Re: [PATCH v4 3.0-rc2-tip 7/22] 7: uprobes: mmap and fork hooks.

[Peter Zijlstra <peterz@infradead.org>]

On Thu, 2011-06-16 at 18:30 +0530, Srikar Dronamraju wrote:

> Now since a register and mmap operations can run in parallel, we could
> have subtle race conditions like this:
> 
> 1. register_uprobe inserts the uprobe in RB tree.
> 2. register_uprobe loops thro vmas and inserts breakpoints.
> 
> 3. mmap is called for same inode, mmap_uprobe() takes reference; 
> 4. mmap completes insertion and releases reference.
> 
> 5. register uprobe tries to install breakpoint on one vma fails and not
> due to -ESRCH or -EEXIST.
> 6. register_uprobe rolls back all install breakpoints except the one
> inserted by mmap.
> 
> We end up with breakpoints that we have inserted by havent cleared.
> 
> Similarly unregister_uprobe might be looping to remove the breakpoints
> when mmap comes in installs the breakpoint and returns.
> unregister_uprobe might erase the uprobe from rbtree after mmap is done.

Well yes, but that's mostly because of how you use those lists.

int __register_uprobe(...)
{
  uprobe = alloc_uprobe(...); // find or insert in tree

  vma_prio_tree_foreach(..) {
    // get mm ref, add to list blah blah
  }

  list_for_each_entry_safe() {
    // del from list etc..
    down_read(mm->mmap_sem);
    ret = install_breakpoint();
    if (ret && (ret != -ESRCH || ret != -EEXIST)) {
      up_read(..);
      goto fail;
  }

  return 0;

fail:
  list_for_each_entry_safe() {
    // del from list, put mm
  }

  return ret;
}

void __unregister_uprobe(...)
{
  uprobe = find_uprobe(); // ref++
  if (delete_consumer(...)); // includes tree removal on last consumer
                             // implies we own the last ref
     return; // consumers

  vma_prio_tree_foreach() {
     // create list
  }

  list_for_each_entry_safe() {
    // remove from list
    remove_breakpoint(); // unconditional, if it wasn't there
                         // its a nop anyway, can't get any new
                         // new probes on account of holding
                         // uprobes_mutex and mmap() doesn't see
                         // it due to tree removal.
  }
}

int register_uprobe(...)
{
  int ret;

  mutex_lock(&uprobes_mutex);
  ret = __register_uprobe(...);
  if (!ret)
    __unregister_uprobe(...);
  mutex_unlock(&uprobes_mutex);

  ret;
}

int mmap_uprobe(...)
{
  spin_lock(&uprobes_treelock);
  for_each_probe_in_inode() {
    // create list;
  }
  spin_unlock(..);

  list_for_each_entry_safe() {
    // remove from list
    ret = install_breakpoint();
    if (ret)
      goto fail;
    if (!uprobe_still_there()) // takes treelock
      remove_breakpoint();
  }

  return 0;

fail:
  list_for_each_entry_safe() {
    // destroy list
  }
  return ret;
}

Should work I think, no?