X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fllite%2Ffile.c;h=14e38cf7050c35306ea957d2579feb1ce9498d90;hb=a7811634d45319dc3812fa652c727367fa290a24;hp=8d04d4e277b1fa29e47520baf418d7a5197c5699;hpb=c42b426c87c3d3b1dc9eda612cc831293dc80d68;p=fs%2Flustre-release.git

diff --git a/lustre/llite/file.c b/lustre/llite/file.c
index 8d04d4e..14e38cf 100644
--- a/lustre/llite/file.c
+++ b/lustre/llite/file.c
@@ -705,8 +705,7 @@ static int ll_lsm_getattr(struct lov_stripe_md *lsm, struct obd_export *exp,
 
         oinfo.oi_md = lsm;
         oinfo.oi_oa = obdo;
-        oinfo.oi_oa->o_id = lsm->lsm_object_id;
-        oinfo.oi_oa->o_seq = lsm->lsm_object_seq;
+	oinfo.oi_oa->o_oi = lsm->lsm_oi;
         oinfo.oi_oa->o_mode = S_IFREG;
         oinfo.oi_oa->o_ioepoch = ioepoch;
         oinfo.oi_oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE |
@@ -756,10 +755,11 @@ int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
 			    capa, obdo, ioepoch, sync);
 	capa_put(capa);
 	if (rc == 0) {
+		struct ost_id *oi = lsm ? &lsm->lsm_oi : &obdo->o_oi;
+
 		obdo_refresh_inode(inode, obdo, obdo->o_valid);
-		CDEBUG(D_INODE,
-		       "objid "LPX64" size %llu, blocks %llu, blksize %lu\n",
-		       lsm ? lsm->lsm_object_id : 0, i_size_read(inode),
+		CDEBUG(D_INODE, "objid "DOSTID" size %llu, blocks %llu,"
+		       " blksize %lu\n", POSTID(oi), i_size_read(inode),
 		       (unsigned long long)inode->i_blocks,
 		       (unsigned long)ll_inode_blksize(inode));
 	}
@@ -767,38 +767,47 @@ int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
 	RETURN(rc);
 }
 
-int ll_merge_lvb(struct inode *inode)
+int ll_merge_lvb(const struct lu_env *env, struct inode *inode)
 {
 	struct ll_inode_info *lli = ll_i2info(inode);
-	struct ll_sb_info *sbi = ll_i2sbi(inode);
-	struct lov_stripe_md *lsm;
+	struct cl_object *obj = lli->lli_clob;
+	struct cl_attr *attr = ccc_env_thread_attr(env);
 	struct ost_lvb lvb;
 	int rc = 0;
 
 	ENTRY;
 
-	lsm = ccc_inode_lsm_get(inode);
 	ll_inode_size_lock(inode);
+	/* merge timestamps the most recently obtained from mds with
+	   timestamps obtained from osts */
+	LTIME_S(inode->i_atime) = lli->lli_lvb.lvb_atime;
+	LTIME_S(inode->i_mtime) = lli->lli_lvb.lvb_mtime;
+	LTIME_S(inode->i_ctime) = lli->lli_lvb.lvb_ctime;
 	inode_init_lvb(inode, &lvb);
 
-	/* merge timestamps the most resently obtained from mds with
-	   timestamps obtained from osts */
-	lvb.lvb_atime = lli->lli_lvb.lvb_atime;
-	lvb.lvb_mtime = lli->lli_lvb.lvb_mtime;
-	lvb.lvb_ctime = lli->lli_lvb.lvb_ctime;
-	if (lsm != NULL) {
-		rc = obd_merge_lvb(sbi->ll_dt_exp, lsm, &lvb, 0);
-		cl_isize_write_nolock(inode, lvb.lvb_size);
+	cl_object_attr_lock(obj);
+	rc = cl_object_attr_get(env, obj, attr);
+	cl_object_attr_unlock(obj);
+
+	if (rc == 0) {
+		if (lvb.lvb_atime < attr->cat_atime)
+			lvb.lvb_atime = attr->cat_atime;
+		if (lvb.lvb_ctime < attr->cat_ctime)
+			lvb.lvb_ctime = attr->cat_ctime;
+		if (lvb.lvb_mtime < attr->cat_mtime)
+			lvb.lvb_mtime = attr->cat_mtime;
 
