linux-kernel - Re: [PATCH 03/10] locking: bring back lglocks

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20180518113925.GB16943@kmo-pixel>
Date:   Fri, 18 May 2018 07:39:25 -0400
From:   Kent Overstreet <kent.overstreet@...il.com>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Dave Chinner <dchinner@...hat.com>, darrick.wong@...cle.com,
        tytso@....edu, linux-btrfs@...r.kernel.org, clm@...com,
        jbacik@...com, viro@...iv.linux.org.uk, willy@...radead.org
Subject: Re: [PATCH 03/10] locking: bring back lglocks

On Fri, May 18, 2018 at 01:03:39PM +0200, Peter Zijlstra wrote:
> On Fri, May 18, 2018 at 06:13:53AM -0400, Kent Overstreet wrote:
> > On Fri, May 18, 2018 at 11:51:02AM +0200, Peter Zijlstra wrote:
> > > On Fri, May 18, 2018 at 03:49:04AM -0400, Kent Overstreet wrote:
> > > > bcachefs makes use of them - also, add a proper lg_lock_init()
> > > 
> > > Why?! lglocks are horrid things, we got rid of them for a reason. They
> > > have terrifying worst case preemption off latencies.
> > 
> > Ah. That was missing from your commit message.
> 
> Yeah, sorry, sometimes it's hard to state what is obvious to oneself :/
> 
> > > Why can't you use something like per-cpu rwsems?
> > 
> > Well,
> > 
> >  a) in my use case, lg_global_lock() pretty much isn't used in normal operation,
> >     it's only called when starting mark and sweep gc (which is not needed
> >     anymore and disabled by default, it'll eventually get rolled into online
> >     fsck) and for device resize
> > 
> >  b) I'm using it in conjection with percpu counters, and technically yes I
> >     certainly _could_ use per-cpu sleepable locks (mutexes would make more sense
> >     for me than rwsems), there's a bit of a clash there and it's going to be a
> >     bit ugly and messy and it's more work for me. (this_cpu_ptr() no longer
> >     makes any sense in that case, so it'd mean auditing and converting all the
> >     code that touches the relevant data structures).
> 
> Well, lg is a reader-writer style lock per definition, as you want
> concurrency on the local and full exclusion against the global, so I'm
> not sure how mutexes fit into this.
> 
> In any case, have a look at percpu_down_read_preempt_disable() and
> percpu_up_read_preempt_enable(); they're a bit of a hack but they should
> work for you I think.
> 
> They will sleep at down_read, but the entire actual critical section
> will be with preemption disabled -- therefore it had better be short and
> bounded, and the RT guys will thank you for not using spinlock_t under
> it (use raw_spinlock_t if you have to).
> 
> The (global) writer side will block and be preemptible like normal.
> 
> > If you're really that dead set against lglocks I might just come up with a new
> > lock with similar semantics, that doesn't break this_cpu_ptr() but sleeps if the
> > global lock is held...
> 
> See above, we already have this ;-)

Ok, I think this might work. I'll have to stare awhile and make sure I remember
everything I'm currently depending on the lglock for...