From a75b9df9d3bfc3cd1083974c045ae31ce5f3434f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 11 May 2011 18:00:51 -0400 Subject: NFSv4.1: Ensure that layoutget uses the correct gfp modes Currently, writebacks may end up recursing back into the filesystem due to GFP_KERNEL direct reclaims in the pnfs subsystem. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4filelayoutdev.c | 34 +++++++++++++++++----------------- 1 file changed, 17 insertions(+), 17 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c index de5350f2b249..db07c7af1395 100644 --- a/fs/nfs/nfs4filelayoutdev.c +++ b/fs/nfs/nfs4filelayoutdev.c @@ -225,11 +225,11 @@ nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) } static struct nfs4_pnfs_ds * -nfs4_pnfs_ds_add(struct inode *inode, u32 ip_addr, u32 port) +nfs4_pnfs_ds_add(struct inode *inode, u32 ip_addr, u32 port, gfp_t gfp_flags) { struct nfs4_pnfs_ds *tmp_ds, *ds; - ds = kzalloc(sizeof(*tmp_ds), GFP_KERNEL); + ds = kzalloc(sizeof(*tmp_ds), gfp_flags); if (!ds) goto out; @@ -261,7 +261,7 @@ out: * Currently only support ipv4, and one multi-path address. */ static struct nfs4_pnfs_ds * -decode_and_add_ds(struct xdr_stream *streamp, struct inode *inode) +decode_and_add_ds(struct xdr_stream *streamp, struct inode *inode, gfp_t gfp_flags) { struct nfs4_pnfs_ds *ds = NULL; char *buf; @@ -303,7 +303,7 @@ decode_and_add_ds(struct xdr_stream *streamp, struct inode *inode) rlen); goto out_err; } - buf = kmalloc(rlen + 1, GFP_KERNEL); + buf = kmalloc(rlen + 1, gfp_flags); if (!buf) { dprintk("%s: Not enough memory\n", __func__); goto out_err; @@ -333,7 +333,7 @@ decode_and_add_ds(struct xdr_stream *streamp, struct inode *inode) sscanf(pstr, "-%d-%d", &tmp[0], &tmp[1]); port = htons((tmp[0] << 8) | (tmp[1])); - ds = nfs4_pnfs_ds_add(inode, ip_addr, port); + ds = nfs4_pnfs_ds_add(inode, ip_addr, port, gfp_flags); dprintk("%s: Decoded address and port %s\n", __func__, buf); out_free: kfree(buf); @@ -343,7 +343,7 @@ out_err: /* Decode opaque device data and return the result */ static struct nfs4_file_layout_dsaddr* -decode_device(struct inode *ino, struct pnfs_device *pdev) +decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) { int i; u32 cnt, num; @@ -362,7 +362,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev) struct page *scratch; /* set up xdr stream */ - scratch = alloc_page(GFP_KERNEL); + scratch = alloc_page(gfp_flags); if (!scratch) goto out_err; @@ -384,7 +384,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev) } /* read stripe indices */ - stripe_indices = kcalloc(cnt, sizeof(u8), GFP_KERNEL); + stripe_indices = kcalloc(cnt, sizeof(u8), gfp_flags); if (!stripe_indices) goto out_err_free_scratch; @@ -423,7 +423,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev) dsaddr = kzalloc(sizeof(*dsaddr) + (sizeof(struct nfs4_pnfs_ds *) * (num - 1)), - GFP_KERNEL); + gfp_flags); if (!dsaddr) goto out_err_free_stripe_indices; @@ -452,7 +452,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev) for (j = 0; j < mp_count; j++) { if (j == 0) { dsaddr->ds_list[i] = decode_and_add_ds(&stream, - ino); + ino, gfp_flags); if (dsaddr->ds_list[i] == NULL) goto out_err_free_deviceid; } else { @@ -503,12 +503,12 @@ out_err: * available devices. */ static struct nfs4_file_layout_dsaddr * -decode_and_add_device(struct inode *inode, struct pnfs_device *dev) +decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_flags) { struct nfs4_file_layout_dsaddr *d, *new; long hash; - new = decode_device(inode, dev); + new = decode_device(inode, dev, gfp_flags); if (!new) { printk(KERN_WARNING "%s: Could not decode or add device\n", __func__); @@ -537,7 +537,7 @@ decode_and_add_device(struct inode *inode, struct pnfs_device *dev) * of available devices, and return it. */ struct nfs4_file_layout_dsaddr * -get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id) +get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags) { struct pnfs_device *pdev = NULL; u32 max_resp_sz; @@ -556,17 +556,17 @@ get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id) dprintk("%s inode %p max_resp_sz %u max_pages %d\n", __func__, inode, max_resp_sz, max_pages); - pdev = kzalloc(sizeof(struct pnfs_device), GFP_KERNEL); + pdev = kzalloc(sizeof(struct pnfs_device), gfp_flags); if (pdev == NULL) return NULL; - pages = kzalloc(max_pages * sizeof(struct