 		CDEBUG(D_VFSTRACE, DFID" updating i_size "LPU64"\n",
-				PFID(&lli->lli_fid), lvb.lvb_size);
-		inode->i_blocks = lvb.lvb_blocks;
+				PFID(&lli->lli_fid), attr->cat_size);
+		cl_isize_write_nolock(inode, attr->cat_size);
+
+		inode->i_blocks = attr->cat_blocks;
+
+		LTIME_S(inode->i_mtime) = lvb.lvb_mtime;
+		LTIME_S(inode->i_atime) = lvb.lvb_atime;
+		LTIME_S(inode->i_ctime) = lvb.lvb_ctime;
 	}
-	LTIME_S(inode->i_mtime) = lvb.lvb_mtime;
-	LTIME_S(inode->i_atime) = lvb.lvb_atime;
-	LTIME_S(inode->i_ctime) = lvb.lvb_ctime;
 	ll_inode_size_unlock(inode);
-	ccc_inode_lsm_put(inode, lsm);
 
 	RETURN(rc);
 }
@@ -827,7 +836,9 @@ void ll_io_init(struct cl_io *io, const struct file *file, int write)
         io->u.ci_rw.crw_nonblock = file->f_flags & O_NONBLOCK;
 	if (write) {
 		io->u.ci_wr.wr_append = !!(file->f_flags & O_APPEND);
-		io->u.ci_wr.wr_sync = file->f_flags & O_SYNC || IS_SYNC(inode);
+		io->u.ci_wr.wr_sync = file->f_flags & O_SYNC ||
+				      file->f_flags & O_DIRECT ||
+				      IS_SYNC(inode);
 	}
         io->ci_obj     = ll_i2info(inode)->lli_clob;
         io->ci_lockreq = CILR_MAYBE;
@@ -850,6 +861,7 @@ ll_file_io_generic(const struct lu_env *env, struct vvp_io_args *args,
         ssize_t               result;
         ENTRY;
 
+restart:
         io = ccc_env_thread_io(env);
         ll_io_init(io, file, iot == CIT_WRITE);
 
@@ -908,6 +920,16 @@ ll_file_io_generic(const struct lu_env *env, struct vvp_io_args *args,
         GOTO(out, result);
 out:
         cl_io_fini(env, io);
+	/* If any bit been read/written (result != 0), we just return
+	 * short read/write instead of restart io. */
+	if (result == 0 && io->ci_need_restart) {
+		CDEBUG(D_VFSTRACE, "Restart %s on %s from %lld, count:%zd\n",
+		       iot == CIT_READ ? "read" : "write",
+		       file->f_dentry->d_name.name, *ppos, count);
+		LASSERTF(io->u.ci_rw.crw_count == count, "%zd != %zd\n",
+			 io->u.ci_rw.crw_count, count);
+		goto restart;
+	}
 
         if (iot == CIT_READ) {
                 if (result >= 0)
@@ -918,7 +940,7 @@ out:
                         ll_stats_ops_tally(ll_i2sbi(file->f_dentry->d_inode),
                                            LPROC_LL_WRITE_BYTES, result);
 			fd->fd_write_failed = false;
-		} else {
+		} else if (result != -ERESTARTSYS) {
 			fd->fd_write_failed = true;
 		}
 	}
@@ -1669,8 +1691,7 @@ int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
 	    !(fiemap->fm_flags & FIEMAP_FLAG_DEVICE_ORDER))
 		GOTO(out, rc = -EOPNOTSUPP);
 
-        fm_key.oa.o_id = lsm->lsm_object_id;
-        fm_key.oa.o_seq = lsm->lsm_object_seq;
+	fm_key.oa.o_oi = lsm->lsm_oi;
         fm_key.oa.o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
 
         obdo_from_inode(&fm_key.oa, inode, OBD_MD_FLSIZE);
@@ -1726,6 +1747,7 @@ int ll_fid2path(struct inode *inode, void *arg)
 	rc = obd_iocontrol(OBD_IOC_FID2PATH, exp, outsize, gfout, NULL);
 	if (rc)
 		GOTO(gf_free, rc);
+
 	if (copy_to_user(arg, gfout, outsize))
 		rc = -EFAULT;
 
@@ -1832,13 +1854,163 @@ int ll_data_version(struct inode *inode, __u64 *data_version,
 	RETURN(rc);
 }
 
