Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / nfs / pnfs_dev.c
diff --git a/kernel/fs/nfs/pnfs_dev.c b/kernel/fs/nfs/pnfs_dev.c
new file mode 100644 (file)
index 0000000..2961fcd
--- /dev/null
@@ -0,0 +1,365 @@
+/*
+ *  Device operations for the pnfs client.
+ *
+ *  Copyright (c) 2002
+ *  The Regents of the University of Michigan
+ *  All Rights Reserved
+ *
+ *  Dean Hildebrand <dhildebz@umich.edu>
+ *  Garth Goodson   <Garth.Goodson@netapp.com>
+ *
+ *  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 <linux/export.h>
+#include <linux/nfs_fs.h>
+#include "nfs4session.h"
+#include "internal.h"
+#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)
+
+#define PNFS_DEVICE_RETRY_TIMEOUT (120*HZ)
+
+static struct hlist_head nfs4_deviceid_cache[NFS4_DEVICE_ID_HASH_SIZE];
+static DEFINE_SPINLOCK(nfs4_deviceid_lock);
+
+#ifdef NFS_DEBUG
+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);
+#endif
+
+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;
+}
+
+static struct nfs4_deviceid_node *
+_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;
+
+       hlist_for_each_entry_rcu(d, &nfs4_deviceid_cache[hash], node)
+               if (d->ld == ld && d->nfs_client == clp &&
+                   !memcmp(&d->deviceid, id, sizeof(*id))) {
+                       if (atomic_read(&d->ref))
+                               return d;
+                       else
+                               continue;
+               }
+       return NULL;
+}
+
+static struct nfs4_deviceid_node *
+nfs4_get_device_info(struct nfs_server *server,
+               const struct nfs4_deviceid *dev_id,
+               struct rpc_cred *cred, gfp_t gfp_flags)
+{
+       struct nfs4_deviceid_node *d = NULL;
+       struct pnfs_device *pdev = NULL;
+       struct page **pages = NULL;
+       u32 max_resp_sz;
+       int max_pages;
+       int rc, i;
+
+       /*
+        * Use the session max response size as the basis for setting
+        * GETDEVICEINFO's maxcount
+        */
+       max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
+       if (server->pnfs_curr_ld->max_deviceinfo_size &&
+           server->pnfs_curr_ld->max_deviceinfo_size < max_resp_sz)
+               max_resp_sz = server->pnfs_curr_ld->max_deviceinfo_size;
+       max_pages = nfs_page_array_len(0, max_resp_sz);
+       dprintk("%s: server %p max_resp_sz %u max_pages %d\n",
+               __func__, server, max_resp_sz, max_pages);
+
+       pdev = kzalloc(sizeof(*pdev), gfp_flags);
+       if (!pdev)
+               return NULL;
+
+       pages = kcalloc(max_pages, sizeof(struct page *), gfp_flags);
+       if (!pages)
+               goto out_free_pdev;
+
+       for (i = 0; i < max_pages; i++) {
+               pages[i] = alloc_page(gfp_flags);
+               if (!pages[i])
+                       goto out_free_pages;
+       }
+
+       memcpy(&pdev->dev_id, dev_id, sizeof(*dev_id));
+       pdev->layout_type = server->pnfs_curr_ld->id;
+       pdev->pages = pages;
+       pdev->pgbase = 0;
+       pdev->pglen = max_resp_sz;
+       pdev->mincount = 0;
+       pdev->maxcount = max_resp_sz - nfs41_maxgetdevinfo_overhead;
+
+       rc = nfs4_proc_getdeviceinfo(server, pdev, cred);
+       dprintk("%s getdevice info returns %d\n", __func__, rc);
+       if (rc)
+               goto out_free_pages;
+
+       /*
+        * Found new device, need to decode it and then add it to the
+        * list of known devices for this mountpoint.
+        */
+       d = server->pnfs_curr_ld->alloc_deviceid_node(server, pdev,
+                       gfp_flags);
+       if (d && pdev->nocache)
+               set_bit(NFS_DEVICEID_NOCACHE, &d->flags);
+
+out_free_pages:
+       for (i = 0; i < max_pages; i++)
+               __free_page(pages[i]);
+       kfree(pages);
+out_free_pdev:
+       kfree(pdev);
+       dprintk("<-- %s d %p\n", __func__, d);
+       return d;
+}
+
+/*
+ * 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
+ */
+static struct nfs4_deviceid_node *
+__nfs4_find_get_deviceid(struct nfs_server *server,
+               const struct nfs4_deviceid *id, long hash)
+{
+       struct nfs4_deviceid_node *d;
+
+       rcu_read_lock();
+       d = _lookup_deviceid(server->pnfs_curr_ld, server->nfs_client, id,
+                       hash);
+       if (d != NULL && !atomic_inc_not_zero(&d->ref))
+               d = NULL;
+       rcu_read_unlock();
+       return d;
+}
+
+struct nfs4_deviceid_node *
+nfs4_find_get_deviceid(struct nfs_server *server,
+               const struct nfs4_deviceid *id, struct rpc_cred *cred,
+               gfp_t gfp_mask)
+{
+       long hash = nfs4_deviceid_hash(id);
+       struct nfs4_deviceid_node *d, *new;
+
+       d = __nfs4_find_get_deviceid(server, id, hash);
+       if (d)
+               return d;
+
+       new = nfs4_get_device_info(server, id, cred, gfp_mask);
+       if (!new)
+               return new;
+
+       spin_lock(&nfs4_deviceid_lock);
+       d = __nfs4_find_get_deviceid(server, id, hash);
+       if (d) {
+               spin_unlock(&nfs4_deviceid_lock);
+               server->pnfs_curr_ld->free_deviceid_node(new);
+               return d;
+       }
+       hlist_add_head_rcu(&new->node, &nfs4_deviceid_cache[hash]);
+       atomic_inc(&new->ref);
+       spin_unlock(&nfs4_deviceid_lock);
+
+       return new;
+}
+EXPORT_SYMBOL_GPL(nfs4_find_get_deviceid);
+
+/*
+ * Remove a deviceid from cache
+ *
+ * @clp nfs_client associated with deviceid
+ * @id the deviceid to unhash
+ *
+ * @ret the unhashed node, if found and dereferenced to zero, NULL otherwise.