page *), GFP_KERNEL); + pages = kzalloc(max_pages * sizeof(struct page *), gfp_flags); if (pages == NULL) { kfree(pdev); return NULL; } for (i = 0; i < max_pages; i++) { - pages[i] = alloc_page(GFP_KERNEL); + pages[i] = alloc_page(gfp_flags); if (!pages[i]) goto out_free; } @@ -587,7 +587,7 @@ get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id) * Found new device, need to decode it and then add it to the * list of known devices for this mountpoint. */ - dsaddr = decode_and_add_device(inode, pdev); + dsaddr = decode_and_add_device(inode, pdev, gfp_flags); out_free: for (i = 0; i < max_pages; i++) __free_page(pages[i]); -- cgit v1.2.3 From 67d51f65bde233b17de304baec4f7c4d086471fe Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Fri, 20 May 2011 10:45:05 +0200 Subject: NFSv4.1: use struct nfs_client to qualify deviceid deviceids are unique per server, per layout type. Therefore, in the global cache in the files layout driver deviceids from different servers may clash so we need to qualify them with a struct nfs_client that represents the nfs server that returned the deviceid. Introduced in 2.6.39 commit ea8eecdd "NFSv4.1 move deviceid cache to filelayout driver" Signed-off-by: Benny Halevy --- fs/nfs/nfs4filelayout.c | 2 +- fs/nfs/nfs4filelayout.h | 3 ++- fs/nfs/nfs4filelayoutdev.c | 9 ++++----- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index dd6ccf007673..571c1b032e9d 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -440,7 +440,7 @@ filelayout_check_layout(struct pnfs_layout_hdr *lo, } /* find and reference the deviceid */ - dsaddr = nfs4_fl_find_get_deviceid(id); + dsaddr = nfs4_fl_find_get_deviceid(NFS_SERVER(lo->plh_inode)->nfs_client, id); if (dsaddr == NULL) { dsaddr = get_device_info(lo->plh_inode, id, gfp_flags); if (dsaddr == NULL) diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h index 2b461d77b43a..301b95568bd2 100644 --- a/fs/nfs/nfs4filelayout.h +++ b/fs/nfs/nfs4filelayout.h @@ -60,6 +60,7 @@ struct nfs4_pnfs_ds { struct nfs4_file_layout_dsaddr { struct hlist_node node; + struct nfs_client *nfs_client; struct nfs4_deviceid deviceid; atomic_t ref; unsigned long flags; @@ -101,7 +102,7 @@ u32 nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j); struct nfs4_pnfs_ds *nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx); extern struct nfs4_file_layout_dsaddr * -nfs4_fl_find_get_deviceid(struct nfs4_deviceid *dev_id); +nfs4_fl_find_get_deviceid(struct nfs_client *, struct nfs4_deviceid *dev_id); extern void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr); struct nfs4_file_layout_dsaddr * get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags); diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c index db07c7af1395..42e326637271 100644 --- a/fs/nfs/nfs4filelayoutdev.c +++ b/fs/nfs/nfs4filelayoutdev.c @@ -431,7 +431,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) dsaddr->stripe_indices = stripe_indices; stripe_indices = NULL; dsaddr->ds_num = num; - + dsaddr->nfs_client = NFS_SERVER(ino)->nfs_client; memcpy(&dsaddr->deviceid, &pdev->dev_id, sizeof(pdev->dev_id)); for (i = 0; i < dsaddr->ds_num; i++) { @@ -516,7 +516,7 @@ decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_fl } spin_lock(&filelayout_deviceid_lock); - d = nfs4_fl_find_get_deviceid(&new->deviceid); + d = nfs4_fl_find_get_deviceid(new->nfs_client, &new->deviceid); if (d) { spin_unlock(&filelayout_deviceid_lock); nfs4_fl_free_deviceid(new); @@ -610,16 +610,15 @@ nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) } struct nfs4_file_layout_dsaddr * -nfs4_fl_find_get_deviceid(struct nfs4_deviceid *id) +nfs4_fl_find_get_deviceid(struct nfs_client *clp, struct nfs4_deviceid *id) { struct nfs4_file_layout_dsaddr *d; struct hlist_node *n; long hash = nfs4_fl_deviceid_hash(id); - rcu_read_lock(); hlist_for_each_entry_rcu(d, n, &filelayout_deviceid_cache[hash], node) { - if (!memcmp(&d->deviceid, id, sizeof(*id))) { + if (d->nfs_client == clp && !memcmp(&d->deviceid, id, sizeof(*id))) { if (!atomic_inc_not_zero(&d->ref)) goto fail; rcu_read_unlock(); -- cgit v1.2.3 From a1eaecbc4c8307e27772d6584ef85a2e93250661 Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Thu, 19 May 2011 22:14:47 -0400 Subject: NFSv4.1: make deviceid cache global Move deviceid cache from the pnfs files layout driver to the generic layer in