Date: Fri, 27 Jan 2017 15:37:57 +0100 From: Julien Charbon <jch@freebsd.org> To: Gleb Smirnoff <glebius@FreeBSD.org>, hiren@FreeBSD.org, Jason Eggleston <jeggleston@llnw.com> Cc: jtl@FreeBSD.org, rrs@FreeBSD.org, net@FreeBSD.org Subject: Re: listening sockets as non sockets Message-ID: <84f9c348-1e65-1b94-37a6-a4c67d195709@freebsd.org> In-Reply-To: <20170127005251.GM2611@FreeBSD.org> References: <20170127005251.GM2611@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --mc4llJNOAcSIKKqbI6nk76vuCmWPKUGma Content-Type: multipart/mixed; boundary="Bia7XafaO9k9cIeGEGeH9sxqELVnt3SLP"; protected-headers="v1" From: Julien Charbon <jch@freebsd.org> To: Gleb Smirnoff <glebius@FreeBSD.org>, hiren@FreeBSD.org, Jason Eggleston <jeggleston@llnw.com> Cc: jtl@FreeBSD.org, rrs@FreeBSD.org, net@FreeBSD.org Message-ID: <84f9c348-1e65-1b94-37a6-a4c67d195709@freebsd.org> Subject: Re: listening sockets as non sockets References: <20170127005251.GM2611@FreeBSD.org> In-Reply-To: <20170127005251.GM2611@FreeBSD.org> --Bia7XafaO9k9cIeGEGeH9sxqELVnt3SLP Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: quoted-printable Hi Gleb, On 1/27/17 1:52 AM, Gleb Smirnoff wrote: > as some of you already heard, I'm trying to separate listening socket= s > into a new file descriptor type. If we look into current struct socket,= > we see that some functional fields belong to normal data flow sockets, > and other belong to listening socket. They are never used simultaneousl= y. > Now, if we look at socket API, we see that once a socket underwent tran= sformation > to a listening socket, only 3 regular syscalls now may be called: liste= n(2), > accept(2) and close(2) and a subset of ioctl() and setsockopt() paramet= ers is > accepted. A listening socket cannot be closed from the protocol side, o= nly from > user side. So, listening socket is so different from a dataflow socket,= that > separating them looks architecturally right thing to do. >=20 > The benefits are: >=20 > 1) Nicer code (I hope). > 2) Smaller 'struct socket'. > 3) Having two different locks for socket and solisten, we can try to ge= t rid > of ACCEPT_LOCK global lock. >=20 > The patch is in a very pre-alpha state. It has been run only in my bhyv= e VM. >=20 > It passes regression tests from tools/regression/sockets and tests/sys,= > including the race tests, and including accept filter ones. >=20 > For TCP it passes basic functionality testing, but likely there are sti= ll races > remaining after ACCEPT_LOCK removal. > <snip> > I've put current snapshot to Phab, so that you can view it there. The s= nap > patch is also attached to this email. >=20 > https://reviews.freebsd.org/D9356 >=20 > At this moment I'd like to start doing some testing (and doing polishin= g > in parallel), and here I seek for your help. Those, who run FreeBSD at > very high connection rates and observe contention on the accept global > mutex, anybody willing to collaborate with me on this? Good idea, we are (obviously) interested by point #3 (Get rid of ACCEPT global lock) and its connection rate scalability improvement. I might be able to look at potential race conditions related to ACCEPT_LOCK and SO_ACCEPTFILTER usage (even if I am more used to INP_INFO lock), but I can certainly provide performance numbers and lock contention metrics using our setup. Do you think it is the right time to start performance testing with your change? Or it is a bit premature? -- Julien --Bia7XafaO9k9cIeGEGeH9sxqELVnt3SLP-- --mc4llJNOAcSIKKqbI6nk76vuCmWPKUGma Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Comment: GPGTools - https://gpgtools.org iQEcBAEBCgAGBQJYi1tLAAoJEKVlQ5Je6dhxK6cH/2pOgTzT9i1b1nvBbZi82p3D ashJsxwEk/2gLcpDNlwrUL7K7Y/qea0Ah//isO47ZHc3qdZOZm7t9D1mr/hmoS6e BJteTgYvvjhOiCradH8lB1LG6kNg+6vQsjsJ/AovepAonRrKnQl2dj6XgFpKDcHi jCqcBW768lIl3FaW/geVwtACXqAeHMIXvg5ikeVM3MtG22H1mTmumMG8GsGu6QYs zWxXdfL3CFAvyFArfUD+Gv2ng5sItCL3gDvo7KeolpHI0nwL297f4BmYkeuks773 4aAogg9Y3e5GkrxF1l56YA3EGg4mwOvLwhjb3OB0m4lW4R2B0Fx1jTFa76np/Ag= =1tCk -----END PGP SIGNATURE----- --mc4llJNOAcSIKKqbI6nk76vuCmWPKUGma--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?84f9c348-1e65-1b94-37a6-a4c67d195709>