summaryrefslogtreecommitdiff
path: root/fs/afs/server.c
diff options
context:
space:
mode:
authorDavid Howells <dhowells@redhat.com>2017-11-02 18:27:50 +0300
committerDavid Howells <dhowells@redhat.com>2017-11-13 18:38:19 +0300
commitd2ddc776a4581d900fc3bdc7803b403daae64d88 (patch)
tree6c5f6bac9e9e9a326ed0a9248914b53c421558e5 /fs/afs/server.c
parent9cc6fc50f7bc69ac28bee45eed13cbc65a86210f (diff)
downloadlinux-d2ddc776a4581d900fc3bdc7803b403daae64d88.tar.xz
afs: Overhaul volume and server record caching and fileserver rotation
The current code assumes that volumes and servers are per-cell and are never shared, but this is not enforced, and, indeed, public cells do exist that are aliases of each other. Further, an organisation can, say, set up a public cell and a private cell with overlapping, but not identical, sets of servers. The difference is purely in the database attached to the VL servers. The current code will malfunction if it sees a server in two cells as it assumes global address -> server record mappings and that each server is in just one cell. Further, each server may have multiple addresses - and may have addresses of different families (IPv4 and IPv6, say). To this end, the following structural changes are made: (1) Server record management is overhauled: (a) Server records are made independent of cell. The namespace keeps track of them, volume records have lists of them and each vnode has a server on which its callback interest currently resides. (b) The cell record no longer keeps a list of servers known to be in that cell. (c) The server records are now kept in a flat list because there's no single address to sort on. (d) Server records are now keyed by their UUID within the namespace. (e) The addresses for a server are obtained with the VL.GetAddrsU rather than with VL.GetEntryByName, using the server's UUID as a parameter. (f) Cached server records are garbage collected after a period of non-use and are counted out of existence before purging is allowed to complete. This protects the work functions against rmmod. (g) The servers list is now in /proc/fs/afs/servers. (2) Volume record management is overhauled: (a) An RCU-replaceable server list is introduced. This tracks both servers and their coresponding callback interests. (b) The superblock is now keyed on cell record and numeric volume ID. (c) The volume record is now tied to the superblock which mounts it, and is activated when mounted and deactivated when unmounted. This makes it easier to handle the cache cookie without causing a double-use in fscache. (d) The volume record is loaded from the VLDB using VL.GetEntryByNameU to get the server UUID list. (e) The volume name is updated if it is seen to have changed when the volume is updated (the update is keyed on the volume ID). (3) The vlocation record is got rid of and VLDB records are no longer cached. Sufficient information is stored in the volume record, though an update to a volume record is now no longer shared between related volumes (volumes come in bundles of three: R/W, R/O and backup). and the following procedural changes are made: (1) The fileserver cursor introduced previously is now fleshed out and used to iterate over fileservers and their addresses. (2) Volume status is checked during iteration, and the server list is replaced if a change is detected. (3) Server status is checked during iteration, and the address list is replaced if a change is detected. (4) The abort code is saved into the address list cursor and -ECONNABORTED returned in afs_make_call() if a remote abort happened rather than translating the abort into an error message. This allows actions to be taken depending on the abort code more easily. (a) If a VMOVED abort is seen then this is handled by rechecking the volume and restarting the iteration. (b) If a VBUSY, VRESTARTING or VSALVAGING abort is seen then this is handled by sleeping for a short period and retrying and/or trying other servers that might serve that volume. A message is also displayed once until the condition has cleared. (c) If a VOFFLINE abort is seen, then this is handled as VBUSY for the moment. (d) If a VNOVOL abort is seen, the volume is rechecked in the VLDB to see if it has been deleted; if not, the fileserver is probably indicating that the volume couldn't be attached and needs salvaging. (e) If statfs() sees one of these aborts, it does not sleep, but rather returns an error, so as not to block the umount program. (5) The fileserver iteration functions in vnode.c are now merged