[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CANaxB-y6_OF4GMKUC=GssEY7Q-sH8y23F3wP2XT31=izSUvHSw@mail.gmail.com>
Date: Thu, 5 Mar 2020 11:54:01 -0800
From: Andrei Vagin <avagin@...il.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>
Cc: Matthew Wilcox <willy@...radead.org>,
LKML <linux-kernel@...r.kernel.org>,
Sasha Levin <sashal@...nel.org>,
stable <stable@...r.kernel.org>,
Josh Triplett <josh@...htriplett.org>,
linux-fsdevel <linux-fsdevel@...r.kernel.org>
Subject: Re: [PATCH AUTOSEL 5.5 542/542] pipe: use exclusive waits when
reading or writing
On Thu, Mar 5, 2020 at 10:41 AM Linus Torvalds
<torvalds@...ux-foundation.org> wrote:
>
> On Thu, Mar 5, 2020 at 12:20 PM Andrei Vagin <avagin@...il.com> wrote:
> >
> > After this change, one more criu test became flaky. This is due to one
> > of corner cases, so I am not sure that we need to fix something in the
> > kernel. I have fixed this issue in the test. I am not sure that this
> > will affect any real applications.
>
> It's an interesting test-case, but it's really not doing anything you
> should rely on.
I'm agree with this.
> But if CRIU itself depends on this behavior (rather than just a test),
> then I guess we need to.
>
> So is it just a test-case, or does CRIU itself depend on that "reads
> get full buffers"? As mentioned, that really _is_ fundamentally broken
> if there is any chance of signals..
No, it doesn't. I'm agree that we can wait an report from a real app.
Thanks!
Andrei
Powered by blists - more mailing lists