* Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* page is empty (contains no dir entry) or hash collide with next page.
* After client receives reply, several pages will be integrated into dir page
* in CFS_PAGE_SIZE (if CFS_PAGE_SIZE greater than LU_PAGE_SIZE), and the
- * lu_dirpage for this integrated page will be adjusted.
+ * lu_dirpage for this integrated page will be adjusted. See
+ * lmv_adjust_dirpages().
*
*/
CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p) hash "LPU64"\n",
inode->i_ino, inode->i_generation, inode, hash);
- LASSERT(max_pages > 0 && max_pages <= PTLRPC_MAX_BRW_PAGES);
+ LASSERT(max_pages > 0 && max_pages <= MD_MAX_BRW_PAGES);
OBD_ALLOC(page_pool, sizeof(page) * max_pages);
if (page_pool != NULL) {
offset = hash_x_index(hash, hash64);
- prefetchw(&page->flags);
- ret = ll_add_to_page_cache_lru(page, inode->i_mapping, offset,
- GFP_KERNEL);
+ prefetchw(&page->flags);
+ ret = add_to_page_cache_lru(page, inode->i_mapping, offset,
+ GFP_KERNEL);
if (ret == 0) {
unlock_page(page);
if (ll_pagevec_add(&lru_pvec, page) == 0)
mode = LCK_PR;
rc = md_lock_match(ll_i2sbi(dir)->ll_md_exp, LDLM_FL_BLOCK_GRANTED,
ll_inode2fid(dir), LDLM_IBITS, &policy, mode, &lockh);
- if (!rc) {
- struct ldlm_enqueue_info einfo = { LDLM_IBITS, mode,
- ll_md_blocking_ast, ldlm_completion_ast,
- NULL, NULL, dir };
- struct lookup_intent it = { .it_op = IT_READDIR };
- struct ptlrpc_request *request;
- struct md_op_data *op_data;
-
- op_data = ll_prep_md_op_data(NULL, dir, NULL, NULL, 0, 0,
- LUSTRE_OPC_ANY, NULL);
- if (IS_ERR(op_data))
- return (void *)op_data;
-
- rc = md_enqueue(ll_i2sbi(dir)->ll_md_exp, &einfo, &it,
- op_data, &lockh, NULL, 0, NULL, 0);
-
- ll_finish_md_op_data(op_data);
-
- request = (struct ptlrpc_request *)it.d.lustre.it_data;
- if (request)
- ptlrpc_req_finished(request);
- if (rc < 0) {
- CERROR("lock enqueue: "DFID" at "LPU64": rc %d\n",
- PFID(ll_inode2fid(dir)), hash, rc);
- return ERR_PTR(rc);
- }
+ if (!rc) {
+ struct ldlm_enqueue_info einfo = {.ei_type = LDLM_IBITS,
+ .ei_mode = mode,
+ .ei_cb_bl =
+ ll_md_blocking_ast,
+ .ei_cb_cp =
+ ldlm_completion_ast,
+ .ei_cb_gl = NULL,
+ .ei_cb_wg = NULL,
+ .ei_cbdata = NULL};
+ struct lookup_intent it = { .it_op = IT_READDIR };
+ struct ptlrpc_request *request;
+ struct md_op_data *op_data;
+
+ op_data = ll_prep_md_op_data(NULL, dir, NULL, NULL, 0, 0,
+ LUSTRE_OPC_ANY, NULL);
+ if (IS_ERR(op_data))
+ return (void *)op_data;
+
+ rc = md_enqueue(ll_i2sbi(dir)->ll_md_exp, &einfo, &it,
+ op_data, &lockh, NULL, 0, NULL, 0);
+
+ ll_finish_md_op_data(op_data);
+