into their callers and more heavily macroised around the cursor. vnode.c is removed. (6) Operations on a particular vnode are serialised on that vnode because the server will lock that vnode whilst it operates on it, so a second op sent will just have to wait. (7) Fileservers are probed with FS.GetCapabilities before being used. This is where service upgrade will be done. (8) A callback interest on a fileserver is set up before an FS operation is performed and passed through to afs_make_call() so that it can be set on the vnode if the operation returns a callback. The callback interest is passed through to afs_iget() also so that it can be set there too. In general, record updating is done on an as-needed basis when we try to access servers, volumes or vnodes rather than offloading it to work items and special threads. Notes: (1) Pre AFS-3.4 servers are no longer supported, though this can be added back if necessary (AFS-3.4 was released in 1998). (2) VBUSY is retried forever for the moment at intervals of 1s. (3) /proc/fs/afs/<cell>/servers no longer exists. Signed-off-by: David Howells <dhowells@redhat.com>
Diffstat (limited to 'fs/afs/server.c')
-rw-r--r--fs/afs/server.c750
1 files changed, 537 insertions, 213 deletions
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 9ca174b24f5b..a6c860bcf391 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -14,7 +14,8 @@
#include "afs_fs.h"
#include "internal.h"
-static unsigned afs_server_timeout = 10; /* server timeout in seconds */
+static unsigned afs_server_gc_delay = 10; /* Server record timeout in seconds */
+static unsigned afs_server_update_delay = 30; /* Time till VLDB recheck in secs */
static void afs_inc_servers_outstanding(struct afs_net *net)
{
@@ -27,60 +28,201 @@ static void afs_dec_servers_outstanding(struct afs_net *net)
wake_up_atomic_t(&net->servers_outstanding);
}
-void afs_server_timer(struct timer_list *timer)
+/*
+ * Find a server by one of its addresses.
+ */
+struct afs_server *afs_find_server(struct afs_net *net,
+ const struct sockaddr_rxrpc *srx)
{
- struct afs_net *net = container_of(timer, struct afs_net, server_timer);
+ const struct sockaddr_in6 *a = &srx->transport.sin6, *b;
+ const struct afs_addr_list *alist;
+ struct afs_server *server = NULL;
+ unsigned int i;
+ bool ipv6 = true;
+ int seq = 0, diff;
+
+ if (srx->transport.sin6.sin6_addr.s6_addr32[0] == 0 ||
+ srx->transport.sin6.sin6_addr.s6_addr32[1] == 0 ||
+ srx->transport.sin6.sin6_addr.s6_addr32[2] == htonl(0xffff))
+ ipv6 = false;
+
+ rcu_read_lock();
+
+ do {
+ if (server)
+ afs_put_server(net, server);
+ server = NULL;
+ read_seqbegin_or_lock(&net->fs_addr_lock, &seq);
+
+ if (ipv6) {
+ hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
+ alist = rcu_dereference(server->addresses);
+ for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
+ b = &alist->addrs[i].transport.sin6;
+ diff = (u16)a->sin6_port - (u16)b->sin6_port;
+ if (diff == 0)
+ diff = memcmp(&a->sin6_addr,
+ &b->sin6_addr,
+ sizeof(struct in6_addr));
+ if (diff == 0)
+ goto found;
+ if (diff < 0) {
+ // TODO: Sort the list
+ //if (i == alist->nr_ipv4)
+ // goto not_found;
+ break;
+ }
+ }
+ }
+ } else {
+ hlist_for_each_entry_rcu(server, &net->fs_addresses4, addr4_link) {
+ alist = rcu_dereference(server->addresses);
+ for (i = 0; i < alist->nr_ipv4; i++) {
+ b = &alist->addrs[i].transport.sin6;
+ diff = (u16)a->sin6_port - (u16)b->sin6_port;
+ if (diff == 0)
+ diff = ((u32)a->sin6_addr.s6_addr32[3] -
+ (u32)b->sin6_addr.s6_addr32[3]);
+ if (diff == 0)
+ goto found;
+ if (diff < 0) {
+ // TODO: Sort the list
+ //if (i == 0)
+ // goto not_found;
+ break;
+ }
+ }
+ }
+ }
- if (!queue_work(afs_wq, &net->server_reaper))
- afs_dec_servers_outstanding(net);
+ //not_found:
+ server = NULL;
+ found:
+ if (server && !atomic_inc_not_zero(&server->usage))
+ server = NULL;
+
+ } while (need_seqretry(&net->fs_addr_lock, seq));
+
+ done_seqretry(&net->fs_addr_lock, seq);
+
+ rcu_read_unlock();
+ return server;
}
/*
- * install a server record in the master tree
+ * Look up a server by its UUID
*/
-static int afs_install_server(struct afs_server *server)
+struct afs_server *afs_find_server_by_uuid(struct afs_net *net, const uuid_t *uuid)
{
- struct afs_server *xserver;
- struct afs_net *net = server->cell->net;
+ struct afs_server *server = NULL;
+ struct rb_node *p;
+ int diff, seq = 0;
+
+ _enter("%pU", uuid);
+
+ do {
+ /* Unfortunately, rbtree walking doesn't give reliable results
+ * under just the RCU read lock, so we have to check for
+ * changes.
