[PATCH 6/6] afs: Try to avoid taking RCU read lock when checking vnode validity

From: David Howells
Date: Wed Sep 08 2021 - 11:58:32 EST

Next message: David Hildenbrand: "Re: [PATCH v1] hugetlbfs: s390 is always 64bit"
Previous message: David Howells: "[PATCH 5/6] afs: Fix mmap coherency vs 3rd-party changes"
In reply to: David Howells: "[PATCH 5/6] afs: Fix mmap coherency vs 3rd-party changes"
Next in thread: markus . suvanto: "Re: [PATCH 0/6] afs: Fixes for 3rd party-induced data corruption"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Try to avoid taking the RCU read lock when checking the validity of a
vnode's callback state. The only thing it's needed for is to pin the
parent volume's server list whilst we search it to find the record of the
server we're currently using to see if it has been reinitialised (ie. it
sent us a CB.InitCallBackState* RPC).

Do this by the following means:

(1) Keep an additional per-cell counter (fs_s_break) that's incremented
each time any of the fileservers in the cell reinitialises.

Since the new counter can be accessed without RCU from the vnode, we
can check that first - and only if it differs, get the RCU read lock
and check the volume's server list.

(2) Replace afs_get_s_break_rcu() with afs_check_server_good() which now
indicates whether the callback promise is still expected to be present
on the server. This does the checks as described in (1).

(3) Restructure afs_check_validity() to take account of the change in (2).

We can also get rid of the valid variable and just use the need_clear
variable with the addition of the afs_cb_break_no_promise reason.

(4) afs_check_validity() probably shouldn't be altering vnode->cb_v_break
and vnode->cb_s_break when it doesn't have cb_lock exclusively locked.

Move the change to vnode->cb_v_break to __afs_break_callback().

Delegate the change to vnode->cb_s_break to afs_select_fileserver()
and set vnode->cb_fs_s_break there also.

(5) afs_validate() no longer needs to get the RCU read lock around its call
to afs_check_validity() - and can skip the call entirely if we don't have
a promise.

Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
cc: linux-afs@xxxxxxxxxxxxxxxxxxx
---

fs/afs/callback.c | 2 +
fs/afs/inode.c | 88 ++++++++++++++++++++++----------------------
fs/afs/internal.h | 2 +
fs/afs/rotate.c | 1 +
include/trace/events/afs.h | 8 +++-
5 files changed, 54 insertions(+), 47 deletions(-)

diff --git a/fs/afs/callback.c b/fs/afs/callback.c
index 1dd7543dbf9f..1b4d5809808d 100644
--- a/fs/afs/callback.c
+++ b/fs/afs/callback.c
@@ -60,6 +60,7 @@ void afs_init_callback_state(struct afs_server *server)
rcu_read_lock();
do {
server->cb_s_break++;
+ atomic_inc(&server->cell->fs_s_break);
if (!list_empty(&server->cell->fs_open_mmaps))
queue_work(system_unbound_wq, &server->initcb_work);

@@ -77,6 +78,7 @@ void __afs_break_callback(struct afs_vnode *vnode, enum afs_cb_break_reason reas
clear_bit(AFS_VNODE_NEW_CONTENT, &vnode->flags);
if (test_and_clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
vnode->cb_break++;
+ vnode->cb_v_break = vnode->volume->cb_v_break;
afs_clear_permits(vnode);

if (vnode->lock_state == AFS_VNODE_LOCK_WAITING_FOR_CB)
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 80b6c8d967d5..126daf9969db 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -587,22 +587,32 @@ static void afs_zap_data(struct afs_vnode *vnode)
}

/*
- * Get the server reinit counter for a vnode's current server.
+ * Check to see if we have a server currently serving this volume and that it
+ * hasn't been reinitialised or dropped from the list.
*/
-static bool afs_get_s_break_rcu(struct afs_vnode *vnode, unsigned int *_s_break)
+static bool afs_check_server_good(struct afs_vnode *vnode)
{
- struct afs_server_list *slist = rcu_dereference(vnode->volume->servers);
+ struct afs_server_list *slist;
struct afs_server *server;
+ bool good;
int i;

