io-uring.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Will Deacon <will@kernel.org>
To: Jens Axboe <axboe@kernel.dk>
Cc: Kees Cook <keescook@chromium.org>, Jann Horn <jannh@google.com>,
	io-uring <io-uring@vger.kernel.org>,
	Kernel Hardening <kernel-hardening@lists.openwall.com>
Subject: Re: [PATCH 07/11] io_uring: use atomic_t for refcounts
Date: Wed, 11 Dec 2019 10:20:13 +0000	[thread overview]
Message-ID: <20191211102012.GA4123@willie-the-truck> (raw)
In-Reply-To: <d6ff9af3-5e72-329c-4aed-cbe6d9373235@kernel.dk>

On Tue, Dec 10, 2019 at 03:55:05PM -0700, Jens Axboe wrote:
> On 12/10/19 3:46 PM, Kees Cook wrote:
> > On Tue, Dec 10, 2019 at 03:21:04PM -0700, Jens Axboe wrote:
> >> On 12/10/19 3:04 PM, Jann Horn wrote:
> >>> [context preserved for additional CCs]
> >>>
> >>> On Tue, Dec 10, 2019 at 4:57 PM Jens Axboe <axboe@kernel.dk> wrote:
> >>>> Recently had a regression that turned out to be because
> >>>> CONFIG_REFCOUNT_FULL was set.
> >>>
> >>> I assume "regression" here refers to a performance regression? Do you
> >>> have more concrete numbers on this? Is one of the refcounting calls
> >>> particularly problematic compared to the others?
> >>
> >> Yes, a performance regression. io_uring is using io-wq now, which does
> >> an extra get/put on the work item to make it safe against async cancel.
> >> That get/put translates into a refcount_inc and refcount_dec per work
> >> item, and meant that we went from 0.5% refcount CPU in the test case to
> >> 1.5%. That's a pretty substantial increase.
> >>
> >>> I really don't like it when raw atomic_t is used for refcounting
> >>> purposes - not only because that gets rid of the overflow checks, but
> >>> also because it is less clear semantically.
> >>
> >> Not a huge fan either, but... It's hard to give up 1% of extra CPU. You
> >> could argue I could just turn off REFCOUNT_FULL, and I could. Maybe
> >> that's what I should do. But I'd prefer to just drop the refcount on the
> >> io_uring side and keep it on for other potential useful cases.
> > 
> > There is no CONFIG_REFCOUNT_FULL any more. Will Deacon's version came
> > out as nearly identical to the x86 asm version. Can you share the
> > workload where you saw this? We really don't want to regression refcount
> > protections, especially in the face of new APIs.
> > 
> > Will, do you have a moment to dig into this?
> 
> Ah, hopefully it'll work out ok, then. The patch came from testing the
> full backport on 5.2.
> 
> Do you have a link to the "nearly identical"? I can backport that
> patch and try on 5.2.

You could try my refcount/full branch, which is what ended up getting merged
during the merge window:

https://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git/log/?h=refcount/full

Will

  reply	other threads:[~2019-12-11 10:20 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-12-10 15:57 [PATCHSET 0/11] io_uring improvements/fixes for 5.5-rc Jens Axboe
2019-12-10 15:57 ` [PATCH 01/11] io_uring: allow unbreakable links Jens Axboe
2019-12-10 21:10   ` Pavel Begunkov
2019-12-10 21:12     ` Jens Axboe
2019-12-10 21:28       ` Pavel Begunkov
2019-12-10 22:17         ` Jens Axboe
2019-12-10 15:57 ` [PATCH 02/11] io-wq: remove worker->wait waitqueue Jens Axboe
2019-12-10 15:57 ` [PATCH 03/11] io-wq: briefly spin for new work after finishing work Jens Axboe
2019-12-10 15:57 ` [PATCH 04/11] io_uring: sqthread should grab ctx->uring_lock for submissions Jens Axboe
2019-12-10 15:57 ` [PATCH 05/11] io_uring: deferred send/recvmsg should assign iov Jens Axboe
2019-12-10 15:57 ` [PATCH 06/11] io_uring: don't dynamically allocate poll data Jens Axboe
2019-12-10 15:57 ` [PATCH 07/11] io_uring: use atomic_t for refcounts Jens Axboe
2019-12-10 22:04   ` Jann Horn
2019-12-10 22:21     ` Jens Axboe
2019-12-10 22:46       ` Kees Cook
2019-12-10 22:55         ` Jens Axboe
2019-12-11 10:20           ` Will Deacon [this message]
2019-12-11 16:56             ` Kees Cook
2019-12-11 17:00               ` Jens Axboe
2019-12-10 15:57 ` [PATCH 08/11] io_uring: run next sqe inline if possible Jens Axboe
2019-12-10 15:57 ` [PATCH 09/11] io_uring: only hash regular files for async work execution Jens Axboe
2019-12-10 15:57 ` [PATCH 10/11] net: make socket read/write_iter() honor IOCB_NOWAIT Jens Axboe
2019-12-10 19:37   ` David Miller
2019-12-10 20:43     ` Jens Axboe
2019-12-10 15:57 ` [PATCH 11/11] io_uring: add sockets to list of files that support non-blocking issue Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191211102012.GA4123@willie-the-truck \
    --to=will@kernel.org \
    --cc=axboe@kernel.dk \
    --cc=io-uring@vger.kernel.org \
    --cc=jannh@google.com \
    --cc=keescook@chromium.org \
    --cc=kernel-hardening@lists.openwall.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).