From owner-freebsd-net@FreeBSD.ORG Thu Jul 8 09:43:10 2010 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 6B0941065674; Thu, 8 Jul 2010 09:43:10 +0000 (UTC) (envelope-from kostikbel@gmail.com) Received: from mail.zoral.com.ua (mx0.zoral.com.ua [91.193.166.200]) by mx1.freebsd.org (Postfix) with ESMTP id DB42F8FC19; Thu, 8 Jul 2010 09:43:09 +0000 (UTC) Received: from deviant.kiev.zoral.com.ua (root@deviant.kiev.zoral.com.ua [10.1.1.148]) by mail.zoral.com.ua (8.14.2/8.14.2) with ESMTP id o689gum8019574 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 8 Jul 2010 12:42:56 +0300 (EEST) (envelope-from kostikbel@gmail.com) Received: from deviant.kiev.zoral.com.ua (kostik@localhost [127.0.0.1]) by deviant.kiev.zoral.com.ua (8.14.4/8.14.4) with ESMTP id o689guxg023776; Thu, 8 Jul 2010 12:42:56 +0300 (EEST) (envelope-from kostikbel@gmail.com) Received: (from kostik@localhost) by deviant.kiev.zoral.com.ua (8.14.4/8.14.4/Submit) id o689gubb023775; Thu, 8 Jul 2010 12:42:56 +0300 (EEST) (envelope-from kostikbel@gmail.com) X-Authentication-Warning: deviant.kiev.zoral.com.ua: kostik set sender to kostikbel@gmail.com using -f Date: Thu, 8 Jul 2010 12:42:56 +0300 From: Kostik Belousov To: Andriy Gapon Message-ID: <20100708094255.GA2408@deviant.kiev.zoral.com.ua> References: <7C3D15DD6E8F464998CA1470D8A322F302BB9F72@ES02CO.wgti.net> <20100707205041.GO13238@deviant.kiev.zoral.com.ua> <4C358843.5000001@freebsd.org> <4C358A01.8080206@icyb.net.ua> <20100708082943.GB2439@deviant.kiev.zoral.com.ua> <4C358EE5.8070003@icyb.net.ua> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="UugvWAfsgieZRqgk" Content-Disposition: inline In-Reply-To: <4C358EE5.8070003@icyb.net.ua> User-Agent: Mutt/1.4.2.3i X-Virus-Scanned: clamav-milter 0.95.2 at skuns.kiev.zoral.com.ua X-Virus-Status: Clean X-Spam-Status: No, score=-2.3 required=5.0 tests=ALL_TRUSTED,AWL,BAYES_50, DNS_FROM_OPENWHOIS autolearn=no version=3.2.5 X-Spam-Checker-Version: SpamAssassin 3.2.5 (2008-06-10) on skuns.kiev.zoral.com.ua Cc: lstewart@freebsd.org, freebsd-net@freebsd.org, bz@freebsd.org, Andre Oppermann , Ming Fu Subject: Re: kern/123095 kern/131602 sendfile X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 08 Jul 2010 09:43:10 -0000 --UugvWAfsgieZRqgk Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Thu, Jul 08, 2010 at 11:40:05AM +0300, Andriy Gapon wrote: > on 08/07/2010 11:29 Kostik Belousov said the following: > > Right, the patch maps the page in sf buffer read-only (on i386 only). > > But note the parallel posting with m_cat() change. It is still not enou= gh, > > and I am not set up for the real network testing ATM. >=20 > Could you also try to experiment with mb_dupcl? > Namely transfer M_RDONLY from source mbuf. Right, it is it. Below is my current patch including debugging facilities that seems to work. Real changes that needed are in m_cat and mb_dupcl. diff --git a/sys/i386/i386/vm_machdep.c b/sys/i386/i386/vm_machdep.c index 6f86895..530f704 100644 --- a/sys/i386/i386/vm_machdep.c +++ b/sys/i386/i386/vm_machdep.c @@ -810,6 +810,12 @@ sf_buf_alloc(struct vm_page *m, int flags) nsfbufsused++; nsfbufspeak =3D