+ if (vnode->cb_fs_s_break == atomic_read(&vnode->volume->cell->fs_s_break))
+ return true;
+
+ rcu_read_lock();
+
+ slist = rcu_dereference(vnode->volume->servers);
for (i = 0; i < slist->nr_servers; i++) {
server = slist->servers[i].server;
if (server == vnode->cb_server) {
- *_s_break = READ_ONCE(server->cb_s_break);
- return true;
+ good = (vnode->cb_s_break == server->cb_s_break);
+ rcu_read_unlock();
+ return good;
}
}

+ rcu_read_unlock();
return false;
}

@@ -611,57 +621,46 @@ static bool afs_get_s_break_rcu(struct afs_vnode *vnode, unsigned int *_s_break)
*/
bool afs_check_validity(struct afs_vnode *vnode)
{
- struct afs_volume *volume = vnode->volume;
enum afs_cb_break_reason need_clear = afs_cb_break_no_break;
time64_t now = ktime_get_real_seconds();
- bool valid;
- unsigned int cb_break, cb_s_break, cb_v_break;
+ unsigned int cb_break;
int seq = 0;

do {
read_seqbegin_or_lock(&vnode->cb_lock, &seq);
- cb_v_break = READ_ONCE(volume->cb_v_break);
cb_break = vnode->cb_break;

- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
- afs_get_s_break_rcu(vnode, &cb_s_break)) {
- if (vnode->cb_s_break != cb_s_break ||
- vnode->cb_v_break != cb_v_break) {
- vnode->cb_s_break = cb_s_break;
- vnode->cb_v_break = cb_v_break;
- need_clear = afs_cb_break_for_vsbreak;
- valid = false;
- } else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) {
+ if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
+ if (vnode->cb_v_break != vnode->volume->cb_v_break)
+ need_clear = afs_cb_break_for_v_break;
+ else if (!afs_check_server_good(vnode))
+ need_clear = afs_cb_break_for_s_reinit;
+ else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
need_clear = afs_cb_break_for_zap;
- valid = false;
- } else if (vnode->cb_expires_at - 10 <= now) {
+ else if (vnode->cb_expires_at - 10 <= now)
need_clear = afs_cb_break_for_lapsed;
- valid = false;
- } else {
- valid = true;
- }
} else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
- valid = true;
+ ;
} else {
- vnode->cb_v_break = cb_v_break;
- valid = false;
+ need_clear = afs_cb_break_no_promise;
}

} while (need_seqretry(&vnode->cb_lock, seq));

done_seqretry(&vnode->cb_lock, seq);

- if (need_clear != afs_cb_break_no_break) {
- write_seqlock(&vnode->cb_lock);
- if (cb_break == vnode->cb_break)
- __afs_break_callback(vnode, need_clear);
- else
- trace_afs_cb_miss(&vnode->fid, need_clear);
- write_sequnlock(&vnode->cb_lock);
- valid = false;
- }
+ if (need_clear == afs_cb_break_no_break)
+ return true;

- return valid;
+ write_seqlock(&vnode->cb_lock);
+ if (need_clear == afs_cb_break_no_promise)
+ vnode->cb_v_break = vnode->volume->cb_v_break;
+ else if (cb_break == vnode->cb_break)
+ __afs_break_callback(vnode, need_clear);
+ else
+ trace_afs_cb_miss(&vnode->fid, need_clear);
+ write_sequnlock(&vnode->cb_lock);
+ return false;
}

/*
@@ -675,21 +674,20 @@ bool afs_check_validity(struct afs_vnode *vnode)
*/
int afs_validate(struct afs_vnode *vnode, struct key *key)
{
- bool valid;
int ret;

