From owner-svn-src-head@FreeBSD.ORG Thu May 14 10:04:43 2015 Return-Path: Delivered-To: svn-src-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id E7F52E8F; Thu, 14 May 2015 10:04:42 +0000 (UTC) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id D2AB81BC6; Thu, 14 May 2015 10:04:42 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.9/8.14.9) with ESMTP id t4EA4gS3099874; Thu, 14 May 2015 10:04:42 GMT (envelope-from mav@FreeBSD.org) Received: (from mav@localhost) by svn.freebsd.org (8.14.9/8.14.9/Submit) id t4EA4g3P099873; Thu, 14 May 2015 10:04:42 GMT (envelope-from mav@FreeBSD.org) Message-Id: <201505141004.t4EA4g3P099873@svn.freebsd.org> X-Authentication-Warning: svn.freebsd.org: mav set sender to mav@FreeBSD.org using -f From: Alexander Motin Date: Thu, 14 May 2015 10:04:42 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r282881 - head/sys/fs/nfsclient X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 14 May 2015 10:04:43 -0000 Author: mav Date: Thu May 14 10:04:42 2015 New Revision: 282881 URL: https://svnweb.freebsd.org/changeset/base/282881 Log: Do not promote large async writes to sync. Present implementation of large sync writes is too strict and so can be quite slow. Instead of doing that, execute large async write in chunks, syncing each chunk separately. It would be good to fix large sync writes too, but I leave it to somebody with more skills in this area. Reviewed by: rmacklem MFC after: 1 week Modified: head/sys/fs/nfsclient/nfs_clbio.c Modified: head/sys/fs/nfsclient/nfs_clbio.c ============================================================================== --- head/sys/fs/nfsclient/nfs_clbio.c Thu May 14 08:13:01 2015 (r282880) +++ head/sys/fs/nfsclient/nfs_clbio.c Thu May 14 10:04:42 2015 (r282881) @@ -864,7 +864,7 @@ ncl_write(struct vop_write_args *ap) struct nfsmount *nmp = VFSTONFS(vp->v_mount); daddr_t lbn; int bcount, noncontig_write, obcount; - int bp_cached, n, on, error = 0, error1; + int bp_cached, n, on, error = 0, error1, wouldcommit; size_t orig_resid, local_resid; off_t orig_size, tmp_off; @@ -908,7 +908,6 @@ ncl_write(struct vop_write_args *ap) if (ioflag & IO_NDELAY) return (EAGAIN); #endif -flush_and_restart: np->n_attrstamp = 0; KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp); error = ncl_vinvalbuf(vp, V_SAVE, td, 1); @@ -965,27 +964,14 @@ flush_and_restart: * IO_UNIT -- we just make all writes atomic anyway, as there's * no point optimizing for something that really won't ever happen. */ + wouldcommit = 0; if (!(ioflag & IO_SYNC)) { int nflag; mtx_lock(&np->n_mtx); nflag = np->n_flag; mtx_unlock(&np->n_mtx); - int needrestart = 0; - if (nmp->nm_wcommitsize < uio->uio_resid) { - /* - * If this request could not possibly be completed - * without exceeding the maximum outstanding write - * commit size, see if we can convert it into a - * synchronous write operation. - */ - if (ioflag & IO_NDELAY) - return (EAGAIN); - ioflag |= IO_SYNC; - if (nflag & NMODIFIED) - needrestart = 1; - } else if (nflag & NMODIFIED) { - int wouldcommit = 0; + if (nflag & NMODIFIED) { BO_LOCK(&vp->v_bufobj); if (vp->v_bufobj.bo_dirty.bv_cnt != 0) { TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, @@ -995,27 +981,22 @@ flush_and_restart: } } BO_UNLOCK(&vp->v_bufobj); - /* - * Since we're not operating synchronously and - * bypassing the buffer cache, we are in a commit - * and holding all of these buffers whether - * transmitted or not. If not limited, this - * will lead to the buffer cache deadlocking, - * as no one else can flush our uncommitted buffers. - */ - wouldcommit += uio->uio_resid; - /* - * If we would initially exceed the maximum - * outstanding write commit size, flush and restart. - */ - if (wouldcommit > nmp->nm_wcommitsize) - needrestart = 1; } - if (needrestart) - goto flush_and_restart; } do { + if (!(ioflag & IO_SYNC)) { + wouldcommit += biosize; + if (wouldcommit > nmp->nm_wcommitsize) { + np->n_attrstamp = 0; + KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp); + error = ncl_vinvalbuf(vp, V_SAVE, td, 1); + if (error) + return (error); + wouldcommit = biosize; + } + } + NFSINCRGLOBAL(newnfsstats.biocache_writes); lbn = uio->uio_offset / biosize; on = uio->uio_offset - (lbn * biosize);