On Sat, Aug 03, 2019 at 09:14:00PM -0300, Jason Gunthorpe wrote:
On Sat, Aug 03, 2019 at 05:36:13PM -0400, Michael S. Tsirkin wrote:I frankly don't know. With SMAP you flip flags twice, and with spectre
On Fri, Aug 02, 2019 at 02:24:18PM -0300, Jason Gunthorpe wrote:You think copy_to_user will be more expensive than the minimum two
On Fri, Aug 02, 2019 at 10:27:21AM -0400, Michael S. Tsirkin wrote:Ever since copy_from_user started playing with flags (for SMAP) and
On Fri, Aug 02, 2019 at 09:46:13AM -0300, Jason Gunthorpe wrote:Yikes, I'm not sure you can ever win against copy_from_user using
On Fri, Aug 02, 2019 at 05:40:07PM +0800, Jason Wang wrote:The topic is whether we should revert
I've also idly wondered if calling synchronize_rcu() under the variousThis must be a proper barrier, like a spinlock, mutex, or
I start with synchronize_rcu() but both you and Michael raise some
mm locks is a deadlock situation.
Then I try spinlock and mutex:I think the topic here is correctness not performance improvement
1) spinlock: add lots of overhead on datapath, this leads 0 performance
commit 7f466032dc9 ("vhost: access vq metadata through kernel virtual address")
or keep it in. The only reason to keep it is performance.
added speculation barriers there's a chance we can win by accessing
memory through the kernel address.
atomics required to synchronize with another thread?
you flush the pipeline. Is that cheaper or more expensive than an atomic
operation? Testing is the only way to tell.
Sorry I misunderstood the question. Permanent GUP breaks lots ofYou didn't answer this.. Why not just use GUP?Also, why can't this just permanently GUP the pages? In fact, where
does it put_page them anyhow? Worrying that 7f466 adds a get_user page
but does not add a put_page??
functionality we need such as THP and numa balancing.
release_pages is used instead of put_page.