+ request = (struct ptlrpc_request *)it.d.lustre.it_data;
+ if (request)
+ ptlrpc_req_finished(request);
+ if (rc < 0) {
+ CERROR("lock enqueue: "DFID" at "LPU64": rc %d\n",
+ PFID(ll_inode2fid(dir)), hash, rc);
+ return ERR_PTR(rc);
+ }
+
+ CDEBUG(D_INODE, "setting lr_lvb_inode to inode %p (%lu/%u)\n",
+ dir, dir->i_ino, dir->i_generation);
+ md_set_lock_data(ll_i2sbi(dir)->ll_md_exp,
+ &it.d.lustre.it_lock_handle, dir, NULL);
} else {
/* for cross-ref object, l_ast_data of the lock may not be set,
* we reset it here */
return rc;
}
+int ll_dir_setdirstripe(struct inode *dir, struct lmv_user_md *lump,
+ char *filename)
+{
+ struct ptlrpc_request *request = NULL;
+ struct md_op_data *op_data;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
+ int mode;
+ int err;
+
+ ENTRY;
+
+ mode = (0755 & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR;
+ op_data = ll_prep_md_op_data(NULL, dir, NULL, filename,
+ strlen(filename), mode, LUSTRE_OPC_MKDIR,
+ lump);
+ if (IS_ERR(op_data))
+ GOTO(err_exit, err = PTR_ERR(op_data));
+
+ op_data->op_cli_flags |= CLI_SET_MEA;
+ err = md_create(sbi->ll_md_exp, op_data, lump, sizeof(*lump), mode,
+ cfs_curproc_fsuid(), cfs_curproc_fsgid(),
+ cfs_curproc_cap_pack(), 0, &request);
+ ll_finish_md_op_data(op_data);
+ if (err)
+ GOTO(err_exit, err);
+err_exit:
+ ptlrpc_req_finished(request);
+ return err;
+}
+
int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
int set_default)
{
lustre_swab_lov_user_md_v1(lump);
lum_size = sizeof(struct lov_user_md_v1);
break;
- }
+ }
case LOV_USER_MAGIC_V3: {
if (lump->lmm_magic != cpu_to_le32(LOV_USER_MAGIC_V3))
lustre_swab_lov_user_md_v3(
(struct lov_user_md_v3 *)lump);
lum_size = sizeof(struct lov_user_md_v3);
break;
- }
+ }
default: {
CDEBUG(D_IOCTL, "bad userland LOV MAGIC:"
" %#08x != %#08x nor %#08x\n",
lump->lmm_magic, LOV_USER_MAGIC_V1,
LOV_USER_MAGIC_V3);
RETURN(-EINVAL);
- }
- }
+ }
+ }
} else {
lum_size = sizeof(struct lov_user_md_v1);
}
if (IS_ERR(op_data))
RETURN(PTR_ERR(op_data));
+ if (lump != NULL && lump->lmm_magic == cpu_to_le32(LMV_USER_MAGIC))
+ op_data->op_cli_flags |= CLI_SET_MEA;
+
/* swabbing is done in lov_setstripe() on server side */
rc = md_setattr(sbi->ll_md_exp, op_data, lump, lum_size,
NULL, 0, &req, NULL);
return mdtidx;
}
+/**
+ * Generic handler to do any pre-copy work.
+ *
+ * It send a first hsm_progress (with extent length == 0) to coordinator as a
+ * first information for it that real work has started.
+ *
+ * Moreover, for a ARCHIVE request, it will sample the file data version and
+ * store it in \a copy.
+ *
+ * \return 0 on success.