+ */
+ if (server)
+ afs_put_server(net, server);
+ server = NULL;
+
+ read_seqbegin_or_lock(&net->fs_lock, &seq);
+
+ p = net->fs_servers.rb_node;
+ while (p) {
+ server = rb_entry(p, struct afs_server, uuid_rb);
+
+ diff = memcmp(uuid, &server->uuid, sizeof(*uuid));
+ if (diff < 0) {
+ p = p->rb_left;
+ } else if (diff > 0) {
+ p = p->rb_right;
+ } else {
+ afs_get_server(server);
+ break;
+ }
+
+ server = NULL;
+ }
+ } while (need_seqretry(&net->fs_lock, seq));
+
+ done_seqretry(&net->fs_lock, seq);
+
+ _leave(" = %p", server);
+ return server;
+}
+
+/*
+ * Install a server record in the namespace tree
+ */
+static struct afs_server *afs_install_server(struct afs_net *net,
+ struct afs_server *candidate)
+{
+ const struct afs_addr_list *alist;
+ struct afs_server *server;
struct rb_node **pp, *p;
- int ret, diff;
+ int ret = -EEXIST, diff;
- _enter("%p", server);
+ _enter("%p", candidate);
- write_lock(&net->servers_lock);
+ write_seqlock(&net->fs_lock);
- ret = -EEXIST;
- pp = &net->servers.rb_node;
+ /* Firstly install the server in the UUID lookup tree */
+ pp = &net->fs_servers.rb_node;
p = NULL;
while (*pp) {
p = *pp;
_debug("- consider %p", p);
- xserver = rb_entry(p, struct afs_server, master_rb);
- diff = memcmp(&server->addrs->addrs[0],
- &xserver->addrs->addrs[0],
- sizeof(sizeof(server->addrs->addrs[0])));
+ server = rb_entry(p, struct afs_server, uuid_rb);
+ diff = memcmp(&candidate->uuid, &server->uuid, sizeof(uuid_t));
if (diff < 0)
pp = &(*pp)->rb_left;
else if (diff > 0)
pp = &(*pp)->rb_right;
else
- goto error;
+ goto exists;
}
- rb_link_node(&server->master_rb, p, pp);
- rb_insert_color(&server->master_rb, &net->servers);
+ server = candidate;
+ rb_link_node(&server->uuid_rb, p, pp);
+ rb_insert_color(&server->uuid_rb, &net->fs_servers);
+ hlist_add_head_rcu(&server->proc_link, &net->fs_proc);
+
+ write_seqlock(&net->fs_addr_lock);
+ alist = rcu_dereference_protected(server->addresses,
+ lockdep_is_held(&net->fs_addr_lock.lock));
+
+ /* Secondly, if the server has any IPv4 and/or IPv6 addresses, install
+ * it in the IPv4 and/or IPv6 reverse-map lists.
+ *
+ * TODO: For speed we want to use something other than a flat list
+ * here; even sorting the list in terms of lowest address would help a
+ * bit, but anything we might want to do gets messy and memory
+ * intensive.