imax(nsfbufspeak, nsfbufsused); } + if ((flags & SFB_RONLY) =3D=3D 0) { + ptep =3D vtopte(sf->kva); + opte =3D *ptep; + if ((opte & PG_RW) =3D=3D 0) + *ptep |=3D PG_RW | PG_A; + } #ifdef SMP goto shootdown;=09 #else @@ -854,8 +860,9 @@ sf_buf_alloc(struct vm_page *m, int flags) PT_SET_MA(sf->kva, xpmap_ptom(VM_PAGE_TO_PHYS(m)) | pgeflag | PG_RW | PG_V | pmap_cache_bits(m->md.pat_mode, 0)); #else - *ptep =3D VM_PAGE_TO_PHYS(m) | pgeflag | PG_RW | PG_V | - pmap_cache_bits(m->md.pat_mode, 0); + *ptep =3D VM_PAGE_TO_PHYS(m) | pgeflag | + ((flags & SFB_RONLY) ? 0 : PG_RW) | PG_V | + pmap_cache_bits(m->md.pat_mode, 0); #endif =20 /* diff --git a/sys/kern/uipc_mbuf.c b/sys/kern/uipc_mbuf.c index f41eb03..1701ef2 100644 --- a/sys/kern/uipc_mbuf.c +++ b/sys/kern/uipc_mbuf.c @@ -301,7 +301,7 @@ mb_dupcl(struct mbuf *n, struct mbuf *m) n->m_ext.ext_size =3D m->m_ext.ext_size; n->m_ext.ref_cnt =3D m->m_ext.ref_cnt; n->m_ext.ext_type =3D m->m_ext.ext_type; - n->m_flags |=3D M_EXT; + n->m_flags |=3D M_EXT | (M_RDONLY & m->m_flags); } =20 /* @@ -911,7 +911,8 @@ m_cat(struct mbuf *m, struct mbuf *n) m =3D m->m_next; while (n) { if (m->m_flags & M_EXT || - m->m_data + m->m_len + n->m_len >=3D &m->m_dat[MLEN]) { + m->m_data + m->m_len + n->m_len >=3D &m->m_dat[MLEN] || + !M_WRITABLE(m)) { /* just join the two chains */ m->m_next =3D n; return; diff --git a/sys/kern/uipc_sockbuf.c b/sys/kern/uipc_sockbuf.c index 2060a2e..95fae7e 100644 --- a/sys/kern/uipc_sockbuf.c +++ b/sys/kern/uipc_sockbuf.c @@ -776,6 +776,8 @@ sbcompress(struct sockbuf *sb, struct mbuf *m, struct m= buf *n) m->m_len <=3D MCLBYTES / 4 && /* XXX: Don't copy too much */ m->m_len <=3D M_TRAILINGSPACE(n) && n->m_type =3D=3D m->m_type) { + KASSERT(n->m_ext.ext_type !=3D EXT_SFBUF, + ("destroying file page %p", n)); bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len, (unsigned)m->m_len); n->m_len +=3D m->m_len; diff --git a/sys/kern/uipc_syscalls.c b/sys/kern/uipc_syscalls.c index 3165dab..df9cd69 100644 --- a/sys/kern/uipc_syscalls.c +++ b/sys/kern/uipc_syscalls.c @@ -2131,7 +2131,8 @@ retry_space: * as necessary, but this wait can be interrupted. */ if ((sf =3D sf_buf_alloc(pg, - (mnw ? SFB_NOWAIT : SFB_CATCH))) =3D=3D NULL) { + (mnw ? SFB_NOWAIT : SFB_CATCH) | SFB_RONLY)) + =3D=3D NULL) { mbstat.sf_allocfail++; vm_page_lock(pg); vm_page_unwire(pg, 0); @@ -2162,6 +2163,8 @@ retry_space: m_cat(m, m0); else m =3D m0; + KASSERT((m0->m_flags & M_RDONLY) !=3D 0, + ("lost M_RDONLY")); =20 /* Keep track of bits processed. */ loopbytes +=3D xfsize; diff --git a/sys/sys/sf_buf.h b/sys/sys/sf_buf.h index af42065..fcb31f8 100644 --- a/sys/sys/sf_buf.h +++ b/sys/sys/sf_buf.h @@ -41,6 +41,7 @@ #define SFB_CPUPRIVATE 2 /* Create a CPU private mapping. */ #define SFB_DEFAULT 0 #define SFB_NOWAIT 4 /* Return NULL if all bufs are used. */ +#define SFB_RONLY 8 /* Map page read-only, if possible. */ =20 struct vm_page; =20 --UugvWAfsgieZRqgk Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (FreeBSD) iEYEARECAAYFAkw1nZ8ACgkQC3+MBN1Mb4iLtQCghSEGzaeG5KqxQ5bAOOOmoX4d jhwAn294ck0E3Or4qCsE19dOtfRTLq3w =U6Au -----END PGP SIGNATURE----- --UugvWAfsgieZRqgk--