+ */
+static int ll_ioc_copy_start(struct super_block *sb, struct hsm_copy *copy)
+{
+ struct ll_sb_info *sbi = ll_s2sbi(sb);
+ struct hsm_progress_kernel hpk;
+ int rc;
+ ENTRY;
+
+ /* Forge a hsm_progress based on data from copy. */
+ hpk.hpk_fid = copy->hc_hai.hai_fid;
+ hpk.hpk_cookie = copy->hc_hai.hai_cookie;
+ hpk.hpk_extent.offset = copy->hc_hai.hai_extent.offset;
+ hpk.hpk_extent.length = 0;
+ hpk.hpk_flags = 0;
+ hpk.hpk_errval = 0;
+ hpk.hpk_data_version = 0;
+
+
+ /* For archive request, we need to read the current file version. */
+ if (copy->hc_hai.hai_action == HSMA_ARCHIVE) {
+ struct inode *inode;
+ __u64 data_version = 0;
+
+ /* Get inode for this fid */
+ inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid);
+ if (IS_ERR(inode)) {
+ hpk.hpk_flags |= HP_FLAG_RETRY;
+ /* hpk_errval is >= 0 */
+ hpk.hpk_errval = -PTR_ERR(inode);
+ GOTO(progress, rc = PTR_ERR(inode));
+ }
+
+ /* Read current file data version */
+ rc = ll_data_version(inode, &data_version, 1);
+ iput(inode);
+ if (rc != 0) {
+ CDEBUG(D_HSM, "Could not read file data version of "
+ DFID" (rc = %d). Archive request ("
+ LPX64") could not be done.\n",
+ PFID(©->hc_hai.hai_fid), rc,
+ copy->hc_hai.hai_cookie);
+ hpk.hpk_flags |= HP_FLAG_RETRY;
+ /* hpk_errval must be >= 0 */
+ hpk.hpk_errval = -rc;
+ GOTO(progress, rc);
+ }
+
+ /* Store it the hsm_copy for later copytool use.
+ * Always modified even if no lsm. */
+ copy->hc_data_version = data_version;
+ }
+
+progress:
+ rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk),
+ &hpk, NULL);
+
+ RETURN(rc);
+}
+
+/**
+ * Generic handler to do any post-copy work.
+ *
+ * It will send the last hsm_progress update to coordinator to inform it
+ * that copy is finished and whether it was successful or not.
+ *
+ * Moreover,
+ * - for ARCHIVE request, it will sample the file data version and compare it
+ * with the version saved in ll_ioc_copy_start(). If they do not match, copy
+ * will be considered as failed.
+ * - for RESTORE request, it will sample the file data version and send it to
+ * coordinator which is useful if the file was imported as 'released'.
+ *
+ * \return 0 on success.
+ */
+static int ll_ioc_copy_end(struct super_block *sb, struct hsm_copy *copy)
+{
+ struct ll_sb_info *sbi = ll_s2sbi(sb);
+ struct hsm_progress_kernel hpk;
+ int rc;
+ ENTRY;
+
+ /* If you modify the logic here, also check llapi_hsm_copy_end(). */
+ /* Take care: copy->hc_hai.hai_action, len, gid and data are not
+ * initialized if copy_end was called with copy == NULL.
+ */
+
+ /* Forge a hsm_progress based on data from copy. */
+ hpk.hpk_fid = copy->hc_hai.hai_fid;
+ hpk.hpk_cookie = copy->hc_hai.hai_cookie;
+ hpk.hpk_extent = copy->hc_hai.hai_extent;
+ hpk.hpk_flags = copy->hc_flags | HP_FLAG_COMPLETED;
+ hpk.hpk_errval = copy->hc_errval;
+ hpk.hpk_data_version = 0;
+
+ /* For archive request, we need to check the file data was not changed.
+ *
+ * For restore request, we need to send the file data version, this is
+ * useful when the file was created using hsm_import.
+ */
+ if (((copy->hc_hai.hai_action == HSMA_ARCHIVE) ||
+ (copy->hc_hai.hai_action == HSMA_RESTORE)) &&
+ (copy->hc_errval == 0)) {
+ struct inode *inode;
+ __u64 data_version = 0;
+
+ /* Get lsm for this fid */
+ inode = search_inode_for_lustre(sb, ©->hc_hai.hai_fid);
+ if (IS_ERR(inode)) {
+ hpk.hpk_flags |= HP_FLAG_RETRY;
+ /* hpk_errval must be >= 0 */
+ hpk.hpk_errval = -PTR_ERR(inode);
+ GOTO(progress, rc = PTR_ERR(inode));
+ }
+
+ rc = ll_data_version(inode, &data_version,
+ copy->hc_hai.hai_action == HSMA_ARCHIVE);
+ iput(inode);
+ if (rc) {
+ CDEBUG(D_HSM, "Could not read file data version. "
+ "Request could not be confirmed.\n");
+ if (hpk.hpk_errval == 0)
+ hpk.hpk_errval = -rc;
+ GOTO(progress, rc);
+ }
+
+ /* Store it the hsm_copy for later copytool use.