preparation for the objects layout driver. Signed-off-by: Benny Halevy --- fs/nfs/Makefile | 2 +- fs/nfs/nfs4filelayout.c | 10 +-- fs/nfs/nfs4filelayout.h | 8 +-- fs/nfs/nfs4filelayoutdev.c | 104 ++++-------------------------- fs/nfs/pnfs.h | 17 +++++ fs/nfs/pnfs_dev.c | 156 +++++++++++++++++++++++++++++++++++++++++++++ 6 files changed, 194 insertions(+), 103 deletions(-) create mode 100644 fs/nfs/pnfs_dev.c (limited to 'fs/nfs/nfs4filelayoutdev.c') diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile index 4776ff9e3814..7516a8aaa42e 100644 --- a/fs/nfs/Makefile +++ b/fs/nfs/Makefile @@ -15,7 +15,7 @@ nfs-$(CONFIG_NFS_V4) += nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \ delegation.o idmap.o \ callback.o callback_xdr.o callback_proc.o \ nfs4namespace.o -nfs-$(CONFIG_NFS_V4_1) += pnfs.o +nfs-$(CONFIG_NFS_V4_1) += pnfs.o pnfs_dev.o nfs-$(CONFIG_SYSCTL) += sysctl.o nfs-$(CONFIG_NFS_FSCACHE) += fscache.o fscache-index.o diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 571c1b032e9d..4c67a6f6519d 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -421,6 +421,7 @@ filelayout_check_layout(struct pnfs_layout_hdr *lo, struct nfs4_deviceid *id, gfp_t gfp_flags) { + struct nfs4_deviceid_node *d; struct nfs4_file_layout_dsaddr *dsaddr; int status = -EINVAL; struct nfs_server *nfss = NFS_SERVER(lo->plh_inode); @@ -440,12 +441,13 @@ filelayout_check_layout(struct pnfs_layout_hdr *lo, } /* find and reference the deviceid */ - dsaddr = nfs4_fl_find_get_deviceid(NFS_SERVER(lo->plh_inode)->nfs_client, id); - if (dsaddr == NULL) { + d = nfs4_find_get_deviceid(NFS_SERVER(lo->plh_inode)->nfs_client, id); + if (d == NULL) { dsaddr = get_device_info(lo->plh_inode, id, gfp_flags); if (dsaddr == NULL) goto out; - } + } else + dsaddr = container_of(d, struct nfs4_file_layout_dsaddr, id_node); fl->dsaddr = dsaddr; if (fl->first_stripe_index < 0 || @@ -535,7 +537,7 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo, memcpy(id, p, sizeof(*id)); p += XDR_QUADLEN(NFS4_DEVICEID4_SIZE); - print_deviceid(id); + nfs4_print_deviceid(id); nfl_util = be32_to_cpup(p++); if (nfl_util & NFL4_UFLG_COMMIT_THRU_MDS) diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h index 301b95568bd2..0ace0a2d6d7d 100644 --- a/fs/nfs/nfs4filelayout.h +++ b/fs/nfs/nfs4filelayout.h @@ -59,10 +59,7 @@ struct nfs4_pnfs_ds { #define NFS4_DEVICE_ID_NEG_ENTRY 0x00000001 struct nfs4_file_layout_dsaddr { - struct hlist_node node; - struct nfs_client *nfs_client; - struct nfs4_deviceid deviceid; - atomic_t ref; + struct nfs4_deviceid_node id_node; unsigned long flags; u32 stripe_count; u8 *stripe_indices; @@ -96,13 +93,10 @@ extern struct nfs_fh * nfs4_fl_select_ds_fh(struct pnfs_layout_segment *lseg, u32 j); extern void print_ds(struct nfs4_pnfs_ds *ds); -extern void print_deviceid(struct nfs4_deviceid *dev_id); u32 nfs4_fl_calc_j_index(struct pnfs_layout_segment *lseg, loff_t offset); u32 nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j); struct nfs4_pnfs_ds *nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx); -extern struct nfs4_file_layout_dsaddr * -nfs4_fl_find_get_deviceid(struct nfs_client *, struct nfs4_deviceid *dev_id); extern void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr); struct nfs4_file_layout_dsaddr * get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags); diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c index 42e326637271..eda4527a57eb 100644 --- a/fs/nfs/nfs4filelayoutdev.c +++ b/fs/nfs/nfs4filelayoutdev.c @@ -36,30 +36,6 @@ #define NFSDBG_FACILITY NFSDBG_PNFS_LD -/* - * Device ID RCU cache. A device ID is unique per client ID and layout type. - */ -#define NFS4_FL_DEVICE_ID_HASH_BITS 5 -#define NFS4_FL_DEVICE_ID_HASH_SIZE (1 << NFS4_FL_DEVICE_ID_HASH_BITS) -#define NFS4_FL_DEVICE_ID_HASH_MASK (NFS4_FL_DEVICE_ID_HASH_SIZE - 1) - -static inline u32 -nfs4_fl_deviceid_hash(struct nfs4_deviceid *id) -{ - unsigned char *cptr = (unsigned char *)id->data; - unsigned int nbytes = NFS4_DEVICEID4_SIZE; - u32 x = 0; - - while (nbytes--) { - x *= 37; - x += *cptr++; - } - return x & NFS4_FL_DEVICE_ID_HASH_MASK; -} - -static struct hlist_head filelayout_deviceid_cache[NFS4_FL_DEVICE_ID_HASH_SIZE]; -static DEFINE_SPINLOCK(filelayout_deviceid_lock); - /* * Data server cache * @@ -89,27 +65,6 @@ print_ds(struct nfs4_pnfs_ds *ds) ds->ds_clp ? ds->ds_clp->cl_exchange_flags : 0); } -void -print_ds_list(struct nfs4_file_layout_dsaddr *dsaddr) -{ - int i; - - ifdebug(FACILITY) { - printk("%s dsaddr->ds_num %d\n", __func__, - dsaddr->ds_num); - for (i = 0; i < dsaddr->ds_num; i++) - print_ds(dsaddr->ds_list[i]); - } -} - -void print_deviceid(struct nfs4_deviceid *id) -{ - u32 *p = (u32 *)id; - - dprintk("%s: device id= [%x%x%x%x]\n", __func__, - p[0], p[1], p[2], p[3]); -} - /* nfs4_ds_cache_lock is held */ static struct nfs4_pnfs_ds * _data_server_lookup_locked(u32 ip_addr, u32 port) @@ -207,7 +162,7 @@ nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) struct nfs4_pnfs_ds *ds; int i; - print_deviceid(&dsaddr->deviceid); + nfs4_print_deviceid(&dsaddr->id_node.deviceid); for (i = 0; i < dsaddr->ds_num; i++) { ds = dsaddr->ds_list[i]; @@ -431,8 +386,8 @@ decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) dsaddr->stripe_indices = stripe_indices; stripe_indices = NULL; dsaddr->ds_num = num; - dsaddr->nfs_client = NFS_SERVER(ino)->nfs_client; - memcpy(&dsaddr->deviceid, &pdev->dev_id, sizeof(pdev->dev_id)); + nfs4_init_deviceid_node(&dsaddr->id_node, NFS_SERVER(ino)->nfs_client, + &pdev->dev_id); for (i = 0; i < dsaddr->ds_num; i++) { int j; @@ -505,8 +460,8 @@ out_err: static struct nfs4_file_layout_dsaddr * decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_flags) { - struct nfs4_file_layout_dsaddr *d, *new; - long hash; + struct nfs4_deviceid_node *d; + struct nfs4_file_layout_dsaddr *n, *new; new = decode_device(inode, dev, gfp_flags); if (!new) { @@ -515,20 +470,13 @@ decode_and_add_device(struct inode *inode, struct pnfs_device *dev, gfp_t gfp_fl return NULL; } - spin_lock(&filelayout_deviceid_lock); - d = nfs4_fl_find_get_deviceid(new->nfs_client, &new->deviceid); - if (d) { - spin_unlock(&filelayout_deviceid_lock); + d = nfs4_insert_deviceid_node(&new->id_node); + n = container_of(d, struct nfs4_file_layout_dsaddr, id_node); + if (n != new) { nfs4_fl_free_deviceid(new); - return d; + return n; } - INIT_HLIST_NODE(&new->node); - atomic_set(&new->ref, 1); - hash = nfs4_fl_deviceid_hash(&new->deviceid); - hlist_add_head_rcu(&new->node, &filelayout_deviceid_cache[hash]); - spin_unlock(&filelayout_deviceid_lock); - return new; } @@ -600,34 +548,8 @@ out_free: void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) { - if (atomic_dec_and_lock(&dsaddr->ref, &filelayout_deviceid_lock)) { - hlist_del_rcu(&dsaddr->node); - spin_unlock(&filelayout_deviceid_lock); - - synchronize_rcu(); + if (nfs4_put_deviceid_node(&dsaddr->id_node)) nfs4_fl_free_deviceid(dsaddr); - } -} - -struct nfs4_file_layout_dsaddr * -nfs4_fl_find_get_deviceid(struct nfs_client *clp, struct nfs4_deviceid *id) -{ - struct nfs4_file_layout_dsaddr *d; - struct hlist_node *n; - long hash = nfs4_fl_deviceid_hash(id); - - rcu_read_lock(); - hlist_for_each_entry_rcu(d, n, &filelayout_deviceid_cache[hash], node) { - if (d->nfs_client == clp && !memcmp(&d->deviceid, id, sizeof(*id))) { - if (!atomic_inc_not_zero(&d->ref)) - goto fail; - rcu_read_unlock(); - return d; - } - } -fail: - rcu_read_unlock(); - return NULL; } /* @@ -675,15 +597,15 @@ static void filelayout_mark_devid_negative(struct nfs4_file_layout_dsaddr *dsaddr, int err, u32 ds_addr) { - u32 *p = (u32 *)&dsaddr->deviceid; + u32 *p = (u32 *)&dsaddr->id_node.deviceid; printk(KERN_ERR "NFS: data server %x connection error %d." " Deviceid [%x%x%x%x] marked out of use.\n", ds_addr, err, p[0], p[1], p[2], p[3]); - spin_lock(&filelayout_deviceid_lock); + spin_lock(&nfs4_ds_cache_lock); dsaddr->flags |= NFS4_DEVICE_ID_NEG_ENTRY; - spin_unlock(&filelayout_deviceid_lock); + spin_unlock(&nfs4_ds_cache_lock); } struct nfs4_pnfs_ds * diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 720bb9da3f9b..3831ad04a231 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -157,6 +157,23 @@ bool pnfs_roc_drain(struct inode *ino, u32 *barrier); void pnfs_set_layoutcommit(struct nfs_write_data *wdata); int pnfs_layoutcommit_inode(struct inode *inode, bool sync); +/* pnfs_dev.c */ +struct nfs4_deviceid_node { + struct hlist_node node; + const struct nfs_client *nfs_client; + struct nfs4_deviceid deviceid; + atomic_t ref; +}; + +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_init_deviceid_node(struct nfs4_deviceid_node *, + const struct nfs_client *, + const struct nfs4_deviceid *); +struct nfs4_deviceid_node *nfs4_insert_deviceid_node(struct nfs4_deviceid_node *); +bool nfs4_put_deviceid_node(struct nfs4_deviceid_node *); + static inline int lo_fail_bit(u32 iomode) { return iomode == IOMODE_RW ? diff --git a/fs/nfs/pnfs_dev.c b/fs/nfs/pnfs_dev.c new file mode 100644 index 000000000000..bf05189a7cbf --- /dev/null +++ b/fs/nfs/pnfs_dev.c @@ -0,0 +1,156 @@ +/* + * Device operations for the pnfs client. + * + * Copyright (c) 2002 + * The Regents of the University of Michigan + * All Rights Reserved + * + * Dean Hildebrand + * Garth Goodson + * + * Permission is granted to use, copy, create derivative works, and + * redistribute this software and such derivative works for any purpose, + * so long as the name of the University of Michigan is not used in + * any advertising or publicity pertaining to the use or distribution + * of this software without specific, written prior authorization. If + * the above copyright notice or any other identification of the + * University of Michigan is included in any copy of any portion of + * this software, then the disclaimer below must also be included. + * + * This software is provided as is, without representation or warranty + * of any kind either express or implied, including without limitation + * the implied warranties of merchantability, fitness for a particular + * purpose, or noninfringement. The Regents of the University of + * Michigan shall not be liable for any damages, including special, + * indirect, incidental, or consequential damages, with respect to any + * claim arising out of or in connection with the use of the software, + * even if it has been or is hereafter advised of the possibility of + * such damages. + */ + +#include "pnfs.h" + +#define NFSDBG_FACILITY NFSDBG_PNFS + +/* + * Device ID RCU cache. A device ID is unique per server and layout type. + */ +#define NFS4_DEVICE_ID_HASH_BITS 5 +#define NFS4_DEVICE_ID_HASH_SIZE (1 << NFS4_DEVICE_ID_HASH_BITS) +#define NFS4_DEVICE_ID_HASH_MASK (NFS4_DEVICE_ID_HASH_SIZE - 1) + +static struct hlist_head nfs4_deviceid_cache[NFS4_DEVICE_ID_HASH_SIZE]; +static DEFINE_SPINLOCK(nfs4_deviceid_lock); + +void +nfs4_print_deviceid(const struct nfs4_deviceid *id) +{ + u32 *p = (u32 *)id; + + dprintk("%s: device id= [%x%x%x%x]\n", __func__, + p[0], p[1], p[2], p[3]); +} +EXPORT_SYMBOL_GPL(nfs4_print_deviceid); + +static inline u32 +nfs4_deviceid_hash(const struct nfs4_deviceid *id) +{ + unsigned char *cptr = (unsigned char *)id->data; + unsigned int nbytes = NFS4_DEVICEID4_SIZE; + u32 x = 0; + + while (nbytes--) { + x *= 37; + x += *cptr++; + } + return x & NFS4_DEVICE_ID_HASH_MASK; +} + +/* + * Lookup a deviceid in cache and get a reference count on it if found + * + * @clp nfs_client associated with deviceid + * @id deviceid to look up + */ +struct nfs4_deviceid_node * +nfs4_find_get_deviceid(const struct nfs_client *clp, const struct nfs4_deviceid *id) +{ + struct nfs4_deviceid_node *d; + struct hlist_node *n; + long hash = nfs4_deviceid_hash(id); + + rcu_read_lock(); + hlist_for_each_entry_rcu(d, n, &nfs4_deviceid_cache[hash], node) { + if (d->nfs_client == clp && !memcmp(&d->deviceid, id, sizeof(*id))) { + if (!atomic_inc_not_zero(&d->ref)) + goto fail; + rcu_read_unlock(); + return d; + } + } +fail: + rcu_read_unlock(); + return NULL; +} +EXPORT_SYMBOL_GPL(nfs4_find_get_deviceid); + +void +nfs4_init_deviceid_node(struct nfs4_deviceid_node *d, + const struct nfs_client *nfs_client, + const struct nfs4_deviceid *id) +{ + d->nfs_client = nfs_client; + d->deviceid = *id; +} +EXPORT_SYMBOL_GPL(nfs4_init_deviceid_node); + +/* + * Uniquely initialize and insert a deviceid node into cache + * + * @new new deviceid node + * Note that the caller must set up new->nfs_client and new->deviceid + * + * @ret the inserted node, if none found, otherwise, the found entry. + */ +struct nfs4_deviceid_node * +nfs4_insert_deviceid_node(struct nfs4_deviceid_node *new) +{ + struct nfs4_deviceid_node *d; + long hash; + + spin_lock(&nfs4_deviceid_lock); + d = nfs4_find_get_deviceid(new->nfs_client, &new->deviceid); + if (d) { + spin_unlock(&nfs4_deviceid_lock); + return d; + } + + INIT_HLIST_NODE(&new->node); + atomic_set(&new->ref, 1); + hash = nfs4_deviceid_hash(&new->deviceid); + hlist_add_head_rcu(&new->node, &nfs4_deviceid_cache[hash]); + spin_unlock(&nfs4_deviceid_lock); + + return