[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190726150322.GB8695@ziepe.ca>
Date: Fri, 26 Jul 2019 12:03:22 -0300
From: Jason Gunthorpe <jgg@...pe.ca>
To: Jason Wang <jasowang@...hat.com>
Cc: "Michael S. Tsirkin" <mst@...hat.com>,
syzbot <syzbot+e58112d71f77113ddb7b@...kaller.appspotmail.com>,
aarcange@...hat.com, akpm@...ux-foundation.org,
christian@...uner.io, davem@...emloft.net, ebiederm@...ssion.com,
elena.reshetova@...el.com, guro@...com, hch@...radead.org,
james.bottomley@...senpartnership.com, jglisse@...hat.com,
keescook@...omium.org, ldv@...linux.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
linux-mm@...ck.org, linux-parisc@...r.kernel.org,
luto@...capital.net, mhocko@...e.com, mingo@...nel.org,
namit@...are.com, peterz@...radead.org,
syzkaller-bugs@...glegroups.com, viro@...iv.linux.org.uk,
wad@...omium.org
Subject: Re: WARNING in __mmdrop
On Fri, Jul 26, 2019 at 10:00:20PM +0800, Jason Wang wrote:
> The question is, MMU notifier are allowed to be blocked on
> invalidate_range_start() which could be much slower than synchronize_rcu()
> to finish.
>
> Looking at amdgpu_mn_invalidate_range_start_gfx() which calls
> amdgpu_mn_invalidate_node() which did:
>
> r = reservation_object_wait_timeout_rcu(bo->tbo.resv,
> true, false, MAX_SCHEDULE_TIMEOUT);
>
> ...
The general guidance has been that invalidate_start should block
minimally, if at all.
I would say synchronize_rcu is outside that guidance.
BTW, always returning EAGAIN for mmu_notifier_range_blockable() is not
good either, it should instead only return EAGAIN if any
vhost_map_range_overlap() is true.
Jason
Powered by blists - more mailing lists