+ * Always modified even if no lsm. */
+ hpk.hpk_data_version = data_version;
+
+ /* File could have been stripped during archiving, so we need
+ * to check anyway. */
+ if ((copy->hc_hai.hai_action == HSMA_ARCHIVE) &&
+ (copy->hc_data_version != data_version)) {
+ CDEBUG(D_HSM, "File data version mismatched. "
+ "File content was changed during archiving. "
+ DFID", start:"LPX64" current:"LPX64"\n",
+ PFID(©->hc_hai.hai_fid),
+ copy->hc_data_version, data_version);
+ /* File was changed, send error to cdt. Do not ask for
+ * retry because if a file is modified frequently,
+ * the cdt will loop on retried archive requests.
+ * The policy engine will ask for a new archive later
+ * when the file will not be modified for some tunable
+ * time */
+ /* we do not notify caller */
+ hpk.hpk_flags &= ~HP_FLAG_RETRY;
+ /* hpk_errval must be >= 0 */
+ hpk.hpk_errval = EBUSY;
+ }
+
+ }
+
+progress:
+ rc = obd_iocontrol(LL_IOC_HSM_PROGRESS, sbi->ll_md_exp, sizeof(hpk),
+ &hpk, NULL);
+
+ RETURN(rc);
+}
+
+
static int copy_and_ioctl(int cmd, struct obd_export *exp, void *data, int len)
{
void *ptr;
obd_ioctl_freedata(buf, len);
return rc;
}
+ case LL_IOC_LMV_SETSTRIPE: {
+ struct lmv_user_md *lum;
+ char *buf = NULL;
+ char *filename;
+ int namelen = 0;
+ int lumlen = 0;
+ int len;
+ int rc;
+
+ rc = obd_ioctl_getdata(&buf, &len, (void *)arg);
+ if (rc)
+ RETURN(rc);
+
+ data = (void *)buf;
+ if (data->ioc_inlbuf1 == NULL || data->ioc_inlbuf2 == NULL ||
+ data->ioc_inllen1 == 0 || data->ioc_inllen2 == 0)
+ GOTO(lmv_out_free, rc = -EINVAL);
+
+ filename = data->ioc_inlbuf1;
+ namelen = data->ioc_inllen1;
+
+ if (namelen < 1) {
+ CDEBUG(D_INFO, "IOC_MDC_LOOKUP missing filename\n");
+ GOTO(lmv_out_free, rc = -EINVAL);
+ }
+ lum = (struct lmv_user_md *)data->ioc_inlbuf2;
+ lumlen = data->ioc_inllen2;
+
+ if (lum->lum_magic != LMV_USER_MAGIC ||
+ lumlen != sizeof(*lum)) {
+ CERROR("%s: wrong lum magic %x or size %d: rc = %d\n",
+ filename, lum->lum_magic, lumlen, -EFAULT);
+ GOTO(lmv_out_free, rc = -EINVAL);
+ }
+
+ /**
+ * ll_dir_setdirstripe will be used to set dir stripe
+ * mdc_create--->mdt_reint_create (with dirstripe)
+ */
+ rc = ll_dir_setdirstripe(inode, lum, filename);
+lmv_out_free:
+ obd_ioctl_freedata(buf, len);
+ RETURN(rc);
+
+ }
case LL_IOC_LOV_SETSTRIPE: {
struct lov_user_md_v3 lumv3;
struct lov_user_md_v1 *lumv1 = (struct lov_user_md_v1 *)&lumv3;