+ */
+ if (alist->nr_ipv4 > 0)
+ hlist_add_head_rcu(&server->addr4_link, &net->fs_addresses4);
+ if (alist->nr_addrs > alist->nr_ipv4)
+ hlist_add_head_rcu(&server->addr6_link, &net->fs_addresses6);
+
+ write_sequnlock(&net->fs_addr_lock);
ret = 0;
-error:
- write_unlock(&net->servers_lock);
- return ret;
+exists:
+ afs_get_server(server);
+ write_sequnlock(&net->fs_lock);
+ return server;
}
/*
* allocate a new server record
*/
-static struct afs_server *afs_alloc_server(struct afs_cell *cell,
- const struct sockaddr_rxrpc *addr)
+static struct afs_server *afs_alloc_server(struct afs_net *net,
+ const uuid_t *uuid,
+ struct afs_addr_list *alist)
{
struct afs_server *server;
@@ -89,194 +231,155 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
server = kzalloc(sizeof(struct afs_server), GFP_KERNEL);
if (!server)
goto enomem;
- server->addrs = kzalloc(sizeof(struct afs_addr_list) +
- sizeof(struct sockaddr_rxrpc),
- GFP_KERNEL);
- if (!server->addrs)
- goto enomem_server;
atomic_set(&server->usage, 1);
- server->net = cell->net;
- server->cell = cell;
-
- INIT_LIST_HEAD(&server->link);
- INIT_LIST_HEAD(&server->grave);
- init_rwsem(&server->sem);
- spin_lock_init(&server->fs_lock);
+ RCU_INIT_POINTER(server->addresses, alist);
+ server->addr_version = alist->version;
+ server->uuid = *uuid;
+ server->flags = (1UL << AFS_SERVER_FL_NEW);
+ server->update_at = ktime_get_real_seconds() + afs_server_update_delay;
+ rwlock_init(&server->fs_lock);
INIT_LIST_HEAD(&server->cb_interests);
rwlock_init(&server->cb_break_lock);
- refcount_set(&server->addrs->usage, 1);
- server->addrs->nr_addrs = 1;
- server->addrs->addrs[0] = *addr;
- afs_inc_servers_outstanding(cell->net);
-
- _leave(" = %p{%d}", server, atomic_read(&server->usage));
+ afs_inc_servers_outstanding(net);
+ _leave(" = %p", server);
return server;
-enomem_server:
- kfree(server);
enomem:
_leave(" = NULL [nomem]");
return NULL;
}
/*
- * get an FS-server record for a cell
+ * Look up an address record for a server
*/
-struct afs_server *afs_lookup_server(struct afs_cell *cell,
- struct sockaddr_rxrpc *addr)
+static struct afs_addr_list *afs_vl_lookup_addrs(struct afs_cell *cell,
+ struct key *key, const uuid_t *uuid)
{
- struct afs_server *server, *candidate;
-
- _enter("%p,%pIS", cell, &addr->transport);
-
- /* quick scan of the list to see if we already have the server */
- read_lock(&cell->servers_lock);
-
- list_for_each_entry(server, &cell->servers, link) {
- if (memcmp(&server->addrs->addrs[0], addr, sizeof(*addr)) == 0)
- goto found_server_quickly;
- }
- read_unlock(&cell->servers_lock);
-
- candidate = afs_alloc_server(cell, addr);
- if (!candidate) {
- _leave(" = -ENOMEM");
- return ERR_PTR(-ENOMEM);
- }
-
- write_lock(&cell->servers_lock);
-
- /* check the cell's server list again */
- list_for_each_entry(server, &cell->servers, link) {
- if (memcmp(&server->addrs->addrs[0], addr, sizeof(*addr)) == 0)
- goto found_server;
- }
-
- _debug("new");
- server = candidate;
- if (afs_install_server(server) < 0)
- goto server_in_two_cells;
-
- afs_get_cell(cell);
- list_add_tail(&server->link, &cell->servers);
-
- write_unlock(&cell->servers_lock);
- _leave(" = %p{%d}", server, atomic_read(&server->usage));
- return server;
-
- /* found a matching server quickly */
-found_server_quickly:
- _debug("found quickly");
- afs_get_server(server);
- read_unlock(&cell->servers_lock);
-no_longer_unused:
- if (!list_empty(&server->grave)) {
- spin_lock(&cell->net->server_graveyard_lock);
- list_del_init(&server->grave);
- spin_unlock(&cell->net->server_graveyard_lock);
+ struct afs_addr_cursor ac;
+ struct afs_addr_list *alist;
+ int ret;
+
+ ret = afs_set_vl_cursor(&ac, cell);
+ if (ret < 0)
+ return ERR_PTR(ret);
+
+ while (afs_iterate_addresses(&ac)) {
+ alist = afs_vl_get_addrs_u(cell->net, &ac, key, uuid);
+ switch (ac.error) {
+ case 0:
+ afs_end_cursor(&ac);
+ return alist;
+ case -ECONNABORTED:
+ ac.error = afs_abort_to_error(ac.abort_code);
+ goto error;
+ case -ENOMEM:
+ case -ENONET:
+ goto error;
+ case -ENETUNREACH:
+ case -EHOSTUNREACH:
+ case -ECONNREFUSED:
+ break;
+ default:
+ ac.error = -EIO;
+ goto error;
+ }
}
- _leave(" = %p{%d}", server, atomic_read(&server->usage));
- return server;
- /* found a matching server on the second pass */
-found_server:
- _debug("found");
- afs_get_server(server);
- write_unlock(&cell->servers_lock);
- kfree(candidate);
- goto no_longer_unused;
-
- /* found a server that seems to be in two cells */
-server_in_two_cells:
- write_unlock(&cell->servers_lock);
- kfree(candidate);
- afs_dec_servers_outstanding(cell->net);
- printk(KERN_NOTICE "kAFS: Server %pI4 appears to be in two cells\n",
- addr);
- _leave(" = -EEXIST");
- return ERR_PTR(-EEXIST);
+error:
+ return ERR_PTR(afs_end_cursor(&ac));
}
/*
- * look up a server by its IP address
+ * Get or create a fileserver record.