new; +} +EXPORT_SYMBOL_GPL(nfs4_insert_deviceid_node); + +/* + * Dereference a deviceid node and delete it when its reference count drops + * to zero. + * + * @d deviceid node to put + * + * @ret true iff the node was deleted + */ +bool +nfs4_put_deviceid_node(struct nfs4_deviceid_node *d) +{ + if (!atomic_dec_and_lock(&d->ref, &nfs4_deviceid_lock)) + return false; + hlist_del_init_rcu(&d->node); + spin_unlock(&nfs4_deviceid_lock); + synchronize_rcu(); + return true; +} +EXPORT_SYMBOL_GPL(nfs4_put_deviceid_node); -- cgit v1.2.3 From 1775bc342c6eacd6304493cbb2e0cda1a0182246 Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Fri, 20 May 2011 13:47:33 +0200 Subject: NFSv4.1: purge deviceid cache on nfs_free_client Use the pnfs_layoutdriver_type both as a qualifier for the deviceid, distinguishing deviceid from different layout types on the server, and for freeing the layout-driver allocated structure containing the nfs4_deviceid_node. [BUG in _deviceid_purge_client] [layout_driver MUST set free_deviceid_node if using dev-cache] [let ver < 4.1 compile] Signed-off-by: Boaz Harrosh [removed EXPORT_SYMBOL_GPL(nfs4_deviceid_purge_client)] Signed-off-by: Benny Halevy --- fs/nfs/client.c | 2 ++ fs/nfs/nfs4filelayout.c | 7 +++++++ fs/nfs/nfs4filelayout.h | 1 + fs/nfs/nfs4filelayoutdev.c | 9 +++++---- fs/nfs/pnfs.h | 11 +++++++++++ fs/nfs/pnfs_dev.c | 47 +++++++++++++++++++++++++++++++++++++++++++--- 6 files changed, 70 insertions(+), 7 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 139be9647d80..b3dc2b88b65b 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -290,6 +290,8 @@ static void nfs_free_client(struct nfs_client *clp) if (clp->cl_machine_cred != NULL) put_rpccred(clp->cl_machine_cred); + nfs4_deviceid_purge_client(clp); + kfree(clp->cl_hostname); kfree(clp); diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 4c67a6f6519d..cd289d9b7de7 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -862,6 +862,12 @@ filelayout_commit_pagelist(struct inode *inode, struct list_head *mds_pages, return -ENOMEM; } +static void +filelayout_free_deveiceid_node(struct nfs4_deviceid_node *d) +{ + nfs4_fl_free_deviceid(container_of(d, struct nfs4_file_layout_dsaddr, id_node)); +} + static struct pnfs_layoutdriver_type filelayout_type = { .id = LAYOUT_NFSV4_1_FILES, .name = "LAYOUT_NFSV4_1_FILES", @@ -874,6 +880,7 @@ static struct pnfs_layoutdriver_type filelayout_type = { .commit_pagelist = filelayout_commit_pagelist, .read_pagelist = filelayout_read_pagelist, .write_pagelist = filelayout_write_pagelist, + .free_deviceid_node = filelayout_free_deveiceid_node, }; static int __init nfs4filelayout_init(void) diff --git a/fs/nfs/nfs4filelayout.h b/fs/nfs/nfs4filelayout.h index 0ace0a2d6d7d..cebe01e3795e 100644 --- a/fs/nfs/nfs4filelayout.h +++ b/fs/nfs/nfs4filelayout.h @@ -98,6 +98,7 @@ u32 nfs4_fl_calc_ds_index(struct pnfs_layout_segment *lseg, u32 j); struct nfs4_pnfs_ds *nfs4_fl_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx); extern void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr); +extern void nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr); struct nfs4_file_layout_dsaddr * get_device_info(struct inode *inode, struct nfs4_deviceid *dev_id, gfp_t gfp_flags); diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c index eda4527a57eb..5914659c8ec5 100644 --- a/fs/nfs/nfs4filelayoutdev.c +++ b/fs/nfs/nfs4filelayoutdev.c @@ -156,7 +156,7 @@ destroy_ds(struct nfs4_pnfs_ds *ds) kfree(ds); } -static void +void nfs4_fl_free_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) { struct nfs4_pnfs_ds *ds; @@ -386,7 +386,9 @@ decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) dsaddr->stripe_indices = stripe_indices; stripe_indices = NULL; dsaddr->ds_num = num; - nfs4_init_deviceid_node(&dsaddr->id_node, NFS_SERVER(ino)->nfs_client, + nfs4_init_deviceid_node(&dsaddr->id_node, + NFS_SERVER(ino)->pnfs_curr_ld, + NFS_SERVER(ino)->nfs_client, &pdev->dev_id); for (i = 0; i < dsaddr->ds_num; i++) { @@ -548,8 +550,7 @@ out_free: void nfs4_fl_put_deviceid(struct nfs4_file_layout_dsaddr *dsaddr) { - if (nfs4_put_deviceid_node(&dsaddr->id_node)) - nfs4_fl_free_deviceid(dsaddr); + nfs4_put_deviceid_node(&dsaddr->id_node); } /* diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 3831ad04a231..80a5d0e2cc43 