Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-27 Thread Mike Marshall
Hi... sorry if I got the attribution wrong... the commit I studied was: commit 76d8aeabfeb1c42641a81c44280177b9a08670d8 Author: David Howells Date: Thu Jun 23 22:00:49 2005 -0700 It was huge ... -Mike On Mon, Feb 27, 2017 at 3:44 AM, David Howells wrote: > Mike Marshall wrote: > >> Using

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-27 Thread David Howells
Mike Marshall wrote: > Using the git log, I searched back to where it seems to me call_rcu was > added, a giant patch from 2005 by David Howells which includes tons of > source and a large amount of documentation. I'm pretty sure Paul McKenney added call_rcu(), not me. David

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-26 Thread Linus Torvalds
On Sun, Feb 26, 2017 at 4:34 PM, Mike Marshall wrote: > Since Orangefs uses ref-walk, not rcu-walk, this patch with call_rcu > has seemed weird to me. Even if orangefs never really allows RCU walking, the VFS layer will look up dentries - and look at their inodes - from RCU. It will then call int

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-26 Thread Mike Marshall
Since Orangefs uses ref-walk, not rcu-walk, this patch with call_rcu has seemed weird to me. Using the git log, I searched back to where it seems to me call_rcu was added, a giant patch from 2005 by David Howells which includes tons of source and a large amount of documentation. It seems that the

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-25 Thread Mike Marshall
After looking through the code and seeing how some other filesystems use call_rcu, it seems that call_rcu has to do with consistency and waiting for stuff to complete before returning an object to the slab cache, whereas we were just calling kmem_cache_free without worrying about that kind of stuff

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-24 Thread Mike Marshall
Thanks Al... I was going to try and evaluate that patch next week, now all I have to do is test it ... -Mike On Fri, Feb 24, 2017 at 3:52 PM, Al Viro wrote: > That, AFAICS, fixes a real bug. Applied, and it needs Cc:stable as well. > > >> Signed-off-by: Peter Zijlstra (Intel) >> --- >> fs/or

Re: [RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-24 Thread Al Viro
That, AFAICS, fixes a real bug. Applied, and it needs Cc:stable as well. > Signed-off-by: Peter Zijlstra (Intel) > --- > fs/orangefs/super.c |9 - > 1 file changed, 8 insertions(+), 1 deletion(-) > > --- a/fs/orangefs/super.c > +++ b/fs/orangefs/super.c > @@ -115,6 +115,13 @@ stat

[RFC][PATCH 07/10] orangefs: Use RCU for destroy_inode

2017-02-24 Thread Peter Zijlstra
Signed-off-by: Peter Zijlstra (Intel) --- fs/orangefs/super.c |9 - 1 file changed, 8 insertions(+), 1 deletion(-) --- a/fs/orangefs/super.c +++ b/fs/orangefs/super.c @@ -115,6 +115,13 @@ static struct inode *orangefs_alloc_inod return &orangefs_inode->vfs_inode; } +static