From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:34741) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UjyW4-0001At-1l for qemu-devel@nongnu.org; Tue, 04 Jun 2013 17:11:42 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UjyW2-0004cY-RI for qemu-devel@nongnu.org; Tue, 04 Jun 2013 17:11:39 -0400 Message-ID: <51AE5808.6000000@redhat.com> Date: Tue, 04 Jun 2013 15:11:36 -0600 From: Eric Blake MIME-Version: 1.0 References: <1370377419-31788-1-git-send-email-alevy@redhat.com> In-Reply-To: <1370377419-31788-1-git-send-email-alevy@redhat.com> Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="----enig2WMGCRMGRCIUORSDJFAJR" Subject: Re: [Qemu-devel] [PATCH 1/5] oslib-posix: add qemu_pipe_non_block List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Alon Levy Cc: qemu-trivial@nongnu.org, qemu-devel@nongnu.org This is an OpenPGP/MIME signed message (RFC 4880 and 3156) ------enig2WMGCRMGRCIUORSDJFAJR Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable On 06/04/2013 02:23 PM, Alon Levy wrote: > Used by the followin patch. s/followin/following/ > =20 > +int qemu_pipe_non_block(int pipefd[2]) > +{ > + int ret; > + > + ret =3D qemu_pipe(pipefd); qemu_pipe() already uses pipe2() when available; it seems like it would be nicer to use pipe2's O_NONBLOCK option directly in one syscall (where supported) instead of having to make additional syscalls after the fact. Would it just be smarter to change the signature of qemu_pipe() to add a bool block parameter, and then change the 5 existing callers to pass false with your later patch in the series passing true, and do it without creating a new wrapper? > + if (ret) { > + return ret; > + } > + if (fcntl(card->pipe[0], F_SETFL, O_NONBLOCK) =3D=3D -1) { > + return -errno; > + } > + if (fcntl(card->pipe[1], F_SETFL, O_NONBLOCK) =3D=3D -1) { > + return -errno; Leaks fds. If you're going to report error, then you must close the fds already created. > + } > + if (fcntl(card->pipe[0], F_SETOWN, getpid()) =3D=3D -1) { > + return -errno; Same comment about fd leaks. This part seems like a useful change, IF you plan on using SIGIO and SIGURG signals; and it is something which pipe2() cannot optimize, so I can see why you are adding a new function instead of changing qemu_pipe() and adjust all its callers to pass an additional parameter. But are you really planning on using SIGIO/SIGURG? Furthermore, this is undefined behavior. According to POSIX, use of F_SETOWN is only portable on sockets, not pipes. It may work on Linux, but you'll need to be aware of what it does on other platforms. http://pubs.opengroup.org/onlinepubs/9699919799/functions/fcntl.html --=20 Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org ------enig2WMGCRMGRCIUORSDJFAJR Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.13 (GNU/Linux) Comment: Public key at http://people.redhat.com/eblake/eblake.gpg Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBCAAGBQJRrlgIAAoJEKeha0olJ0Nqb4sIAJ4vj/nKYf2B65JPThJ7we56 Sfw1BxFH+bg5CyHgyqQ46nGeWH10t1lA+9VfFkTXlhD7N2dIPNEBhy7p5tO8Efj0 YDKcLsFPxGX1t/luCFzZdlcsit+C+Sz4lqdAmf9O16WemeByOpBPJnJFB3ec5nnU +5CqsLyfnThneJk8PqSSmhLmImzVSdJeLN5vC2Yni3Qkt1k84EWrteatpGKvPowt Ksp2Mi0nR42TymmnXbUn4kAHYDSsfR+KU0+c0st/K64qnjkivlX/+rscyk2wMd2H 6lXTbUKPACV6IOt406MZBaz7LWjnxz00wBaiFGlfPJxCK0sCuX4NOOxdbwfEtu4= =evgq -----END PGP SIGNATURE----- ------enig2WMGCRMGRCIUORSDJFAJR--