X-Git-Url: https://err.no/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=fs%2Fxfs%2Fxfs_vnodeops.c;h=e475e3717eb3cf114a61a6cbbe4f2339df5d4a7d;hb=b4ba0ba24b57ec975482f4ba2d350fbee7557240;hp=2ebfc60097d1aff9c8b4b56154cd7ece30bdb908;hpb=82dab941a192d081dd0b7cde3ed32603372d5acc;p=linux-2.6 diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c index 2ebfc60097..e475e3717e 100644 --- a/fs/xfs/xfs_vnodeops.c +++ b/fs/xfs/xfs_vnodeops.c @@ -856,18 +856,14 @@ xfs_readlink( /* * xfs_fsync * - * This is called to sync the inode and its data out to disk. - * We need to hold the I/O lock while flushing the data, and - * the inode lock while flushing the inode. The inode lock CANNOT - * be held while flushing the data, so acquire after we're done - * with that. + * This is called to sync the inode and its data out to disk. We need to hold + * the I/O lock while flushing the data, and the inode lock while flushing the + * inode. The inode lock CANNOT be held while flushing the data, so acquire + * after we're done with that. */ int xfs_fsync( - xfs_inode_t *ip, - int flag, - xfs_off_t start, - xfs_off_t stop) + xfs_inode_t *ip) { xfs_trans_t *tp; int error; @@ -875,103 +871,79 @@ xfs_fsync( xfs_itrace_entry(ip); - ASSERT(start >= 0 && stop >= -1); - if (XFS_FORCED_SHUTDOWN(ip->i_mount)) return XFS_ERROR(EIO); - if (flag & FSYNC_DATA) - filemap_fdatawait(vn_to_inode(XFS_ITOV(ip))->i_mapping); + /* capture size updates in I/O completion before writing the inode. */ + error = filemap_fdatawait(vn_to_inode(XFS_ITOV(ip))->i_mapping); + if (error) + return XFS_ERROR(error); /* - * We always need to make sure that the required inode state - * is safe on disk. The vnode might be clean but because - * of committed transactions that haven't hit the disk yet. - * Likewise, there could be unflushed non-transactional - * changes to the inode core that have to go to disk. + * We always need to make sure that the required inode state is safe on + * disk. The vnode might be clean but we still might need to force the + * log because of committed transactions that haven't hit the disk yet. + * Likewise, there could be unflushed non-transactional changes to the + * inode core that have to go to disk and this requires us to issue + * a synchronous transaction to capture these changes correctly. * - * The following code depends on one assumption: that - * any transaction that changes an inode logs the core - * because it has to change some field in the inode core - * (typically nextents or nblocks). That assumption - * implies that any transactions against an inode will - * catch any non-transactional updates. If inode-altering - * transactions exist that violate this assumption, the - * code breaks. Right now, it figures that if the involved - * update_* field is clear and the inode is unpinned, the - * inode is clean. Either it's been flushed or it's been - * committed and the commit has hit the disk unpinning the inode. - * (Note that xfs_inode_item_format() called at commit clears - * the update_* fields.) + * This code relies on the assumption that if the update_* fields + * of the inode are clear and the inode is unpinned then it is clean + * and no action is required. */ xfs_ilock(ip, XFS_ILOCK_SHARED); - /* If we are flushing data then we care about update_size - * being set, otherwise we care about update_core - */ - if ((flag & FSYNC_DATA) ? - (ip->i_update_size == 0) : - (ip->i_update_core == 0)) { + if (!