DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org C4D512146D
MIME-Version: 1.0
References: <20181115083817.20986-1-gbenhaim@augury.com> <CABBYNZJ_TEM18yfO6O4HFdyo7sm-cE8iL4Y8_DEa1e6xZ7a=Bg@mail.gmail.com>
 <CAHotPr-jS457EKX=7dvcwA6eB12zr7d2dZG5Prjh6+fMdAT=+Q@mail.gmail.com>
 <9007400A-D10D-4906-AB75-6D6395AF98FD@holtmann.org> <CAHotPr9Jb3Prqk_tkaeTRsJk0VgsGWbj8tsSdEp-Q9mb9Dt=-A@mail.gmail.com>
 <CAHotPr-BO=RE6HZYWsC1+MT9oSZ58Xf-8XLz53BJq0Z6Em9Hsw@mail.gmail.com> <CAO1O6seRNsBvz_sdZHKZhSYc3wbeZPOf-_0qXEG6PAnom0qcvA@mail.gmail.com>
In-Reply-To: <CAO1O6seRNsBvz_sdZHKZhSYc3wbeZPOf-_0qXEG6PAnom0qcvA@mail.gmail.com>
From:   Gal Ben Haim <gbenhaim@augury.com>
Date:   Thu, 15 Nov 2018 17:43:32 +0200
Message-ID: <CAHotPr_3F-wCDyKryYa-K_=tmLP7cwRDqhbzWNDJ2P7EmYxRVg@mail.gmail.com>
Subject: Re: [PATCH BlueZ] gatt: Ignore SIGPIPE
To:     emil.lenngren@gmail.com
Cc:     Marcel Holtmann <marcel@holtmann.org>,
        Luiz Augusto von Dentz <luiz.dentz@gmail.com>,
        linux-bluetooth@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Sender: linux-bluetooth-owner@vger.kernel.org
Precedence: bulk

should it be added to setup_signalfd in main.c ?

On Thu, Nov 15, 2018 at 2:56 PM Emil Lenngren <emil.lenngren@gmail.com> wrote:
>
> Den tors 15 nov. 2018 kl 12:55 skrev Gal Ben Haim <gbenhaim@augury.com>:
> >
> > I submitted a different patch -
> > https://marc.info/?l=linux-bluetooth&m=154228281730667&w=2
> > On Thu, Nov 15, 2018 at 12:20 PM Gal Ben Haim <gbenhaim@augury.com> wrote:
> > >
> > > I can try to create another patch for it
> > >
> > > On Thu, Nov 15, 2018 at 12:18 PM Marcel Holtmann <marcel@holtmann.org> wrote:
> > > >
> > > > Hi Gal,
> > > >
> > > > > io_send can use poll and check for POLLOUT and POLLRDHUP before
> > > > > writing to the pipe
> > > >
> > > > that should be done anyway since code that writes to a pipe descriptor (or socket for that matter) without checking that it is still valid is broken code.
> > > >
> > > > Regards
> > > >
> > > > Marcel
> > > >
>
> Even if you first check for writability with poll, you still need to
> also ignore the SIGPIPE signal, since the connection could be reset
> after the check is made but before the write is executed, I guess. But
> the "signal(SIGPIPE, SIG_IGN)" should usually be made once at
> application initialization.
>
> /Emil