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/nfs4filelayoutdev.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') 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/nfs4filelayoutdev.c | 104 ++++++--------------------------------------- 1 file changed, 13 insertions(+), 91 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') 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 * -- 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/nfs4filelayoutdev.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') 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); } /* -- 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/nfs4filelayoutdev.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) (limited to 'fs/nfs/nfs4filelayoutdev.c') 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) */ -- cgit v1.2.3