Re: [PATCH v2 2.6.38-rc8-tip 4/20] 4: uprobes: Adding and remove auprobe in a rb tree.

[Thomas Gleixner]

On Tue, 15 Mar 2011, Srikar Dronamraju wrote:
> * Thomas Gleixner <tglx@xxxxxxxxxxxxx> [2011-03-15 14:38:33]:
> > > +/*
> > > + * Find a uprobe corresponding to a given inode:offset
> > > + * Acquires treelock
> > > + */
> > > +static struct uprobe *find_uprobe(struct inode * inode, loff_t offset)
> > > +{
> > > +	struct uprobe *uprobe;
> > > +	unsigned long flags;
> > > +
> > > +	spin_lock_irqsave(&treelock, flags);
> > > +	uprobe = __find_uprobe(inode, offset, NULL);
> > > +	spin_unlock_irqrestore(&treelock, flags);
> > 
> > What's the calling context ? Do we really need a spinlock here for
> > walking the rb tree ?
> > 
> 
> find_uprobe() gets called from unregister_uprobe and on probe hit from
> uprobe_notify_resume. I am not sure if its a good idea to walk the tree
> as and when the tree is changing either because of a insertion or
> deletion of a probe.

I know that you cannot walk the tree lockless except you would use
some rcu based container for your probes.

Though my question is more whether this needs to be a spinlock or if
that could be replaced by a mutex. At least there is no reason to
disable interrupts. You cannot trap into a probe from the thread in
which you are installing/removing it.

Thanks,

	tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/