Something like this. What do you think?
>
> I think I can live with the simple set MMF_HAS_UPROBE and leave it at
> that. The better optimization seems to be to not install breakpoints
> when ->filter() excludes the task..
>
> It looks like we currently install the breakpoint unconditionally and
> only ->filter() once we hit the breakpoint, which is somewhat
> sub-optimal.
>

Yes, We install breakpoints unconditionally, I think we had already
discussed this and Oleg had proposed a solution too.
http://lkml.org/lkml/2011/6/16/470 where we move the mm struct from task
struct to signal struct.

--
Thanks and Regards
Srikar



\
 
 \ /
  Last update: 2012-04-20 12:29    [W:0.232 / U:0.120 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site