Re: [PATCH] anobjrmap 9 priority mjb tree

[Rajesh Venkatasubramanian]

> Unless we see a plausible way forward on your SDET numbers, I
> think it casts this project in doubt - but even so I do need

We can try a few fancy locking tricks. But, we don't know whether
such tricks will help.

> i_shared_lock changed to i_shared_sem to allow that cond_resched_lock
> in unmap_vmas to solve vmtruncate latency problems?  With i_mmap and
> i_mmap_shared as lists, isn't it easy to insert a dummy marker vma
> and drop the lock if we need resched?  Resuming from marker after.
>
> But, sadly, I doubt that can be done with the prio tree: Rajesh?

Yeap. With prio_tree it is tricky. We already have the marker for
prio_tree, i.e., prio_tree_iter. But, when you drop a lock new tree
nodes may be added to the prio_tree, and the marker does not provide
any consistent meaning after the node additions.

Rajesh

-
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/