*/
-struct afs_server *afs_find_server(struct afs_net *net,
- const struct sockaddr_rxrpc *srx)
+struct afs_server *afs_lookup_server(struct afs_cell *cell, struct key *key,
+ const uuid_t *uuid)
{
- struct afs_server *server = NULL;
- struct rb_node *p;
- int diff;
+ struct afs_addr_list *alist;
+ struct afs_server *server, *candidate;
- _enter("{%d,%pIS}", srx->transport.family, &srx->transport);
+ _enter("%p,%pU", cell->net, uuid);
- read_lock(&net->servers_lock);
+ server = afs_find_server_by_uuid(cell->net, uuid);
+ if (server)
+ return server;
- p = net->servers.rb_node;
- while (p) {
- server = rb_entry(p, struct afs_server, master_rb);
+ alist = afs_vl_lookup_addrs(cell, key, uuid);
+ if (IS_ERR(alist))
+ return ERR_CAST(alist);
- _debug("- consider %p", p);
+ candidate = afs_alloc_server(cell->net, uuid, alist);
+ if (!candidate) {
+ afs_put_addrlist(alist);
+ return ERR_PTR(-ENOMEM);
+ }
- diff = memcmp(srx, &server->addrs->addrs[0], sizeof(*srx));
- if (diff < 0) {
- p = p->rb_left;
- } else if (diff > 0) {
- p = p->rb_right;
- } else {
- afs_get_server(server);
- goto found;
- }
+ server = afs_install_server(cell->net, candidate);
+ if (server != candidate) {
+ afs_put_addrlist(alist);
+ kfree(candidate);
}
- server = NULL;
-found:
- read_unlock(&net->servers_lock);
- _leave(" = %p", server);
+ _leave(" = %p{%d}", server, atomic_read(&server->usage));
return server;
}
+/*
+ * Set the server timer to fire after a given delay, assuming it's not already
+ * set for an earlier time.
+ */
static void afs_set_server_timer(struct afs_net *net, time64_t delay)
{
- afs_inc_servers_outstanding(net);
if (net->live) {
- if (timer_reduce(&net->server_timer, jiffies + delay * HZ))
- afs_dec_servers_outstanding(net);
- } else {
- if (!queue_work(afs_wq, &net->server_reaper))
+ afs_inc_servers_outstanding(net);
+ if (timer_reduce(&net->fs_timer, jiffies + delay * HZ))
afs_dec_servers_outstanding(net);
}
}
/*
- * destroy a server record
- * - removes from the cell list
+ * Server management timer. We have an increment on fs_outstanding that we
+ * need to pass along to the work item.
+ */
+void afs_servers_timer(struct timer_list *timer)
+{
+ struct afs_net *net = container_of(timer, struct afs_net, fs_timer);
+
+ _enter("");
+ if (!queue_work(afs_wq, &net->fs_manager))
+ afs_dec_servers_outstanding(net);
+}
+
+/*
+ * Release a reference on a server record.
