From owner-svn-src-vendor@FreeBSD.ORG Tue Nov 26 20:09:44 2013 Return-Path: Delivered-To: svn-src-vendor@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id B7324325; Tue, 26 Nov 2013 20:09:44 +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)) (No client certificate requested) by mx1.freebsd.org (Postfix) with ESMTPS id 96D8B2962; Tue, 26 Nov 2013 20:09:44 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.7/8.14.7) with ESMTP id rAQK9iwr018201; Tue, 26 Nov 2013 20:09:44 GMT (envelope-from avg@svn.freebsd.org) Received: (from avg@localhost) by svn.freebsd.org (8.14.7/8.14.5/Submit) id rAQK9iDR018197; Tue, 26 Nov 2013 20:09:44 GMT (envelope-from avg@svn.freebsd.org) Message-Id: <201311262009.rAQK9iDR018197@svn.freebsd.org> From: Andriy Gapon Date: Tue, 26 Nov 2013 20:09:44 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-vendor@freebsd.org Subject: svn commit: r258665 - vendor-sys/illumos/dist/uts/common/fs/zfs X-SVN-Group: vendor-sys MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-vendor@freebsd.org X-Mailman-Version: 2.1.16 Precedence: list List-Id: SVN commit messages for the vendor work area tree List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 26 Nov 2013 20:09:44 -0000 Author: avg Date: Tue Nov 26 20:09:43 2013 New Revision: 258665 URL: http://svnweb.freebsd.org/changeset/base/258665 Log: 4347 ZPL can use dmu_tx_assign(TXG_WAIT) illumos/illumos-gate@e722410c49fe67cbf0f639cbcc288bd6cbcf7dd1 Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_dir.c vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_znode.c Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_dir.c ============================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_dir.c Tue Nov 26 20:03:25 2013 (r258664) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_dir.c Tue Nov 26 20:09:43 2013 (r258665) @@ -937,7 +937,6 @@ zfs_make_xattrdir(znode_t *zp, vattr_t * return (SET_ERROR(EDQUOT)); } -top: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); @@ -946,13 +945,8 @@ top: fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); - error = dmu_tx_assign(tx, TXG_NOWAIT); + error = dmu_tx_assign(tx, TXG_WAIT); if (error) { - if (error == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto top; - } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); return (error); Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c ============================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c Tue Nov 26 20:03:25 2013 (r258664) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c Tue Nov 26 20:09:43 2013 (r258665) @@ -111,11 +111,18 @@ * (3) All range locks must be grabbed before calling dmu_tx_assign(), * as they can span dmu_tx_assign() calls. * - * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign(). - * This is critical because we don't want to block while holding locks. - * Note, in particular, that if a lock is sometimes acquired before - * the tx assigns, and sometimes after (e.g. z_lock), then failing to - * use a non-blocking assign can deadlock the system. The scenario: + * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to + * dmu_tx_assign(). This is critical because we don't want to block + * while holding locks. + * + * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This + * reduces lock contention and CPU usage when we must wait (note that if + * throughput is constrained by the storage, nearly every transaction + * must wait). + * + * Note, in particular, that if a lock is sometimes acquired before + * the tx assigns, and sometimes after (e.g. z_lock), then failing + * to use a non-blocking assign can deadlock the system. The scenario: * * Thread A has grabbed a lock before calling dmu_tx_assign(). * Thread B is in an already-assigned tx, and blocks for this lock. @@ -728,7 +735,6 @@ zfs_write(vnode_t *vp, uio_t *uio, int i while (n > 0) { abuf = NULL; woff = uio->uio_loffset; -again: if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { if (abuf != NULL) @@ -780,13 +786,8 @@ again: dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); zfs_sa_upgrade_txholds(tx, zp); - error = dmu_tx_assign(tx, TXG_NOWAIT); + error = dmu_tx_assign(tx, TXG_WAIT); if (error) { - if (error == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto again; - } dmu_tx_abort(tx); if (abuf != NULL) dmu_return_arcbuf(abuf); @@ -3043,12 +3044,9 @@ top: zfs_sa_upgrade_txholds(tx, zp); - err = dmu_tx_assign(tx, TXG_NOWAIT); - if (err) { - if (err == ERESTART) - dmu_tx_wait(tx); + err = dmu_tx_assign(tx, TXG_WAIT); + if (err) goto out; - } count = 0; /* @@ -4140,19 +4138,13 @@ zfs_putapage(vnode_t *vp, page_t *pp, u_ err = SET_ERROR(EDQUOT); goto out; } -top: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_write(tx, zp->z_id, off, len); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); - err = dmu_tx_assign(tx, TXG_NOWAIT); + err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { - if (err == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto top; - } dmu_tx_abort(tx); goto out; } Modified: vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_znode.c ============================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_znode.c Tue Nov 26 20:03:25 2013 (r258664) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_znode.c Tue Nov 26 20:09:43 2013 (r258665) @@ -1469,7 +1469,6 @@ zfs_extend(znode_t *zp, uint64_t end) zfs_range_unlock(rl); return (0); } -top: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); @@ -1489,13 +1488,8 @@ top: newblksz = 0; } - error = dmu_tx_assign(tx, TXG_NOWAIT); + error = dmu_tx_assign(tx, TXG_WAIT); if (error) { - if (error == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto top; - } dmu_tx_abort(tx); zfs_range_unlock(rl); return (error); @@ -1592,17 +1586,11 @@ zfs_trunc(znode_t *zp, uint64_t end) zfs_range_unlock(rl); return (error); } -top: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); - error = dmu_tx_assign(tx, TXG_NOWAIT); + error = dmu_tx_assign(tx, TXG_WAIT); if (error) { - if (error == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto top; - } dmu_tx_abort(tx); zfs_range_unlock(rl); return (error); @@ -1709,13 +1697,8 @@ log: tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); - error = dmu_tx_assign(tx, TXG_NOWAIT); + error = dmu_tx_assign(tx, TXG_WAIT); if (error) { - if (error == ERESTART) { - dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto log; - } dmu_tx_abort(tx); return (error); }