From owner-svn-src-all@FreeBSD.ORG Wed Aug 29 15:41:18 2012 Return-Path: Delivered-To: svn-src-all@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 1C6A71065670; Wed, 29 Aug 2012 15:41:18 +0000 (UTC) (envelope-from kib@FreeBSD.org) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:4f8:fff6::2c]) by mx1.freebsd.org (Postfix) with ESMTP id 064FF8FC0A; Wed, 29 Aug 2012 15:41:18 +0000 (UTC) Received: from svn.freebsd.org (localhost [127.0.0.1]) by svn.freebsd.org (8.14.4/8.14.4) with ESMTP id q7TFfHt0070053; Wed, 29 Aug 2012 15:41:17 GMT (envelope-from kib@svn.freebsd.org) Received: (from kib@localhost) by svn.freebsd.org (8.14.4/8.14.4/Submit) id q7TFfHsR070051; Wed, 29 Aug 2012 15:41:17 GMT (envelope-from kib@svn.freebsd.org) Message-Id: <201208291541.q7TFfHsR070051@svn.freebsd.org> From: Konstantin Belousov Date: Wed, 29 Aug 2012 15:41:17 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-9@freebsd.org X-SVN-Group: stable-9 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Cc: Subject: svn commit: r239842 - stable/9/sys/kern X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 29 Aug 2012 15:41:18 -0000 Author: kib Date: Wed Aug 29 15:41:17 2012 New Revision: 239842 URL: http://svn.freebsd.org/changeset/base/239842 Log: MFC r237365: Fix locking for f_offset, vn_read() and vn_write() cases only. Modified: stable/9/sys/kern/vfs_vnops.c Directory Properties: stable/9/sys/ (props changed) Modified: stable/9/sys/kern/vfs_vnops.c ============================================================================== --- stable/9/sys/kern/vfs_vnops.c Wed Aug 29 15:38:18 2012 (r239841) +++ stable/9/sys/kern/vfs_vnops.c Wed Aug 29 15:41:17 2012 (r239842) @@ -512,6 +512,66 @@ vn_rdwr_inchunks(rw, vp, base, len, offs return (error); } +static void +foffset_lock(struct file *fp, struct uio *uio, int flags) +{ + struct mtx *mtxp; + + if ((flags & FOF_OFFSET) != 0) + return; + + /* + * According to McKusick the vn lock was protecting f_offset here. + * It is now protected by the FOFFSET_LOCKED flag. + */ + mtxp = mtx_pool_find(mtxpool_sleep, fp); + mtx_lock(mtxp); + while (fp->f_vnread_flags & FOFFSET_LOCKED) { + fp->f_vnread_flags |= FOFFSET_LOCK_WAITING; + msleep(&fp->f_vnread_flags, mtxp, PUSER -1, + "vnread offlock", 0); + } + fp->f_vnread_flags |= FOFFSET_LOCKED; + uio->uio_offset = fp->f_offset; + mtx_unlock(mtxp); +} + +static int +get_advice(struct file *fp, struct uio *uio) +{ + struct mtx *mtxp; + int ret; + + ret = POSIX_FADV_NORMAL; + if (fp->f_advice == NULL) + return (ret); + + mtxp = mtx_pool_find(mtxpool_sleep, fp); + mtx_lock(mtxp); + if (uio->uio_offset >= fp->f_advice->fa_start && + uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end) + ret = fp->f_advice->fa_advice; + mtx_unlock(mtxp); + return (ret); +} + +static void +foffset_unlock(struct file *fp, struct uio *uio, int flags) +{ + struct mtx *mtxp; + + if ((flags & FOF_OFFSET) != 0) + return; + + fp->f_offset = uio->uio_offset; + mtxp = mtx_pool_find(mtxpool_sleep, fp); + mtx_lock(mtxp); + if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING) + wakeup(&fp->f_vnread_flags); + fp->f_vnread_flags = 0; + mtx_unlock(mtxp); +} + /* * File table vnode read routine. */ @@ -524,44 +584,22 @@ vn_read(fp, uio, active_cred, flags, td) struct thread *td; { struct vnode *vp; - int error, ioflag; struct mtx *mtxp; + int error, ioflag; int advice, vfslocked; off_t offset, start, end; KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); - mtxp = NULL; + KASSERT(flags & FOF_OFFSET, ("No FOF_OFFSET")); vp = fp->f_vnode; ioflag = 0; if (fp->f_flag & FNONBLOCK) ioflag |= IO_NDELAY; if (fp->f_flag & O_DIRECT) ioflag |= IO_DIRECT; - advice = POSIX_FADV_NORMAL; + advice = get_advice(fp, uio); vfslocked = VFS_LOCK_GIANT(vp->v_mount); - /* - * According to McKusick the vn lock was protecting f_offset here. - * It is now protected by the FOFFSET_LOCKED flag. - */ - if ((flags & FOF_OFFSET) == 0 || fp->f_advice != NULL) { - mtxp = mtx_pool_find(mtxpool_sleep, fp); - mtx_lock(mtxp); - if ((flags & FOF_OFFSET) == 0) { - while (fp->f_vnread_flags & FOFFSET_LOCKED) { - fp->f_vnread_flags |= FOFFSET_LOCK_WAITING; - msleep(&fp->f_vnread_flags, mtxp, PUSER -1, - "vnread offlock", 0); - } - fp->f_vnread_flags |= FOFFSET_LOCKED; - uio->uio_offset = fp->f_offset; - } - if (fp->f_advice != NULL && - uio->uio_offset >= fp->f_advice->fa_start && - uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end) - advice = fp->f_advice->fa_advice; - mtx_unlock(mtxp); - } vn_lock(vp, LK_SHARED | LK_RETRY); switch (advice) { @@ -581,14 +619,6 @@ vn_read(fp, uio, active_cred, flags, td) if (error == 0) #endif error = VOP_READ(vp, uio, ioflag, fp->f_cred); - if ((flags & FOF_OFFSET) == 0) { - fp->f_offset = uio->uio_offset; - mtx_lock(mtxp); - if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING) - wakeup(&fp->f_vnread_flags); - fp->f_vnread_flags = 0; - mtx_unlock(mtxp); - } fp->f_nextoff = uio->uio_offset; VOP_UNLOCK(vp, 0); if (error == 0 && advice == POSIX_FADV_NOREUSE && @@ -610,6 +640,7 @@ vn_read(fp, uio, active_cred, flags, td) */ start = offset; end = uio->uio_offset - 1; + mtxp = mtx_pool_find(mtxpool_sleep, fp); mtx_lock(mtxp); if (fp->f_advice != NULL && fp->f_advice->fa_advice == POSIX_FADV_NOREUSE) { @@ -641,13 +672,14 @@ vn_write(fp, uio, active_cred, flags, td { struct vnode *vp; struct mount *mp; - int error, ioflag, lock_flags; struct mtx *mtxp; + int error, ioflag, lock_flags; int advice, vfslocked; off_t offset, start, end; KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); + KASSERT(flags & FOF_OFFSET, ("No FOF_OFFSET")); vp = fp->f_vnode; vfslocked = VFS_LOCK_GIANT(vp->v_mount); if (vp->v_type == VREG) @@ -666,6 +698,8 @@ vn_write(fp, uio, active_cred, flags, td if (vp->v_type != VCHR && (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) goto unlock; + + advice = get_advice(fp, uio); if ((MNT_SHARED_WRITES(mp) || ((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) && @@ -676,19 +710,6 @@ vn_write(fp, uio, active_cred, flags, td } vn_lock(vp, lock_flags | LK_RETRY); - if ((flags & FOF_OFFSET) == 0) - uio->uio_offset = fp->f_offset; - advice = POSIX_FADV_NORMAL; - mtxp = NULL; - if (fp->f_advice != NULL) { - mtxp = mtx_pool_find(mtxpool_sleep, fp); - mtx_lock(mtxp); - if (fp->f_advice != NULL && - uio->uio_offset >= fp->f_advice->fa_start && - uio->uio_offset + uio->uio_resid <= fp->f_advice->fa_end) - advice = fp->f_advice->fa_advice; - mtx_unlock(mtxp); - } switch (advice) { case POSIX_FADV_NORMAL: case POSIX_FADV_SEQUENTIAL: @@ -706,8 +727,6 @@ vn_write(fp, uio, active_cred, flags, td if (error == 0) #endif error = VOP_WRITE(vp, uio, ioflag, fp->f_cred); - if ((flags & FOF_OFFSET) == 0) - fp->f_offset = uio->uio_offset; fp->f_nextoff = uio->uio_offset; VOP_UNLOCK(vp, 0); if (vp->v_type != VCHR) @@ -746,6 +765,7 @@ vn_write(fp, uio, active_cred, flags, td */ start = offset; end = uio->uio_offset - 1; + mtxp = mtx_pool_find(mtxpool_sleep, fp); mtx_lock(mtxp); if (fp->f_advice != NULL && fp->f_advice->fa_advice == POSIX_FADV_NOREUSE) { @@ -830,11 +850,15 @@ vn_io_fault(struct file *fp, struct uio else doio = vn_write; vp = fp->f_vnode; + foffset_lock(fp, uio, flags); + if (uio->uio_segflg != UIO_USERSPACE || vp->v_type != VREG || ((mp = vp->v_mount) != NULL && (mp->mnt_kern_flag & MNTK_NO_IOPF) == 0) || - !vn_io_fault_enable) - return (doio(fp, uio, active_cred, flags, td)); + !vn_io_fault_enable) { + error = doio(fp, uio, active_cred, flags | FOF_OFFSET, td); + goto out_last; + } /* * The UFS follows IO_UNIT directive and replays back both @@ -867,7 +891,7 @@ vn_io_fault(struct file *fp, struct uio } save = vm_fault_disable_pagefaults(); - error = doio(fp, uio, active_cred, flags, td); + error = doio(fp, uio, active_cred, flags | FOF_OFFSET, td); if (error != EFAULT) goto out; @@ -918,7 +942,8 @@ vn_io_fault(struct file *fp, struct uio td->td_ma = ma; td->td_ma_cnt = cnt; - error = doio(fp, &short_uio, active_cred, flags, td); + error = doio(fp, &short_uio, active_cred, flags | FOF_OFFSET, + td); vm_page_unhold_pages(ma, cnt); adv = len - short_uio.uio_resid; @@ -941,6 +966,8 @@ out: vm_fault_enable_pagefaults(save); vn_rangelock_unlock(vp, rl_cookie); free(uio_clone, M_IOV); +out_last: + foffset_unlock(fp, uio, flags); return (error); }