*/
void afs_put_server(struct afs_net *net, struct afs_server *server)
{
+ unsigned int usage;
+
if (!server)
return;
- _enter("%p{%d}", server, atomic_read(&server->usage));
+ server->put_time = ktime_get_real_seconds();
- _debug("PUT SERVER %d", atomic_read(&server->usage));
+ usage = atomic_dec_return(&server->usage);
- ASSERTCMP(atomic_read(&server->usage), >, 0);
+ _enter("{%u}", usage);
- if (likely(!atomic_dec_and_test(&server->usage))) {
- _leave("");
+ if (likely(usage > 0))
return;
- }
- spin_lock(&net->server_graveyard_lock);
- if (atomic_read(&server->usage) == 0) {
- list_move_tail(&server->grave, &net->server_graveyard);
- server->time_of_death = ktime_get_real_seconds();
- afs_set_server_timer(net, afs_server_timeout);
- }
- spin_unlock(&net->server_graveyard_lock);
- _leave(" [dead]");
+ afs_set_server_timer(net, afs_server_gc_delay);
+}
+
+static void afs_server_rcu(struct rcu_head *rcu)
+{
+ struct afs_server *server = container_of(rcu, struct afs_server, rcu);
+
+ afs_put_addrlist(server->addresses);
+ kfree(server);
}
/*
@@ -284,7 +387,7 @@ void afs_put_server(struct afs_net *net, struct afs_server *server)
*/
static void afs_destroy_server(struct afs_net *net, struct afs_server *server)
{
- struct afs_addr_list *alist = server->addrs;
+ struct afs_addr_list *alist = server->addresses;
struct afs_addr_cursor ac = {
.alist = alist,
.addr = &alist->addrs[0],
@@ -294,79 +397,300 @@ static void afs_destroy_server(struct afs_net *net, struct afs_server *server)
};
_enter("%p", server);
- afs_fs_give_up_all_callbacks(server, &ac, NULL, false);
- afs_put_cell(net, server->cell);
- afs_put_addrlist(server->addrs);
- kfree(server);
+ afs_fs_give_up_all_callbacks(net, server, &ac, NULL);
+ call_rcu(&server->rcu, afs_server_rcu);
afs_dec_servers_outstanding(net);
}
/*
- * reap dead server records
+ * Garbage collect any expired servers.
*/
-void afs_reap_server(struct work_struct *work)
+static void afs_gc_servers(struct afs_net *net, struct afs_server *gc_list)
{
- LIST_HEAD(corpses);
struct afs_server *server;
- struct afs_net *net = container_of(work, struct afs_net, server_reaper);
- unsigned long delay, expiry;
- time64_t now;
-
- now = ktime_get_real_seconds();
- spin_lock(&net->server_graveyard_lock);
-
- while (!list_empty(&net->server_graveyard)) {
- server = list_entry(net->server_graveyard.next,
- struct afs_server, grave);
-
- /* the queue is ordered most dead first */
- if (net->live) {
- expiry = server->time_of_death + afs_server_timeout;
- if (expiry > now) {
- delay = (expiry - now);
- afs_set_server_timer(net, delay);
- break;
- }
+ bool deleted;
+ int usage;
+
+ while ((server = gc_list)) {
+ gc_list = server->gc_next;
+
+ write_seqlock(&net->fs_lock);
+ usage = 1;
+ deleted = atomic_try_cmpxchg(&server->usage, &usage, 0);
+ if (deleted) {
+ rb_erase(&server->uuid_rb, &net->fs_servers);
+ hlist_del_rcu(&server->proc_link);
}
+ write_sequnlock(&net->fs_lock);
- write_lock(&server->cell->servers_lock);
- write_lock(&net->servers_lock);
- if (atomic_read(&server->usage) > 0) {
- list_del_init(&server->grave);
- } else {
- list_move_tail(&server->grave, &corpses);
- list_del_init(&server->link);
- rb_erase(&server->master_rb, &net->servers);
+ if (deleted)
+ afs_destroy_server(net, server);
+ }
+}
+
+/*
+ * Manage the records of servers known to be within a network namespace. This
+ * includes garbage collecting unused servers.
+ *
+ * Note also that we were given an increment on net->servers_outstanding by
+ * whoever queued us that we need to deal with before returning.
+ */
+void afs_manage_servers(struct work_struct *work)
+{
+ struct afs_net *net = container_of(work, struct afs_net, fs_manager);
+ struct afs_server *gc_list = NULL;
+ struct rb_node *cursor;
+ time64_t now = ktime_get_real_seconds(), next_manage = TIME64_MAX;
+ bool purging = !net->live;
+
+ _enter("");
+
+ /* Trawl the server list looking for servers that have expired from
+ * lack of use.