-long ll_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+struct ll_swap_stack {
+	struct iattr		 ia1, ia2;
+	__u64			 dv1, dv2;
+	struct inode		*inode1, *inode2;
+	bool			 check_dv1, check_dv2;
+};
+
+static int ll_swap_layouts(struct file *file1, struct file *file2,
+			   struct lustre_swap_layouts *lsl)
 {
-        struct inode *inode = file->f_dentry->d_inode;
-        struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
-        int flags;
+	struct mdc_swap_layouts	 msl;
+	struct md_op_data	*op_data;
+	__u32			 gid;
+	__u64			 dv;
+	struct ll_swap_stack	*llss = NULL;
+	int			 rc, rc1;
 
-        ENTRY;
+	OBD_ALLOC_PTR(llss);
+	if (llss == NULL)
+		RETURN(-ENOMEM);
+
+	llss->inode1 = file1->f_dentry->d_inode;
+	llss->inode2 = file2->f_dentry->d_inode;
+
+	if (!S_ISREG(llss->inode2->i_mode))
+		GOTO(free, rc = -EINVAL);
+
+	if (ll_permission(llss->inode1, MAY_WRITE, NULL) ||
+	    ll_permission(llss->inode2, MAY_WRITE, NULL))
+		GOTO(free, rc = -EPERM);
+
+	if (llss->inode2->i_sb != llss->inode1->i_sb)
+		GOTO(free, rc = -EXDEV);
+
+	/* we use 2 bool because it is easier to swap than 2 bits */
+	if (lsl->sl_flags & SWAP_LAYOUTS_CHECK_DV1)
+		llss->check_dv1 = true;
+
+	if (lsl->sl_flags & SWAP_LAYOUTS_CHECK_DV2)
+		llss->check_dv2 = true;
+
+	/* we cannot use lsl->sl_dvX directly because we may swap them */
+	llss->dv1 = lsl->sl_dv1;
+	llss->dv2 = lsl->sl_dv2;
+
+	rc = lu_fid_cmp(ll_inode2fid(llss->inode1), ll_inode2fid(llss->inode2));
+	if (rc == 0) /* same file, done! */
+		GOTO(free, rc = 0);
+
+	if (rc < 0) { /* sequentialize it */
+		swap(llss->inode1, llss->inode2);
+		swap(file1, file2);
+		swap(llss->dv1, llss->dv2);
+		swap(llss->check_dv1, llss->check_dv2);
+	}
+
+	gid = lsl->sl_gid;
+	if (gid != 0) { /* application asks to flush dirty cache */
+		rc = ll_get_grouplock(llss->inode1, file1, gid);
+		if (rc < 0)
+			GOTO(free, rc);
+
+		rc = ll_get_grouplock(llss->inode2, file2, gid);
+		if (rc < 0) {
+			ll_put_grouplock(llss->inode1, file1, gid);
+			GOTO(free, rc);
+		}
+	}
+
+	/* to be able to restore mtime and atime after swap
+	 * we need to first save them */
+	if (lsl->sl_flags &
+	    (SWAP_LAYOUTS_KEEP_MTIME | SWAP_LAYOUTS_KEEP_ATIME)) {
+		llss->ia1.ia_mtime = llss->inode1->i_mtime;
+		llss->ia1.ia_atime = llss->inode1->i_atime;
+		llss->ia1.ia_valid = ATTR_MTIME | ATTR_ATIME;
+		llss->ia2.ia_mtime = llss->inode2->i_mtime;
+		llss->ia2.ia_atime = llss->inode2->i_atime;
+		llss->ia2.ia_valid = ATTR_MTIME | ATTR_ATIME;
+	}
+
+	/* ultimate check, before swaping the layouts we check if
+	 * dataversion has changed (if requested) */
+	if (llss->check_dv1) {
+		rc = ll_data_version(llss->inode1, &dv, 0);
+		if (rc)
+			GOTO(putgl, rc);
+		if (dv != llss->dv1)
+			GOTO(putgl, rc = -EAGAIN);
+	}
+
+	if (llss->check_dv2) {
+		rc = ll_data_version(llss->inode2, &dv, 0);
+		if (rc)
+			GOTO(putgl, rc);
+		if (dv != llss->dv2)
+			GOTO(putgl, rc = -EAGAIN);
+	}
+
+	/* struct md_op_data is used to send the swap args to the mdt
+	 * only flags is missing, so we use struct mdc_swap_layouts
+	 * through the md_op_data->op_data */
+	/* flags from user space have to be converted before they are send to
+	 * server, no flag is sent today, they are only used on the client */
+	msl.msl_flags = 0;
+	rc = -ENOMEM;
+	op_data = ll_prep_md_op_data(NULL, llss->inode1, llss->inode2, NULL, 0,
+				     0, LUSTRE_OPC_ANY, &msl);
+	if (op_data != NULL) {
+		rc = obd_iocontrol(LL_IOC_LOV_SWAP_LAYOUTS,
+				   ll_i2mdexp(llss->inode1),
+				   sizeof(*op_data), op_data, NULL);
+		ll_finish_md_op_data(op_data);
+	}
+
+putgl:
+	if (gid != 0) {
+		ll_put_grouplock(llss->inode2, file2, gid);
+		ll_put_grouplock(llss->inode1, file1, gid);
+	}
+
+	/* rc can be set from obd_iocontrol() or from a GOTO(putgl, ...) */
+	if (rc != 0)
+		GOTO(free, rc);
+
+	/* clear useless flags */
+	if (!(lsl->sl_flags & SWAP_LAYOUTS_KEEP_MTIME)) {
+		llss->ia1.ia_valid &= ~ATTR_MTIME;
+		llss->ia2.ia_valid &= ~ATTR_MTIME;
+	}
+
+	if (!(lsl->sl_flags & SWAP_LAYOUTS_KEEP_ATIME)) {
+		llss->ia1.ia_valid &= ~ATTR_ATIME;
+		llss->ia2.ia_valid &= ~ATTR_ATIME;
+	}
+
+	/* update time if requested */
+	rc = rc1 = 0;
+	if (llss->ia2.ia_valid != 0)
+		rc = ll_setattr(file1->f_dentry, &llss->ia2);
+
+	if (llss->ia1.ia_valid != 0)
+		rc1 = ll_setattr(file2->f_dentry, &llss->ia1);
+
+free:
+	if (llss != NULL)
+		OBD_FREE_PTR(llss);
+
+	RETURN(rc);
+}
+
+long ll_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	struct inode		*inode = file->f_dentry->d_inode;
+	struct ll_file_data	*fd = LUSTRE_FPRIVATE(file);
+	int			 flags, rc;
+	ENTRY;
 
         CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),cmd=%x\n", inode->i_ino,
                inode->i_generation, inode, cmd);
@@ -1878,6 +2050,27 @@ long ll_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
                 RETURN(ll_lov_setstripe(inode, file, arg));
         case LL_IOC_LOV_SETEA:
                 RETURN(ll_lov_setea(inode, file, arg));
+	case LL_IOC_LOV_SWAP_LAYOUTS: {
+		struct file *file2;
+		struct lustre_swap_layouts lsl;
+
+		if (cfs_copy_from_user(&lsl, (char *)arg,
+				       sizeof(struct lustre_swap_layouts)))
+			RETURN(-EFAULT);
+
+		if ((file->f_flags & O_ACCMODE) == 0) /* O_RDONLY */
+			RETURN(-EPERM);
+
+		file2 = fget(lsl.sl_fd);
+		if (file2 == NULL)
+			RETURN(-EBADF);
+
+		rc = -EPERM;
+		if ((file2->f_flags & O_ACCMODE) != 0) /* O_WRONLY or O_RDWR */
+			rc = ll_swap_layouts(file, file2, &lsl);
+		fput(file2);
+		RETURN(rc);
+	}
         case LL_IOC_LOV_GETSTRIPE:
                 RETURN(ll_lov_getstripe(inode, arg));
         case LL_IOC_RECREATE_OBJ:
@@ -2009,12 +2202,37 @@ long ll_file_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		OBD_FREE_PTR(hss);
 		RETURN(rc);
 	}
+	case LL_IOC_HSM_ACTION: {
+		struct md_op_data		*op_data;
+		struct hsm_current_action	*hca;
+		int				 rc;
+
+		OBD_ALLOC_PTR(hca);
+		if (hca == NULL)
+			RETURN(-ENOMEM);
+
+		op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0,
+					     LUSTRE_OPC_ANY, hca);
+		if (op_data == NULL) {
+			OBD_FREE_PTR(hca);
+			RETURN(-ENOMEM);
+		}
 