if (cfs_copy_from_user(lumv1, lumv1p, sizeof(*lumv1)))
RETURN(-EFAULT);
- if (lumv1->lmm_magic == LOV_USER_MAGIC_V3) {
+ if ((lumv1->lmm_magic == LOV_USER_MAGIC_V3) ) {
if (cfs_copy_from_user(&lumv3, lumv3p, sizeof(lumv3)))
RETURN(-EFAULT);
}
RETURN(rc);
}
+ case LL_IOC_LMV_GETSTRIPE: {
+ struct lmv_user_md *lump = (struct lmv_user_md *)arg;
+ struct lmv_user_md lum;
+ struct lmv_user_md *tmp;
+ int lum_size;
+ int rc = 0;
+ int mdtindex;
+
+ if (copy_from_user(&lum, lump, sizeof(struct lmv_user_md)))
+ RETURN(-EFAULT);
+
+ if (lum.lum_magic != LMV_MAGIC_V1)
+ RETURN(-EINVAL);
+
+ lum_size = lmv_user_md_size(1, LMV_MAGIC_V1);
+ OBD_ALLOC(tmp, lum_size);
+ if (tmp == NULL)
+ GOTO(free_lmv, rc = -ENOMEM);
+
+ memcpy(tmp, &lum, sizeof(lum));
+ tmp->lum_type = LMV_STRIPE_TYPE;
+ tmp->lum_stripe_count = 1;
+ mdtindex = ll_get_mdt_idx(inode);
+ if (mdtindex < 0)
+ GOTO(free_lmv, rc = -ENOMEM);
+
+ tmp->lum_stripe_offset = mdtindex;
+ tmp->lum_objects[0].lum_mds = mdtindex;
+ memcpy(&tmp->lum_objects[0].lum_fid, ll_inode2fid(inode),
+ sizeof(struct lu_fid));
+ if (copy_to_user((void *)arg, tmp, lum_size))
+ GOTO(free_lmv, rc = -EFAULT);
+free_lmv:
+ if (tmp)
+ OBD_FREE(tmp, lum_size);
+ RETURN(rc);
+ }
+ case LL_IOC_REMOVE_ENTRY: {
+ char *filename = NULL;
+ int namelen = 0;
+ int rc;
+
+ /* Here is a little hack to avoid sending REINT_RMENTRY to
+ * unsupported server, which might crash the server(LU-2730),
+ * Because both LVB_TYPE and REINT_RMENTRY will be supported
+ * on 2.4, we use OBD_CONNECT_LVB_TYPE to detect whether the
+ * server will support REINT_RMENTRY XXX*/
+ if (!(exp_connect_flags(sbi->ll_md_exp) & OBD_CONNECT_LVB_TYPE))
+ return -ENOTSUPP;
+
+ filename = getname((const char *)arg);
+ if (IS_ERR(filename))
+ RETURN(PTR_ERR(filename));
+
+ namelen = strlen(filename);
+ if (namelen < 1)
+ GOTO(out_rmdir, rc = -EINVAL);
+
+ rc = ll_rmdir_entry(inode, filename, namelen);
+out_rmdir:
+ if (filename)
+ putname(filename);
+ RETURN(rc);
+ }
+ case LL_IOC_LOV_SWAP_LAYOUTS:
+ RETURN(-EPERM);
case LL_IOC_OBD_STATFS:
RETURN(ll_obd_statfs(inode, (void *)arg));
case LL_IOC_LOV_GETSTRIPE:
RETURN(rc);
case OBD_IOC_FID2PATH:
RETURN(ll_fid2path(inode, (void *)arg));
- case LL_IOC_HSM_CT_START:
- rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void *)arg,
- sizeof(struct lustre_kernelcomm));
- RETURN(rc);
-
- default:
- RETURN(obd_iocontrol(cmd, sbi->ll_dt_exp,0,NULL,(void *)arg));