+ */
+ read_seqlock_excl(&net->fs_lock);
+
+ for (cursor = rb_first(&net->fs_servers); cursor; cursor = rb_next(cursor)) {
+ struct afs_server *server =
+ rb_entry(cursor, struct afs_server, uuid_rb);
+ int usage = atomic_read(&server->usage);
+
+ _debug("manage %pU %u", &server->uuid, usage);
+
+ ASSERTCMP(usage, >=, 1);
+ ASSERTIFCMP(purging, usage, ==, 1);
+
+ if (usage == 1) {
+ time64_t expire_at = server->put_time;
+
+ if (!test_bit(AFS_SERVER_FL_VL_FAIL, &server->flags) &&
+ !test_bit(AFS_SERVER_FL_NOT_FOUND, &server->flags))
+ expire_at += afs_server_gc_delay;
+ if (purging || expire_at <= now) {
+ server->gc_next = gc_list;
+ gc_list = server;
+ } else if (expire_at < next_manage) {
+ next_manage = expire_at;
+ }
}
- write_unlock(&net->servers_lock);
- write_unlock(&server->cell->servers_lock);
}
- spin_unlock(&net->server_graveyard_lock);
+ read_sequnlock_excl(&net->fs_lock);
- /* now reap the corpses we've extracted */
- while (!list_empty(&corpses)) {
- server = list_entry(corpses.next, struct afs_server, grave);
- list_del(&server->grave);
- afs_destroy_server(net, server);
+ /* Update the timer on the way out. We have to pass an increment on
+ * servers_outstanding in the namespace that we are in to the timer or
+ * the work scheduler.
+ */
+ if (!purging && next_manage < TIME64_MAX) {
+ now = ktime_get_real_seconds();
+
+ if (next_manage - now <= 0) {
+ if (queue_work(afs_wq, &net->fs_manager))
+ afs_inc_servers_outstanding(net);
+ } else {
+ afs_set_server_timer(net, next_manage - now);
+ }
}
+ afs_gc_servers(net, gc_list);
+
afs_dec_servers_outstanding(net);
+ _leave(" [%d]", atomic_read(&net->servers_outstanding));
+}
+
+static void afs_queue_server_manager(struct afs_net *net)
+{
+ afs_inc_servers_outstanding(net);
+ if (!queue_work(afs_wq, &net->fs_manager))
+ afs_dec_servers_outstanding(net);
}
/*
- * Discard all the server records from a net namespace when it is destroyed or
- * the afs module is removed.
+ * Purge list of servers.
*/
-void __net_exit afs_purge_servers(struct afs_net *net)
+void afs_purge_servers(struct afs_net *net)
{
- if (del_timer_sync(&net->server_timer))
+ _enter("");
+
+ if (del_timer_sync(&net->fs_timer))
atomic_dec(&net->servers_outstanding);
- afs_inc_servers_outstanding(net);
- if (!queue_work(afs_wq, &net->server_reaper))
- afs_dec_servers_outstanding(net);
+ afs_queue_server_manager(net);
+ _debug("wait");
wait_on_atomic_t(&net->servers_outstanding, atomic_t_wait,
TASK_UNINTERRUPTIBLE);
+ _leave("");
+}
+
+/*
+ * Probe a fileserver to find its capabilities.
+ *
+ * TODO: Try service upgrade.
+ */
+static bool afs_do_probe_fileserver(struct afs_fs_cursor *fc)
+{
+ _enter("");
+
+ fc->ac.addr = NULL;
+ fc->ac.start = READ_ONCE(fc->ac.alist->index);
+ fc->ac.index = fc->ac.start;
+ fc->ac.error = 0;
+ fc->ac.begun = false;
+
+ while (afs_iterate_addresses(&fc->ac)) {
+ afs_fs_get_capabilities(afs_v2net(fc->vnode), fc->cbi->server,
+ &fc->ac, fc->key);
+ switch (fc->ac.error) {
+ case 0:
+ afs_end_cursor(&fc->ac);
+ set_bit(AFS_SERVER_FL_PROBED, &fc->cbi->server->flags);
+ return true;
+ case -ECONNABORTED:
+ fc->ac.error = afs_abort_to_error(fc->ac.abort_code);
+ goto error;
+ case -ENOMEM:
+ case -ENONET:
+ goto error;
+ case -ENETUNREACH:
+ case -EHOSTUNREACH:
+ case -ECONNREFUSED:
+ case -ETIMEDOUT:
+ case -ETIME:
+ break;
+ default:
+ fc->ac.error = -EIO;
+ goto error;
+ }
+ }
+
+error:
+ afs_end_cursor(&fc->ac);
+ return false;
+}
+
+/*
+ * If we haven't already, try probing the fileserver to get its capabilities.