(ip->i_update_size || ip->i_update_core)) { /* - * Timestamps/size haven't changed since last inode - * flush or inode transaction commit. That means - * either nothing got written or a transaction - * committed which caught the updates. If the - * latter happened and the transaction hasn't - * hit the disk yet, the inode will be still - * be pinned. If it is, force the log. + * Timestamps/size haven't changed since last inode flush or + * inode transaction commit. That means either nothing got + * written or a transaction committed which caught the updates. + * If the latter happened and the transaction hasn't hit the + * disk yet, the inode will be still be pinned. If it is, + * force the log. */ xfs_iunlock(ip, XFS_ILOCK_SHARED); if (xfs_ipincount(ip)) { - _xfs_log_force(ip->i_mount, (xfs_lsn_t)0, - XFS_LOG_FORCE | - ((flag & FSYNC_WAIT) - ? XFS_LOG_SYNC : 0), + error = _xfs_log_force(ip->i_mount, (xfs_lsn_t)0, + XFS_LOG_FORCE | XFS_LOG_SYNC, &log_flushed); } else { /* - * If the inode is not pinned and nothing - * has changed we don't need to flush the - * cache. + * If the inode is not pinned and nothing has changed + * we don't need to flush the cache. */ changed = 0; } - error = 0; } else { /* - * Kick off a transaction to log the inode - * core to get the updates. Make it - * sync if FSYNC_WAIT is passed in (which - * is done by everybody but specfs). The - * sync transaction will also force the log. + * Kick off a transaction to log the inode core to get the + * updates. The sync transaction will also force the log. */ xfs_iunlock(ip, XFS_ILOCK_SHARED); tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS); - if ((error = xfs_trans_reserve(tp, 0, - XFS_FSYNC_TS_LOG_RES(ip->i_mount), - 0, 0, 0))) { + error = xfs_trans_reserve(tp, 0, + XFS_FSYNC_TS_LOG_RES(ip->i_mount), 0, 0, 0); + if (error) { xfs_trans_cancel(tp, 0); return error; } xfs_ilock(ip, XFS_ILOCK_EXCL); /* - * Note - it's possible that we might have pushed - * ourselves out of the way during trans_reserve - * which would flush the inode. But there's no - * guarantee that the inode buffer has actually - * gone out yet (it's delwri). Plus the buffer - * could be pinned anyway if it's part of an - * inode in another recent transaction. So we - * play it safe and fire off the transaction anyway. + * Note - it's possible that we might have pushed ourselves out + * of the way during trans_reserve which would flush the inode. + * But there's no guarantee that the inode buffer has actually + * gone out yet (it's delwri). Plus the buffer could be pinned + * anyway if it's part of an inode in another recent + * transaction. So we play it safe and fire off the + * transaction anyway. */ xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); xfs_trans_ihold(tp, ip); xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); - if (flag & FSYNC_WAIT) - xfs_trans_set_sync(tp); + xfs_trans_set_sync(tp); error = _xfs_trans_commit(tp, 0, &log_flushed); xfs_iunlock(ip, XFS_ILOCK_EXCL); @@ -2162,20 +2134,6 @@ xfs_remove( return error; } - /* - * We need to get a reference to ip before we get our log - * reservation. The reason for this is that we cannot call - * xfs_iget for an inode for which we do not have a reference - * once we've acquired a log reservation. This is because the - * inode we are trying to get might be in xfs_inactive going - * for a log reservation. Since we'll have to wait for the - * inactive code to complete before returning from xfs_iget, - * we need to make sure that we don't have log space reserved - * when we call xfs_iget. Instead we get an unlocked reference - * to the inode before getting our log reservation. - */ - IHOLD(ip); - xfs_itrace_entry(ip); xfs_itrace_ref(ip); @@ -2184,7 +2142,6 @@ xfs_remove( error = XFS_QM_DQATTACH(mp, ip, 0); if (error) { REMOVE_DEBUG_TRACE(__LINE__); - IRELE(ip); goto std_return; } @@ -2211,7 +2168,6 @@ xfs_remove( ASSERT(error != ENOSPC); REMOVE_DEBUG_TRACE(__LINE__); xfs_trans_cancel(tp, 0); - IRELE(ip); return error; } @@ -2219,7 +2175,6 @@ xfs_remove( if (error) { REMOVE_DEBUG_TRACE(__LINE__); xfs_trans_cancel(tp, cancel_flags); - IRELE(ip); goto std_return; } @@ -2227,6 +2182,7 @@ xfs_remove( * At this point, we've gotten both the directory and the entry * inodes locked. */ + IHOLD(ip); xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL); IHOLD(dp); @@ -2259,12 +2215,6 @@ xfs_remove( */ link_zero = (ip)->i_d.di_nlink==0; - /* - * Take an extra ref on the inode so that it doesn't - * go to xfs_inactive() from within the commit. - */ - IHOLD(ip); - /* * If this is a synchronous mount, make sure that the * remove transaction goes to disk before returning to @@ -2281,10 +2231,8 @@ xfs_remove( } error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); - if (error) { - IRELE(ip); + if (error) goto std_return; - } /* * If we are using filestreams, kill the stream association. @@ -2296,7 +2244,6 @@ xfs_remove( xfs_filestream_deassociate(ip); xfs_itrace_exit(ip); - IRELE(ip); /* Fall through to std_return with error = 0 */ std_return: @@ -2325,8 +2272,6 @@ xfs_remove( cancel_flags |= XFS_TRANS_ABORT; xfs_trans_cancel(tp, cancel_flags); - IRELE(ip); - goto std_return; } @@ -2698,7 +2643,6 @@ xfs_rmdir( struct xfs_name *name, xfs_inode_t *cdp) { - bhv_vnode_t *dir_vp = XFS_ITOV(dp); xfs_mount_t *mp = dp->i_mount; xfs_trans_t *tp; int error; @@ -2723,20 +2667,6 @@ xfs_rmdir( return XFS_ERROR(error); } - /* - * We need to get a reference to cdp before we get our log - * reservation. The reason for this is that we cannot call - * xfs_iget for an inode for which we do not have a reference - * once we've acquired a log reservation. This is because the - * inode we are trying to get might be in xfs_inactive going - * for a log reservation. Since we'll have to wait for the - * inactive code to complete before returning from xfs_iget, - * we need to make sure that we don't have log space reserved - * when we call xfs_iget. Instead we get an unlocked reference - * to the inode before getting our log reservation. - */ - IHOLD(cdp); - /* * Get the dquots for the inodes. */ @@ -2744,7 +2674,6 @@ xfs_rmdir( if (!error) error = XFS_QM_DQATTACH(mp, cdp, 0); if (error) { - IRELE(cdp); REMOVE_DEBUG_TRACE(__LINE__); goto std_return; } @@ -2771,7 +2700,6 @@ xfs_rmdir( if (error) { ASSERT(error != ENOSPC); cancel_flags = 0; - IRELE(cdp); goto error_return; } XFS_BMAP_INIT(&free_list, &first_block); @@ -2785,14 +2713,13 @@ xfs_rmdir( error = xfs_lock_dir_and_entry(dp, cdp); if (error) { xfs_trans_cancel(tp, cancel_flags); - IRELE(cdp); goto std_return; } + IHOLD(dp); xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL); - VN_HOLD(dir_vp); - xfs_itrace_ref(cdp); + IHOLD(cdp); xfs_trans_ijoin(tp, cdp, XFS_ILOCK_EXCL); ASSERT(cdp->i_d.di_nlink >= 2); @@ -2845,12 +2772,6 @@ xfs_rmdir( /* Determine these before committing transaction */ last_cdp_link = (cdp)->i_d.di_nlink==0; - /* - * Take an extra ref on the child vnode so that it - * does not go to xfs_inactive() from within the commit. - */ - IHOLD(cdp); - /* * If this is a synchronous mount, make sure that the * rmdir transaction goes to disk before returning to @@ -2865,19 +2786,15 @@ xfs_rmdir( xfs_bmap_cancel(&free_list); xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT)); - IRELE(cdp); goto std_return; } error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); if (error) { - IRELE(cdp); goto std_return; } - IRELE(cdp); - /* Fall through to std_return with error = 0 or the errno * from xfs_trans_commit. */ std_return: