staging: lustre: lmv: honor MDT index when creating volatile file

LMV should honor MDT index embedded in the name of volatile
file, then during hsm restore, the file under striped dir can
be restored to the right MDT.
Signed-off-by: wang di <di.wang@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4691
Reviewed-on: http://review.whamcloud.com/10866Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Henri Doreau <henri.doreau@cea.fr>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: James Simmons <jsimmons@infradead.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/staging/lustre/lustre/include/lustre/lustre_user.h

@@ -244,6 +244,7 @@ struct ost_id {
 #define LL_IOC_HSM_IMPORT		_IOWR('f', 245, struct hsm_user_import)
 #define LL_IOC_LMV_SET_DEFAULT_STRIPE	_IOWR('f', 246, struct lmv_user_md)
 #define LL_IOC_MIGRATE			_IOR('f', 247, int)
+#define LL_IOC_FID2MDTIDX		_IOWR('f', 248, struct lu_fid)
 
 #define LL_STATFS_LMV	   1
 #define LL_STATFS_LOV	   2

drivers/staging/lustre/lustre/llite/dir.c

@@ -1568,6 +1568,23 @@ static long ll_dir_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		return rc;
 	case OBD_IOC_FID2PATH:
 		return ll_fid2path(inode, (void __user *)arg);
+	case LL_IOC_FID2MDTIDX: {
+		struct obd_export *exp = ll_i2mdexp(inode);
+		struct lu_fid fid;
+		__u32 index;
+
+		if (copy_from_user(&fid, (const struct lu_fid __user *)arg,
+				   sizeof(fid)))
+			return -EFAULT;
+
+		/* Call mdc_iocontrol */
+		rc = obd_iocontrol(LL_IOC_FID2MDTIDX, exp, sizeof(fid), &fid,
+				   &index);
+		if (rc)
+			return rc;
+
+		return index;
+	}
 	case LL_IOC_HSM_REQUEST: {
 		struct hsm_user_request	*hur;
 		ssize_t			 totalsize;

drivers/staging/lustre/lustre/llite/llite_lib.c

@@ -2373,8 +2373,9 @@ struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
 	op_data->op_bias = 0;
 	op_data->op_cli_flags = 0;
 	if ((opc == LUSTRE_OPC_CREATE) && name &&
-	    filename_is_volatile(name, namelen, NULL))
+	    filename_is_volatile(name, namelen, &op_data->op_mds))
 		op_data->op_bias |= MDS_CREATE_VOLATILE;
+	else
 		op_data->op_mds = 0;
 	op_data->op_data = data;
 

drivers/staging/lustre/lustre/lmv/lmv_fld.c

@@ -47,18 +47,20 @@
 #include "../include/lprocfs_status.h"
 #include "lmv_internal.h"
 
-int lmv_fld_lookup(struct lmv_obd *lmv,
-		   const struct lu_fid *fid,
-		   u32 *mds)
+int lmv_fld_lookup(struct lmv_obd *lmv, const struct lu_fid *fid, u32 *mds)
 {
+	struct obd_device *obd = lmv2obd_dev(lmv);
 	int rc;
 
-	/* FIXME: Currently ZFS still use local seq for ROOT unfortunately, and
+	/*
+	 * FIXME: Currently ZFS still use local seq for ROOT unfortunately, and
 	 * this fid_is_local check should be removed once LU-2240 is fixed
 	 */
-	LASSERTF((fid_seq_in_fldb(fid_seq(fid)) ||
-		  fid_seq_is_local_file(fid_seq(fid))) &&
-		 fid_is_sane(fid), DFID" is insane!\n", PFID(fid));
+	if (!fid_is_sane(fid) || !(fid_seq_in_fldb(fid_seq(fid)) ||
+				   fid_seq_is_local_file(fid_seq(fid)))) {
+		CERROR("%s: invalid FID " DFID "\n", obd->obd_name, PFID(fid));
+		return -EINVAL;
+	}
 
 	rc = fld_client_lookup(&lmv->lmv_fld, fid_seq(fid), mds,
 			       LU_SEQ_RANGE_MDT, NULL);

drivers/staging/lustre/lustre/lmv/lmv_internal.h

@@ -62,6 +62,11 @@ int lmv_revalidate_slaves(struct obd_export *exp, struct mdt_body *mbody,
 			  ldlm_blocking_callback cb_blocking,
 			  int extra_lock_flags);
 
+static inline struct obd_device *lmv2obd_dev(struct lmv_obd *lmv)
+{
+	return container_of0(lmv, struct obd_device, u.lmv);
+}
+
 static inline struct lmv_tgt_desc *
 lmv_get_target(struct lmv_obd *lmv, u32 mdt_idx, int *index)
 {

drivers/staging/lustre/lustre/lmv/lmv_obd.c

@@ -1073,6 +1073,21 @@ static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
 		rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
 		break;
 	}
+	case LL_IOC_FID2MDTIDX: {
+		struct lu_fid *fid = karg;
+		int mdt_index;
+
+		rc = lmv_fld_lookup(lmv, fid, &mdt_index);
+		if (rc)
+			return rc;
+
+		/*
+		 * Note: this is from llite(see ll_dir_ioctl()), @uarg does not
+		 * point to user space memory for FID2MDTIDX.
+		 */
+		*(__u32 *)uarg = mdt_index;
+		break;
+	}
 	case OBD_IOC_FID2PATH: {
 		rc = lmv_fid2path(exp, len, karg, uarg);
 		break;
@@ -1671,13 +1686,44 @@ lmv_locate_target_for_name(struct lmv_obd *lmv, struct lmv_stripe_md *lsm,
  * retval		pointer to the lmv_tgt_desc if succeed.
  *			ERR_PTR(errno) if failed.
  */
-struct lmv_tgt_desc
-*lmv_locate_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
+struct lmv_tgt_desc*
+lmv_locate_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
 	       struct lu_fid *fid)
 {
 	struct lmv_stripe_md *lsm = op_data->op_mea1;
 	struct lmv_tgt_desc *tgt;
 
+	/*
+	 * During creating VOLATILE file, it should honor the mdt
+	 * index if the file under striped dir is being restored, see
+	 * ct_restore().
+	 */
+	if (op_data->op_bias & MDS_CREATE_VOLATILE &&
+	    (int)op_data->op_mds != -1 && lsm) {
+		int i;
+
+		tgt = lmv_get_target(lmv, op_data->op_mds, NULL);
+		if (IS_ERR(tgt))
+			return tgt;
+
+		/* refill the right parent fid */
+		for (i = 0; i < lsm->lsm_md_stripe_count; i++) {
+			struct lmv_oinfo *oinfo;
+
+			oinfo = &lsm->lsm_md_oinfo[i];
+			if (oinfo->lmo_mds == op_data->op_mds) {
+				*fid = oinfo->lmo_fid;
+				break;
+			}
+		}
+
+		/* Hmm, can not find the stripe by mdt_index(op_mds) */
+		if (i == lsm->lsm_md_stripe_count)
+			tgt = ERR_PTR(-EINVAL);
+
+		return tgt;
+	}
+
 	if (!lsm || !op_data->op_namelen) {
 		tgt = lmv_find_target(lmv, fid);
 		if (IS_ERR(tgt))