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -65,6 +65,8 @@ enum { NFS_LAYOUT_DESTROYED, /* no new use of layout allowed */ }; +struct nfs4_deviceid_node; + /* Per-layout driver specific registration structure */ struct pnfs_layoutdriver_type { struct list_head pnfs_tblid; @@ -90,6 +92,8 @@ struct pnfs_layoutdriver_type { */ enum pnfs_try_status (*read_pagelist) (struct nfs_read_data *nfs_data); enum pnfs_try_status (*write_pagelist) (struct nfs_write_data *nfs_data, int how); + + void (*free_deviceid_node) (struct nfs4_deviceid_node *); }; struct pnfs_layout_hdr { @@ -160,6 +164,7 @@ int pnfs_layoutcommit_inode(struct inode *inode, bool sync); /* pnfs_dev.c */ struct nfs4_deviceid_node { struct hlist_node node; + const struct pnfs_layoutdriver_type *ld; const struct nfs_client *nfs_client; struct nfs4_deviceid deviceid; atomic_t ref; @@ -169,10 +174,12 @@ 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_init_deviceid_node(struct nfs4_deviceid_node *, + const struct pnfs_layoutdriver_type *, const struct nfs_client *, const struct nfs4_deviceid *); struct nfs4_deviceid_node *nfs4_insert_deviceid_node(struct nfs4_deviceid_node *); bool nfs4_put_deviceid_node(struct nfs4_deviceid_node *); +void nfs4_deviceid_purge_client(const struct nfs_client *); static inline int lo_fail_bit(u32 iomode) { @@ -349,6 +356,10 @@ static inline int pnfs_layoutcommit_inode(struct inode *inode, bool sync) { return 0; } + +static inline void nfs4_deviceid_purge_client(struct nfs_client *ncl) +{ +} #endif /* CONFIG_NFS_V4_1 */ #endif /* FS_NFS_PNFS_H */ diff --git a/fs/nfs/pnfs_dev.c b/fs/nfs/pnfs_dev.c index bf05189a7cbf..64a4b85c7dbc 100644 --- a/fs/nfs/pnfs_dev.c +++ b/fs/nfs/pnfs_dev.c @@ -96,11 +96,15 @@ EXPORT_SYMBOL_GPL(nfs4_find_get_deviceid); void nfs4_init_deviceid_node(struct nfs4_deviceid_node *d, + const struct pnfs_layoutdriver_type *ld, const struct nfs_client *nfs_client, const struct nfs4_deviceid *id) { + INIT_HLIST_NODE(&d->node); + d->ld = ld; d->nfs_client = nfs_client; d->deviceid = *id; + atomic_set(&d->ref, 1); } EXPORT_SYMBOL_GPL(nfs4_init_deviceid_node); @@ -108,7 +112,10 @@ EXPORT_SYMBOL_GPL(nfs4_init_deviceid_node); * Uniquely initialize and insert a deviceid node into cache * * @new new deviceid node - * Note that the caller must set up new->nfs_client and new->deviceid + * Note that the caller must set up the following members: + * new->ld + * new->nfs_client + * new->deviceid * * @ret the inserted node, if none found, otherwise, the found entry. */ @@ -125,8 +132,6 @@ nfs4_insert_deviceid_node(struct nfs4_deviceid_node *new) return d; } - INIT_HLIST_NODE(&new->node); - atomic_set(&new->ref, 1); hash = nfs4_deviceid_hash(&new->deviceid); hlist_add_head_rcu(&new->node, &nfs4_deviceid_cache[hash]); spin_unlock(&nfs4_deviceid_lock); @@ -151,6 +156,42 @@ nfs4_put_deviceid_node(struct nfs4_deviceid_node *d) hlist_del_init_rcu(&d->node); spin_unlock(&nfs4_deviceid_lock); synchronize_rcu(); + d->ld->free_deviceid_node(d); return true; } EXPORT_SYMBOL_GPL(nfs4_put_deviceid_node); + +static void +_deviceid_purge_client(const struct nfs_client *clp, long hash) +{ + struct nfs4_deviceid_node *d; + struct hlist_node *n, *next; + HLIST_HEAD(tmp); + + rcu_read_lock(); + hlist_for_each_entry_rcu(d, n, &nfs4_deviceid_cache[hash], node) + if (d->nfs_client == clp && atomic_read(&d->ref)) { + hlist_del_init_rcu(&d->node); + hlist_add_head(&d->node, &tmp); + } + rcu_read_unlock(); + + if (hlist_empty(&tmp)) + return; + + synchronize_rcu(); + hlist_for_each_entry_safe(d, n, next, &tmp, node) + if (atomic_dec_and_test(&d->ref)) + d->ld->free_deviceid_node(d); +} + +void +nfs4_deviceid_purge_client(const struct nfs_client *clp) +{ + long h; + + spin_lock(&nfs4_deviceid_lock); + for (h = 0; h < NFS4_DEVICE_ID_HASH_SIZE; h++) + _deviceid_purge_client(clp, h); + spin_unlock(&nfs4_deviceid_lock); +} -- cgit v1.2.3 From f7da7a129d57bfe0f74573dc03531c63e1360fae Mon Sep 17 00:00:00 2001 From: Benny Halevy Date: Thu, 19 May 2011 14:16:47 -0400 Subject: SUNRPC: introduce xdr_init_decode_pages Initialize xdr_stream and xdr_buf using an array of page pointers and length of buffer. Signed-off-by: Benny Halevy --- fs/nfs/dir.c | 9 ++------- fs/nfs/nfs4filelayout.c | 9 ++------- fs/nfs/nfs4filelayoutdev.c | 9 ++------- include/linux/sunrpc/xdr.h | 2 ++ net/sunrpc/xdr.c | 19 +++++++++++++++++++ 5 files changed, 27 insertions(+), 21 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 7237672216c8..f673a9e1d95d 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -512,12 +512,7 @@ int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *en struct page **xdr_pages, struct page *page, unsigned int buflen) { struct xdr_stream stream; - struct xdr_buf buf = { - .pages = xdr_pages, - .page_len = buflen, - .buflen = buflen, - .len = buflen, - }; + struct xdr_buf buf; struct page *scratch; struct nfs_cache_array *array; unsigned int count = 0; @@ -527,7 +522,7 @@ int nfs_readdir_page_filler(nfs_readdir_descriptor_t *desc, struct nfs_entry *en if (scratch == NULL) return -ENOMEM; - xdr_init_decode(&stream, &buf, NULL); + xdr_init_decode_pages(&stream, &buf, xdr_pages, buflen); xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); do { diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c index 501a9b86b318..33bda24e8cd2 100644 --- a/fs/nfs/nfs4filelayout.c +++ b/fs/nfs/nfs4filelayout.c @@ -510,12 +510,7 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo, gfp_t gfp_flags) { struct xdr_stream stream; - struct xdr_buf buf = { - .pages = lgr->layoutp->pages, - .page_len = lgr->layoutp->len, - .buflen = lgr->layoutp->len, - .len = lgr->layoutp->len, - }; + struct xdr_buf buf; struct page *scratch; __be32 *p; uint32_t nfl_util; @@ -527,7 +522,7 @@ filelayout_decode_layout(struct pnfs_layout_hdr *flo, if (!scratch) return -ENOMEM; - xdr_init_decode(&stream, &buf, NULL); + xdr_init_decode_pages(&stream, &buf, lgr->layoutp->pages, lgr->layoutp->len); xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); /* 20 = ufl_util (4), first_stripe_index (4), pattern_offset (8), diff --git a/fs/nfs/nfs4filelayoutdev.c b/fs/nfs/nfs4filelayoutdev.c index 5914659c8ec5..3b7bf1377264 100644 --- a/fs/nfs/nfs4filelayoutdev.c +++ b/fs/nfs/nfs4filelayoutdev.c @@ -308,12 +308,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) u8 max_stripe_index; struct nfs4_file_layout_dsaddr *dsaddr = NULL; struct xdr_stream stream; - struct xdr_buf buf = { - .pages = pdev->pages, - .page_len = pdev->pglen, - .buflen = pdev->pglen, - .len = pdev->pglen, - }; + struct xdr_buf buf; struct page *scratch; /* set up xdr stream */ @@ -321,7 +316,7 @@ decode_device(struct inode *ino, struct pnfs_device *pdev, gfp_t gfp_flags) if (!scratch) goto out_err; - xdr_init_decode(&stream, &buf, NULL); + xdr_init_decode_pages(&stream, &buf, pdev->pages, pdev->pglen); xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); /* Get the stripe count (number of stripe index) */ diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h index fc84b7a19ca3..a20970ef9e4e 100644 --- a/include/linux/sunrpc/xdr.h +++ b/include/linux/sunrpc/xdr.h @@ -216,6 +216,8 @@ extern __be32 *xdr_reserve_space(struct xdr_stream *xdr, size_t nbytes); extern void xdr_write_pages(struct xdr_stream *xdr, struct page **pages, unsigned int base, unsigned int len); extern void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p); +extern void xdr_init_decode_pages(struct xdr_stream *xdr, struct xdr_buf *buf, + struct page **pages, unsigned int len); extern void xdr_set_scratch_buffer(struct xdr_stream *xdr, void *buf, size_t buflen); extern __be32 *xdr_inline_decode(struct xdr_stream *xdr, size_t nbytes); extern void xdr_read_pages(struct xdr_stream *xdr, unsigned int len); diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c index 679cd674b81d..f008c14ad34c 100644 --- a/net/sunrpc/xdr.c +++ b/net/sunrpc/xdr.c @@ -638,6 +638,25 @@ void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p) } EXPORT_SYMBOL_GPL(xdr_init_decode); +/** + * xdr_init_decode - Initialize an xdr_stream for decoding data. + * @xdr: pointer to xdr_stream struct + * @buf: pointer to XDR buffer from which to decode data + * @pages: list of pages to decode into + * @len: length in bytes of buffer in pages + */ +void xdr_init_decode_pages(struct xdr_stream *xdr, struct xdr_buf *buf, + struct page **pages, unsigned int len) +{ + memset(buf, 0, sizeof(*buf)); + buf->pages = pages; + buf->page_len = len; + buf->buflen = len; + buf->len = len; + xdr_init_decode(xdr, buf, NULL); +} +EXPORT_SYMBOL_GPL(xdr_init_decode_pages); + static __be32 * __xdr_inline_decode(struct xdr_stream *xdr, size_t nbytes) { __be32 *p = xdr->p; -- cgit v1.2.3