+ * We try not to instigate parallel probes, but it's possible that the parallel
+ * probes will fail due to authentication failure when ours would succeed.
+ *
+ * TODO: Try sending an anonymous probe if an authenticated probe fails.
+ */
+bool afs_probe_fileserver(struct afs_fs_cursor *fc)
+{
+ bool success;
+ int ret, retries = 0;
+
+ _enter("");
+
+retry:
+ if (test_bit(AFS_SERVER_FL_PROBED, &fc->cbi->server->flags)) {
+ _leave(" = t");
+ return true;
+ }
+
+ if (!test_and_set_bit_lock(AFS_SERVER_FL_PROBING, &fc->cbi->server->flags)) {
+ success = afs_do_probe_fileserver(fc);
+ clear_bit_unlock(AFS_SERVER_FL_PROBING, &fc->cbi->server->flags);
+ wake_up_bit(&fc->cbi->server->flags, AFS_SERVER_FL_PROBING);
+ _leave(" = t");
+ return success;
+ }
+
+ _debug("wait");
+ ret = wait_on_bit(&fc->cbi->server->flags, AFS_SERVER_FL_PROBING,
+ TASK_INTERRUPTIBLE);
+ if (ret == -ERESTARTSYS) {
+ fc->ac.error = ret;
+ _leave(" = f [%d]", ret);
+ return false;
+ }
+
+ retries++;
+ if (retries == 4) {
+ fc->ac.error = -ESTALE;
+ _leave(" = f [stale]");
+ return false;
+ }
+ _debug("retry");
+ goto retry;
+}
+
+/*
+ * Get an update for a server's address list.
+ */
+static noinline bool afs_update_server_record(struct afs_fs_cursor *fc, struct afs_server *server)
+{
+ struct afs_addr_list *alist, *discard;
+
+ _enter("");
+
+ alist = afs_vl_lookup_addrs(fc->vnode->volume->cell, fc->key,
+ &server->uuid);
+ if (IS_ERR(alist)) {
+ fc->ac.error = PTR_ERR(alist);
+ _leave(" = f [%d]", fc->ac.error);
+ return false;
+ }
+
+ discard = alist;
+ if (server->addr_version != alist->version) {
+ write_lock(&server->fs_lock);
+ discard = rcu_dereference_protected(server->addresses,
+ lockdep_is_held(&server->fs_lock));
+ rcu_assign_pointer(server->addresses, alist);
+ server->addr_version = alist->version;
+ write_unlock(&server->fs_lock);
+ }
+
+ server->update_at = ktime_get_real_seconds() + afs_server_update_delay;
+ afs_put_addrlist(discard);
+ _leave(" = t");
+ return true;
+}
+
+/*
+ * See if a server's address list needs updating.
+ */
+bool afs_check_server_record(struct afs_fs_cursor *fc, struct afs_server *server)
+{
+ time64_t now = ktime_get_real_seconds();
+ long diff;
+ bool success;
+ int ret, retries = 0;
+
+ _enter("");
+
+ ASSERT(server);
+
+retry:
+ diff = READ_ONCE(server->update_at) - now;
+ if (diff > 0) {
+ _leave(" = t [not now %ld]", diff);
+ return true;
+ }
+
+ if (!test_and_set_bit_lock(AFS_SERVER_FL_UPDATING, &server->flags)) {
+ success = afs_update_server_record(fc, server);
+ clear_bit_unlock(AFS_SERVER_FL_UPDATING, &server->flags);
+ wake_up_bit(&server->flags, AFS_SERVER_FL_UPDATING);
+ _leave(" = %d", success);
+ return success;
+ }
+
+ ret = wait_on_bit(&server->flags, AFS_SERVER_FL_UPDATING,
+ TASK_INTERRUPTIBLE);
+ if (ret == -ERESTARTSYS) {
+ fc->ac.error = ret;
+ _leave(" = f [intr]");
+ return false;
+ }
+
+ retries++;
+ if (retries == 4) {
+ _leave(" = f [stale]");
+ ret = -ESTALE;
+ return false;
+ }
+ goto retry;
}