- }
+ case LL_IOC_HSM_REQUEST: {
+ struct hsm_user_request *hur;
+ int totalsize;
+
+ OBD_ALLOC_PTR(hur);
+ if (hur == NULL)
+ RETURN(-ENOMEM);
+
+ /* We don't know the true size yet; copy the fixed-size part */
+ if (cfs_copy_from_user(hur, (void *)arg, sizeof(*hur))) {
+ OBD_FREE_PTR(hur);
+ RETURN(-EFAULT);
+ }
+
+ /* Compute the whole struct size */
+ totalsize = hur_len(hur);
+ OBD_FREE_PTR(hur);
+ OBD_ALLOC_LARGE(hur, totalsize);
+ if (hur == NULL)
+ RETURN(-ENOMEM);
+
+ /* Copy the whole struct */
+ if (cfs_copy_from_user(hur, (void *)arg, totalsize)) {
+ OBD_FREE_LARGE(hur, totalsize);
+ RETURN(-EFAULT);
+ }
+
+ rc = obd_iocontrol(cmd, ll_i2mdexp(inode), totalsize,
+ hur, NULL);
+
+ OBD_FREE_LARGE(hur, totalsize);
+
+ RETURN(rc);
+ }
+ case LL_IOC_HSM_PROGRESS: {
+ struct hsm_progress_kernel hpk;
+ struct hsm_progress hp;
+
+ if (cfs_copy_from_user(&hp, (void *)arg, sizeof(hp)))
+ RETURN(-EFAULT);
+
+ hpk.hpk_fid = hp.hp_fid;
+ hpk.hpk_cookie = hp.hp_cookie;
+ hpk.hpk_extent = hp.hp_extent;
+ hpk.hpk_flags = hp.hp_flags;
+ hpk.hpk_errval = hp.hp_errval;
+ hpk.hpk_data_version = 0;
+
+ /* File may not exist in Lustre; all progress
+ * reported to Lustre root */
+ rc = obd_iocontrol(cmd, sbi->ll_md_exp, sizeof(hpk), &hpk,
+ NULL);
+ RETURN(rc);
+ }
+ case LL_IOC_HSM_CT_START:
+ rc = copy_and_ioctl(cmd, sbi->ll_md_exp, (void *)arg,
+ sizeof(struct lustre_kernelcomm));
+ RETURN(rc);
+
+ case LL_IOC_HSM_COPY_START: {
+ struct hsm_copy *copy;
+ int rc;
+
+ OBD_ALLOC_PTR(copy);
+ if (copy == NULL)
+ RETURN(-ENOMEM);
+ if (cfs_copy_from_user(copy, (char *)arg, sizeof(*copy))) {
+ OBD_FREE_PTR(copy);
+ RETURN(-EFAULT);
+ }
+
+ rc = ll_ioc_copy_start(inode->i_sb, copy);
+ if (cfs_copy_to_user((char *)arg, copy, sizeof(*copy)))
+ rc = -EFAULT;
+
+ OBD_FREE_PTR(copy);
+ RETURN(rc);
+ }
+ case LL_IOC_HSM_COPY_END: {
+ struct hsm_copy *copy;
+ int rc;
+
+ OBD_ALLOC_PTR(copy);
+ if (copy == NULL)
+ RETURN(-ENOMEM);
+ if (cfs_copy_from_user(copy, (char *)arg, sizeof(*copy))) {
+ OBD_FREE_PTR(copy);
+ RETURN(-EFAULT);
+ }
+
+ rc = ll_ioc_copy_end(inode->i_sb, copy);
+ if (cfs_copy_to_user((char *)arg, copy, sizeof(*copy)))
+ rc = -EFAULT;
+
+ OBD_FREE_PTR(copy);
+ RETURN(rc);
+ }
+ default:
+ RETURN(obd_iocontrol(cmd, sbi->ll_dt_exp, 0, NULL,
+ (void *)arg));
+ }
}
static loff_t ll_dir_seek(struct file *file, loff_t offset, int origin)