From 1be5683b03a766670b3b629bf6bfeab3ca9239d8 Mon Sep 17 00:00:00 2001 From: Marc Eshel Date: Sun, 22 May 2011 19:47:09 +0300 Subject: pnfs: CB_NOTIFY_DEVICEID Note: This functionlaity is incomplete as all layout segments referring to the 'to be removed device id' need to be reaped, and all in flight I/O drained. [use be32 res in nfs4_callback_devicenotify] [use nfs_client to qualify deviceid for cb_notify_deviceid] [use global deviceid cache for CB_NOTIFY_DEVICEID] [refactor device cache _lookup_deviceid] [refactor device cache _find_get_deviceid] Signed-off-by: Benny Halevy [Bug in new global-device-cache code] [layout_driver MUST set free_deviceid_node if using dev-cache] Signed-off-by: Boaz Harrosh Signed-off-by: Benny Halevy --- fs/nfs/callback_proc.c | 47 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 2f41dccea18e..fb5e5b9a97ae 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -241,6 +241,53 @@ static void pnfs_recall_all_layouts(struct nfs_client *clp) do_callback_layoutrecall(clp, &args); } +__be32 nfs4_callback_devicenotify(struct cb_devicenotifyargs *args, + void *dummy, struct cb_process_state *cps) +{ + int i; + __be32 res = 0; + struct nfs_client *clp = cps->clp; + struct nfs_server *server = NULL; + + dprintk("%s: -->\n", __func__); + + if (!clp) { + res = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION); + goto out; + } + + for (i = 0; i < args->ndevs; i++) { + struct cb_devicenotifyitem *dev = &args->devs[i]; + + if (!server || + server->pnfs_curr_ld->id != dev->cbd_layout_type) { + rcu_read_lock(); + list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) + if (server->pnfs_curr_ld && + server->pnfs_curr_ld->id == dev->cbd_layout_type) { + rcu_read_unlock(); + goto found; + } + rcu_read_unlock(); + dprintk("%s: layout type %u not found\n", + __func__, dev->cbd_layout_type); + continue; + } + + found: + if (dev->cbd_notify_type == NOTIFY_DEVICEID4_CHANGE) + dprintk("%s: NOTIFY_DEVICEID4_CHANGE not supported, " + "deleting instead\n", __func__); + nfs4_delete_deviceid(clp, &dev->cbd_dev_id); + } + +out: + kfree(args->devs); + dprintk("%s: exit with status = %u\n", + __func__, be32_to_cpu(res)); + return res; +} + int nfs41_validate_delegation_stateid(struct nfs_delegation *delegation, const nfs4_stateid *stateid) { if (delegation == NULL) -- cgit v1.2.3 From 35c8bb543c9e83197e6375142d1d1c2ee3cf017d Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Tue, 24 May 2011 18:04:02 +0300 Subject: NFSv4.1: use layout driver in global device cache pnfs deviceids are unique per server, per layout type. struct nfs_client is currently used to distinguish deviceids from different nfs servers, yet these may clash between different layout types on the same server. Therefore, use the layout driver associated with each deviceid at insertion time to look it up, unhash, or delete it. Signed-off-by: Benny Halevy --- fs/nfs/callback_proc.c | 2 +- fs/nfs/nfs4filelayout.c | 3 ++- fs/nfs/pnfs.h | 6 +++--- fs/nfs/pnfs_dev.c | 28 +++++++++++++++++----------- 4 files changed, 23 insertions(+), 16 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index fb5e5b9a97ae..c73e7b2fb8e2 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -278,7 +278,7 @@ __be32 nfs4_callback_devicenotify(struct cb_devicenotifyargs *args, if (dev->cbd_notify_type == NOTIFY_DEVICEID4_CHANGE) dprintk("%s: NOTIFY_DEVICEID4_CHANGE not supported, " "deleting instead\n", __func__); - nfs4_delete_deviceid(clp, &dev->cbd_dev_id); + nfs4_delete_deviceid(server->pnfs_curr_ld, clp, &dev->cbd_dev_id); } out: diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index cd289d9b7de7..501a9b86b318 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -441,7 +441,8 @@ filelayout_check_layout(struct pnfs_layout_hdr *lo, } /* find and reference the deviceid */ - d = nfs4_find_get_deviceid(NFS_SERVER(lo->plh_inode)->nfs_client, id); + d = nfs4_find_get_deviceid(NFS_SERVER(lo->plh_inode)->pnfs_curr_ld, + NFS_SERVER(lo->plh_inode)->nfs_client, id); if (d == NULL) { dsaddr = get_device_info(lo->plh_inode, id, gfp_flags); if (dsaddr == NULL) diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index fbd3f7cd9e71..5b083d295334 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -171,9 +171,9 @@ struct nfs4_deviceid_node { }; void nfs4_print_deviceid(const struct nfs4_deviceid *dev_id); -struct nfs4_deviceid_node *nfs4_find_get_deviceid(const struct nfs_client *, const struct nfs4_deviceid *); -struct nfs4_deviceid_node *nfs4_unhash_put_deviceid(const struct nfs_client *, const struct nfs4_deviceid *); -void nfs4_delete_deviceid(const struct nfs_client *, const struct nfs4_deviceid *); +struct nfs4_deviceid_node *nfs4_find_get_deviceid(const struct pnfs_layoutdriver_type *, const struct nfs_client *, const struct nfs4_deviceid *); +struct nfs4_deviceid_node *nfs4_unhash_put_deviceid(const struct pnfs_layoutdriver_type *, const struct nfs_client *, const struct nfs4_deviceid *); +void nfs4_delete_deviceid(const struct pnfs_layoutdriver_type *, const struct nfs_client *, const struct nfs4_deviceid *); void nfs4_init_deviceid_node(struct nfs4_deviceid_node *, const struct pnfs_layoutdriver_type *, const struct nfs_client *, diff --git a/fs/nfs/pnfs_dev.c b/fs/nfs/pnfs_dev.c index 8fd3839df299..c65e133ce9c0 100644 --- a/fs/nfs/pnfs_dev.c +++ b/fs/nfs/pnfs_dev.c @@ -67,14 +67,16 @@ nfs4_deviceid_hash(const struct nfs4_deviceid *id) } static struct nfs4_deviceid_node * -_lookup_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id, +_lookup_deviceid(const struct pnfs_layoutdriver_type *ld, + const struct nfs_client *clp, const struct nfs4_deviceid *id, long hash) { struct nfs4_deviceid_node *d; struct hlist_node *n; hlist_for_each_entry_rcu(d, n, &nfs4_deviceid_cache[hash], node) - if (d->nfs_client == clp && !memcmp(&d->deviceid, id, sizeof(*id))) { + if (d->ld == ld && d->nfs_client == clp && + !memcmp(&d->deviceid, id, sizeof(*id))) { if (atomic_read(&d->ref)) return d; else @@ -90,13 +92,14 @@ _lookup_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id, * @id deviceid to look up */ struct nfs4_deviceid_node * -_find_get_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id, +_find_get_deviceid(const struct pnfs_layoutdriver_type *ld, + const struct nfs_client *clp, const struct nfs4_deviceid *id, long hash) { struct nfs4_deviceid_node *d; rcu_read_lock(); - d = _lookup_deviceid(clp, id, hash); + d = _lookup_deviceid(ld, clp, id, hash); if (d && !atomic_inc_not_zero(&d->ref)) d = NULL; rcu_read_unlock(); @@ -104,9 +107,10 @@ _find_get_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id, } struct nfs4_deviceid_node * -nfs4_find_get_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id) +nfs4_find_get_deviceid(const struct pnfs_layoutdriver_type *ld, + const struct nfs_client *clp, const struct nfs4_deviceid *id) { - return _find_get_deviceid(clp, id, nfs4_deviceid_hash(id)); + return _find_get_deviceid(ld, clp, id, nfs4_deviceid_hash(id)); } EXPORT_SYMBOL_GPL(nfs4_find_get_deviceid); @@ -119,13 +123,14 @@ EXPORT_SYMBOL_GPL(nfs4_find_get_deviceid); * @ret the unhashed node, if found and dereferenced to zero, NULL otherwise. */ struct nfs4_deviceid_node * -nfs4_unhash_put_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id) +nfs4_unhash_put_deviceid(const struct pnfs_layoutdriver_type *ld, + const struct nfs_client *clp, const struct nfs4_deviceid *id) { struct nfs4_deviceid_node *d; spin_lock(&nfs4_deviceid_lock); rcu_read_lock(); - d = _lookup_deviceid(clp, id, nfs4_deviceid_hash(id)); + d = _lookup_deviceid(ld, clp, id, nfs4_deviceid_hash(id)); rcu_read_unlock(); if (!d) { spin_unlock(&nfs4_deviceid_lock); @@ -150,11 +155,12 @@ EXPORT_SYMBOL_GPL(nfs4_unhash_put_deviceid); * @id deviceid to delete */ void -nfs4_delete_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id) +nfs4_delete_deviceid(const struct pnfs_layoutdriver_type *ld, + const struct nfs_client *clp, const struct nfs4_deviceid *id) { struct nfs4_deviceid_node *d; - d = nfs4_unhash_put_deviceid(clp, id); + d = nfs4_unhash_put_deviceid(ld, clp, id); if (!d) return; d->ld->free_deviceid_node(d); @@ -194,7 +200,7 @@ nfs4_insert_deviceid_node(struct nfs4_deviceid_node *new) spin_lock(&nfs4_deviceid_lock); hash = nfs4_deviceid_hash(&new->deviceid); - d = _find_get_deviceid(new->nfs_client, &new->deviceid, hash); + d = _find_get_deviceid(new->ld, new->nfs_client, &new->deviceid, hash); if (d) { spin_unlock(&nfs4_deviceid_lock); return d; -- cgit v1.2.3 From 778b5502fdba5b183553f3f2ef1672ba78ac58b6 Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Sun, 22 May 2011 19:48:02 +0300 Subject: pnfs: Use byte-range for cb_layoutrecall Use recalled range to invalidate particular layout segments in the layout cache. Signed-off-by: Benny Halevy --- fs/nfs/callback_proc.c | 4 ++-- fs/nfs/pnfs.c | 15 +++++++++------ fs/nfs/pnfs.h | 2 +- 3 files changed, 12 insertions(+), 9 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index c73e7b2fb8e2..d4d1954e9bb9 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -139,7 +139,7 @@ static u32 initiate_file_draining(struct nfs_client *clp, spin_lock(&ino->i_lock); if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) || mark_matching_lsegs_invalid(lo, &free_me_list, - args->cbl_range.iomode)) + &args->cbl_range)) rv = NFS4ERR_DELAY; else rv = NFS4ERR_NOMATCHING_LAYOUT; @@ -184,7 +184,7 @@ static u32 initiate_bulk_draining(struct nfs_client *clp, ino = lo->plh_inode; spin_lock(&ino->i_lock); set_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); - if (mark_matching_lsegs_invalid(lo, &free_me_list, range.iomode)) + if (mark_matching_lsegs_invalid(lo, &free_me_list, &range)) rv = NFS4ERR_DELAY; list_del_init(&lo->plh_bulk_recall); spin_unlock(&ino->i_lock); diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 2357ee343f4a..20436a5e76cd 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -321,10 +321,12 @@ lo_seg_intersecting(struct pnfs_layout_range *l1, } static bool -should_free_lseg(u32 lseg_iomode, u32 recall_iomode) +should_free_lseg(struct pnfs_layout_range *lseg_range, + struct pnfs_layout_range *recall_range) { - return (recall_iomode == IOMODE_ANY || - lseg_iomode == recall_iomode); + return (recall_range->iomode == IOMODE_ANY || + lseg_range->iomode == recall_range->iomode) && + lo_seg_intersecting(lseg_range, recall_range); } /* Returns 1 if lseg is removed from list, 0 otherwise */ @@ -355,7 +357,7 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg, int mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - u32 iomode) + struct pnfs_layout_range *recall_range) { struct pnfs_layout_segment *lseg, *next; int invalid = 0, removed = 0; @@ -368,7 +370,8 @@ mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, return 0; } list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list) - if (should_free_lseg(lseg->pls_range.iomode, iomode)) { + if (!recall_range || + should_free_lseg(&lseg->pls_range, recall_range)) { dprintk("%s: freeing lseg %p iomode %d " "offset %llu length %llu\n", __func__, lseg, lseg->pls_range.iomode, lseg->pls_range.offset, @@ -417,7 +420,7 @@ pnfs_destroy_layout(struct nfs_inode *nfsi) lo = nfsi->layout; if (lo) { lo->plh_block_lgets++; /* permanently block new LAYOUTGETs */ - mark_matching_lsegs_invalid(lo, &tmp_list, IOMODE_ANY); + mark_matching_lsegs_invalid(lo, &tmp_list, NULL); } spin_unlock(&nfsi->vfs_inode.i_lock); pnfs_free_lseg_list(&tmp_list); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 78f8a4a171b3..f37ab3539cbf 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -154,7 +154,7 @@ int pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct nfs4_state *open_state); int mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, struct list_head *tmp_list, - u32 iomode); + struct pnfs_layout_range *recall_range); bool pnfs_roc(struct inode *ino); void pnfs_roc_release(struct inode *ino); void pnfs_roc_set_barrier(struct inode *ino, u32 barrier); -- cgit v1.2.3