+		rc = obd_iocontrol(cmd, ll_i2mdexp(inode), sizeof(*op_data),
+				   op_data, NULL);
+
+		if (cfs_copy_to_user((char *)arg, hca, sizeof(*hca)))
+			rc = -EFAULT;
+
+		ll_finish_md_op_data(op_data);
+		OBD_FREE_PTR(hca);
+		RETURN(rc);
+	}
 	default: {
 		int err;
 
 		if (LLIOC_STOP ==
-			ll_iocontrol_call(inode, file, cmd, arg, &err))
+		     ll_iocontrol_call(inode, file, cmd, arg, &err))
 			RETURN(err);
 
 		RETURN(obd_iocontrol(cmd, ll_i2dtexp(inode), 0, NULL,
@@ -2110,7 +2328,7 @@ loff_t ll_file_seek(struct file *file, loff_t offset, int origin)
 		eof = i_size_read(inode);
 	}
 
-	retval = generic_file_llseek_size(file, offset, origin,
+	retval = ll_generic_file_llseek_size(file, offset, origin,
 					  ll_file_maxbytes(inode), eof);
 	RETURN(retval);
 }
@@ -2193,15 +2411,25 @@ int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
 	RETURN(result);
 }
 
+/*
+ * When dentry is provided (the 'else' case), *file->f_dentry may be
+ * null and dentry must be used directly rather than pulled from
+ * *file->f_dentry as is done otherwise.
+ */
+
 #ifdef HAVE_FILE_FSYNC_4ARGS
 int ll_fsync(struct file *file, loff_t start, loff_t end, int data)
+{
+	struct dentry *dentry = file->f_dentry;
 #elif defined(HAVE_FILE_FSYNC_2ARGS)
 int ll_fsync(struct file *file, int data)
+{
+	struct dentry *dentry = file->f_dentry;
 #else
 int ll_fsync(struct file *file, struct dentry *dentry, int data)
-#endif
 {
-        struct inode *inode = file->f_dentry->d_inode;
+#endif
+        struct inode *inode = dentry->d_inode;
         struct ll_inode_info *lli = ll_i2info(inode);
         struct ptlrpc_request *req;
         struct obd_capa *oc;
@@ -2502,7 +2730,7 @@ int __ll_inode_revalidate_it(struct dentry *dentry, struct lookup_intent *it,
         /* XXX: Enable OBD_CONNECT_ATTRFID to reduce unnecessary getattr RPC.
          *      But under CMD case, it caused some lock issues, should be fixed
          *      with new CMD ibits lock. See bug 12718 */
-        if (exp->exp_connect_flags & OBD_CONNECT_ATTRFID) {
+	if (exp_connect_flags(exp) & OBD_CONNECT_ATTRFID) {
                 struct lookup_intent oit = { .it_op = IT_GETATTR };
                 struct md_op_data *op_data;
 
@@ -3050,6 +3278,7 @@ static int ll_layout_lock_set(struct lustre_handle *lockh, ldlm_mode_t mode,
 		rc = obd_unpackmd(sbi->ll_dt_exp, &md.lsm,
 				  lock->l_lvb_data, lock->l_lvb_len);
 		if (rc >= 0) {
+			*gen = LL_LAYOUT_GEN_EMPTY;
 			if (md.lsm != NULL)
 				*gen = md.lsm->lsm_layout_gen;
 			rc = 0;
@@ -3125,11 +3354,11 @@ int ll_layout_refresh(struct inode *inode, __u32 *gen)
 					   .ei_mode = LCK_CR,
 					   .ei_cb_bl = ll_md_blocking_ast,
 					   .ei_cb_cp = ldlm_completion_ast,
-					   .ei_cbdata = inode };
+					   .ei_cbdata = NULL };
 	int rc;
 	ENTRY;
 
-	*gen = LL_LAYOUT_GEN_ZERO;
+	*gen = LL_LAYOUT_GEN_NONE;
 	if (!(sbi->ll_flags & LL_SBI_LAYOUT_LOCK))
 		RETURN(0);
 
@@ -3188,6 +3417,8 @@ again:
 
 	ll_finish_md_op_data(op_data);
 
+	md_set_lock_data(sbi->ll_md_exp, &it.d.lustre.it_lock_handle, inode, NULL);
+
 	mode = it.d.lustre.it_lock_mode;
 	it.d.lustre.it_lock_mode = 0;
 	ll_intent_drop_lock(&it);