_enter("{v={%llx:%llu} fl=%lx},%x",
vnode->fid.vid, vnode->fid.vnode, vnode->flags,
key_serial(key));

- rcu_read_lock();
- valid = afs_check_validity(vnode);
- rcu_read_unlock();
-
- if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
- clear_nlink(&vnode->vfs_inode);
+ if (unlikely(test_bit(AFS_VNODE_DELETED, &vnode->flags))) {
+ if (vnode->vfs_inode.i_nlink)
+ clear_nlink(&vnode->vfs_inode);
+ goto valid;
+ }

- if (valid)
+ if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
+ afs_check_validity(vnode))
goto valid;

down_write(&vnode->validate_lock);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 0deeb76c67d0..c97618855b46 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -392,6 +392,7 @@ struct afs_cell {
seqlock_t fs_lock; /* For fs_servers */
struct rw_semaphore fs_open_mmaps_lock;
struct list_head fs_open_mmaps; /* List of vnodes that are mmapped */
+ atomic_t fs_s_break; /* Counter of CB.InitCallBackState messages */

/* VL server list. */
rwlock_t vl_servers_lock; /* Lock on vl_servers */
@@ -664,6 +665,7 @@ struct afs_vnode {
struct list_head cb_mmap_link; /* Link in cell->fs_open_mmaps */
void *cb_server; /* Server with callback/filelock */
atomic_t cb_nr_mmap; /* Number of mmaps */
+ unsigned int cb_fs_s_break; /* Mass server break counter (cell->fs_s_break) */
unsigned int cb_s_break; /* Mass break counter on ->server */
unsigned int cb_v_break; /* Mass break counter on ->volume */
unsigned int cb_break; /* Break counter on vnode */
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index d83f13c44b92..79e1a5f6701b 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -374,6 +374,7 @@ bool afs_select_fileserver(struct afs_operation *op)
if (vnode->cb_server != server) {
vnode->cb_server = server;
vnode->cb_s_break = server->cb_s_break;
+ vnode->cb_fs_s_break = atomic_read(&server->cell->fs_s_break);
vnode->cb_v_break = vnode->volume->cb_v_break;
clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
}
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 9f73ed2cf061..bca73e8c8cde 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -306,11 +306,13 @@ enum afs_flock_operation {

enum afs_cb_break_reason {
afs_cb_break_no_break,
+ afs_cb_break_no_promise,
afs_cb_break_for_callback,
afs_cb_break_for_deleted,
afs_cb_break_for_lapsed,
+ afs_cb_break_for_s_reinit,
afs_cb_break_for_unlink,
- afs_cb_break_for_vsbreak,
+ afs_cb_break_for_v_break,
afs_cb_break_for_volume_callback,
afs_cb_break_for_zap,
};
@@ -602,11 +604,13 @@ enum afs_cb_break_reason {

#define afs_cb_break_reasons \
EM(afs_cb_break_no_break, "no-break") \
+ EM(afs_cb_break_no_promise, "no-promise") \
EM(afs_cb_break_for_callback, "break-cb") \
EM(afs_cb_break_for_deleted, "break-del") \
EM(afs_cb_break_for_lapsed, "break-lapsed") \
+ EM(afs_cb_break_for_s_reinit, "s-reinit") \
EM(afs_cb_break_for_unlink, "break-unlink") \
- EM(afs_cb_break_for_vsbreak, "break-vs") \
+ EM(afs_cb_break_for_v_break, "break-v") \
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
E_(afs_cb_break_for_zap, "break-zap")

Next message: David Hildenbrand: "Re: [PATCH v1] hugetlbfs: s390 is always 64bit"
Previous message: David Howells: "[PATCH 5/6] afs: Fix mmap coherency vs 3rd-party changes"
In reply to: David Howells: "[PATCH 5/6] afs: Fix mmap coherency vs 3rd-party changes"
Next in thread: markus . suvanto: "Re: [PATCH 0/6] afs: Fixes for 3rd party-induced data corruption"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]