+ */
+void
+nfs4_delete_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(ld, clp, id, nfs4_deviceid_hash(id));
+       rcu_read_unlock();
+       if (!d) {
+               spin_unlock(&nfs4_deviceid_lock);
+               return;
+       }
+       hlist_del_init_rcu(&d->node);
+       clear_bit(NFS_DEVICEID_NOCACHE, &d->flags);
+       spin_unlock(&nfs4_deviceid_lock);
+
+       /* balance the initial ref set in pnfs_insert_deviceid */
+       nfs4_put_deviceid_node(d);
+}
+EXPORT_SYMBOL_GPL(nfs4_delete_deviceid);
+
+void
+nfs4_init_deviceid_node(struct nfs4_deviceid_node *d, struct nfs_server *server,
+                       const struct nfs4_deviceid *id)
+{
+       INIT_HLIST_NODE(&d->node);
+       INIT_HLIST_NODE(&d->tmpnode);
+       d->ld = server->pnfs_curr_ld;
+       d->nfs_client = server->nfs_client;
+       d->flags = 0;
+       d->deviceid = *id;
+       atomic_set(&d->ref, 1);
+}
+EXPORT_SYMBOL_GPL(nfs4_init_deviceid_node);
+
+/*
+ * Dereference a deviceid node and delete it when its reference count drops
+ * to zero.
+ *
+ * @d deviceid node to put
+ *
+ * return true iff the node was deleted
+ * Note that since the test for d->ref == 0 is sufficient to establish
+ * that the node is no longer hashed in the global device id cache.
+ */
+bool
+nfs4_put_deviceid_node(struct nfs4_deviceid_node *d)
+{
+       if (test_bit(NFS_DEVICEID_NOCACHE, &d->flags)) {
+               if (atomic_add_unless(&d->ref, -1, 2))
+                       return false;
+               nfs4_delete_deviceid(d->ld, d->nfs_client, &d->deviceid);
+       }
+       if (!atomic_dec_and_test(&d->ref))
+               return false;
+       d->ld->free_deviceid_node(d);
+       return true;
+}
+EXPORT_SYMBOL_GPL(nfs4_put_deviceid_node);
+
+void
+nfs4_mark_deviceid_unavailable(struct nfs4_deviceid_node *node)
+{
+       node->timestamp_unavailable = jiffies;
+       set_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags);
+}
+EXPORT_SYMBOL_GPL(nfs4_mark_deviceid_unavailable);
+
+bool
+nfs4_test_deviceid_unavailable(struct nfs4_deviceid_node *node)
+{
+       if (test_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags)) {
+               unsigned long start, end;
+
+               end = jiffies;
+               start = end - PNFS_DEVICE_RETRY_TIMEOUT;
+               if (time_in_range(node->timestamp_unavailable, start, end))
+                       return true;
+               clear_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags);
+       }
+       return false;
+}
+EXPORT_SYMBOL_GPL(nfs4_test_deviceid_unavailable);
+
+static void
+_deviceid_purge_client(const struct nfs_client *clp, long hash)
+{
+       struct nfs4_deviceid_node *d;
+       HLIST_HEAD(tmp);
+
+       spin_lock(&nfs4_deviceid_lock);
+       rcu_read_lock();
+       hlist_for_each_entry_rcu(d, &nfs4_deviceid_cache[hash], node)
+               if (d->nfs_client == clp && atomic_read(&d->ref)) {
+                       hlist_del_init_rcu(&d->node);
+                       hlist_add_head(&d->tmpnode, &tmp);
+                       clear_bit(NFS_DEVICEID_NOCACHE, &d->flags);
+               }
+       rcu_read_unlock();
+       spin_unlock(&nfs4_deviceid_lock);
+
+       if (hlist_empty(&tmp))
+               return;
+
+       while (!hlist_empty(&tmp)) {
+               d = hlist_entry(tmp.first, struct nfs4_deviceid_node, tmpnode);
+               hlist_del(&d->tmpnode);
+               nfs4_put_deviceid_node(d);
+       }
+}
+
+void
+nfs4_deviceid_purge_client(const struct nfs_client *clp)
+{
+       long h;
+
+       if (!(clp->cl_exchange_flags & EXCHGID4_FLAG_USE_PNFS_MDS))
+               return;
+       for (h = 0; h < NFS4_DEVICE_ID_HASH_SIZE; h++)
+               _deviceid_purge_client(clp, h);
+}
+
+/*
+ * Stop use of all deviceids associated with an nfs_client
+ */
+void
+nfs4_deviceid_mark_client_invalid(struct nfs_client *clp)
+{
+       struct nfs4_deviceid_node *d;
+       int i;
+
+       rcu_read_lock();
+       for (i = 0; i < NFS4_DEVICE_ID_HASH_SIZE; i ++){
+               hlist_for_each_entry_rcu(d, &nfs4_deviceid_cache[i], node)
+                       if (d->nfs_client == clp)
+                               set_bit(NFS_DEVICEID_INVALID, &d->flags);
+       }
+       rcu_read_unlock();
+}