@@ -165,7 +165,11 @@ struct afs_status_cb {
* AFS volume synchronisation information
*/
struct afs_volsync {
+ unsigned int mask; /* Bitmask of supplied fields */
+#define AFS_VOLSYNC_CREATION 0x01
+#define AFS_VOLSYNC_UPDATE 0x02
time64_t creation; /* volume creation time */
+ time64_t update; /* Volume update time */
};
/*
@@ -81,7 +81,7 @@ 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;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
afs_clear_permits(vnode);
if (vnode->lock_state == AFS_VNODE_LOCK_WAITING_FOR_CB)
@@ -158,12 +158,13 @@ static void afs_break_one_callback(struct afs_volume *volume,
struct super_block *sb;
struct afs_vnode *vnode;
struct inode *inode;
+ unsigned int cb_v_break;
if (fid->vnode == 0 && fid->unique == 0) {
/* The callback break applies to an entire volume. */
write_lock(&volume->cb_v_break_lock);
- volume->cb_v_break++;
- trace_afs_cb_break(fid, volume->cb_v_break,
+ cb_v_break = atomic_inc_return(&volume->cb_v_break);
+ trace_afs_cb_break(fid, cb_v_break,
afs_cb_break_for_volume_callback, false);
write_unlock(&volume->cb_v_break_lock);
return;
@@ -35,11 +35,14 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
key_get(key);
}
- op->key = key;
- op->volume = afs_get_volume(volume, afs_volume_trace_get_new_op);
- op->net = volume->cell->net;
- op->cb_v_break = volume->cb_v_break;
- op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
+ op->key = key;
+ op->volume = afs_get_volume(volume, afs_volume_trace_get_new_op);
+ op->net = volume->cell->net;
+ op->cb_v_break = atomic_read(&volume->cb_v_break);
+ op->pre_volsync.mask = READ_ONCE(volume->volsync.mask);
+ op->pre_volsync.creation = READ_ONCE(volume->volsync.creation);
+ op->pre_volsync.update = READ_ONCE(volume->volsync.update);
+ op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
op->nr_iterations = -1;
afs_op_set_error(op, -EDESTADDRREQ);
@@ -147,7 +150,7 @@ bool afs_begin_vnode_operation(struct afs_operation *op)
afs_prepare_vnode(op, &op->file[0], 0);
afs_prepare_vnode(op, &op->file[1], 1);
- op->cb_v_break = op->volume->cb_v_break;
+ op->cb_v_break = atomic_read(&op->volume->cb_v_break);
_leave(" = true");
return true;
}
@@ -165,8 +165,10 @@ static void xdr_decode_AFSVolSync(const __be32 **_bp,
bp++; /* spare6 */
*_bp = bp;
- if (volsync)
+ if (volsync) {
volsync->creation = creation;
+ volsync->mask |= AFS_VOLSYNC_CREATION;
+ }
}
/*
@@ -1870,7 +1872,10 @@ static int afs_deliver_fs_inline_bulk_status(struct afs_call *call)
return ret;
bp = call->buffer;
- xdr_decode_AFSVolSync(&bp, &op->volsync);
+ /* Unfortunately, prior to OpenAFS-1.6, volsync here is filled
+ * with rubbish.
+ */
+ xdr_decode_AFSVolSync(&bp, NULL);
call->unmarshall++;
fallthrough;
@@ -329,6 +329,8 @@ static void afs_fetch_status_success(struct afs_operation *op)
struct afs_vnode *vnode = vp->vnode;
int ret;
+ afs_update_volume_state(op, vp);
+
if (vnode->netfs.inode.i_state & I_NEW) {
ret = afs_inode_init_from_status(op, vp, vnode);
afs_op_set_error(op, ret);
@@ -542,7 +544,7 @@ struct inode *afs_root_iget(struct super_block *sb, struct key *key)
BUG_ON(!(inode->i_state & I_NEW));
vnode = AFS_FS_I(inode);
- vnode->cb_v_break = as->volume->cb_v_break,
+ vnode->cb_v_break = atomic_read(&as->volume->cb_v_break),
afs_set_netfs_context(vnode);
op = afs_alloc_operation(key, as->volume);
@@ -631,6 +631,7 @@ struct afs_volume {
afs_volid_t vids[AFS_MAXTYPES]; /* All associated volume IDs */
refcount_t ref;
time64_t update_at; /* Time at which to next update */
+ struct afs_volsync volsync; /* Volume info from server */
struct afs_cell *cell; /* Cell to which belongs (pins ref) */
struct rb_node cell_node; /* Link in cell->volumes */
struct hlist_node proc_link; /* Link in cell->proc_volumes */
@@ -651,7 +652,10 @@ struct afs_volume {
rwlock_t servers_lock; /* Lock for ->servers */
unsigned int servers_seq; /* Incremented each time ->servers changes */
- unsigned cb_v_break; /* Break-everything counter. */
+ /* Callback management */
+ atomic_t cb_ro_snapshot; /* RO volume update-from-snapshot counter */
+ atomic_t cb_v_break; /* Volume-break event counter. */
+ atomic_t cb_scrub; /* Scrub-all-data event counter. */
rwlock_t cb_v_break_lock;
afs_voltype_t type; /* type of volume */
@@ -845,7 +849,8 @@ struct afs_operation {
struct afs_volume *volume; /* Volume being accessed */
struct afs_vnode_param file[2];
struct afs_vnode_param *more_files;
- struct afs_volsync volsync;
+ struct afs_volsync pre_volsync; /* Volsync before op */
+ struct afs_volsync volsync; /* Volsync returned by op */
struct dentry *dentry; /* Dentry to be altered */
struct dentry *dentry_2; /* Second dentry to be altered */
struct timespec64 mtime; /* Modification time to record */
@@ -1052,7 +1057,7 @@ static inline unsigned int afs_calc_vnode_cb_break(struct afs_vnode *vnode)
static inline bool afs_cb_is_broken(unsigned int cb_break,
const struct afs_vnode *vnode)
{
- return cb_break != (vnode->cb_break + vnode->volume->cb_v_break);
+ return cb_break != (vnode->cb_break + atomic_read(&vnode->volume->cb_v_break));
}
/*
@@ -1544,6 +1549,7 @@ extern void afs_fs_exit(void);
/*
* validation.c
*/
+void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp);
bool afs_check_validity(struct afs_vnode *vnode);
bool afs_pagecache_valid(struct afs_vnode *vnode);
int afs_validate(struct afs_vnode *vnode, struct key *key);
@@ -482,7 +482,7 @@ bool afs_select_fileserver(struct afs_operation *op)
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;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
}
@@ -10,6 +10,109 @@
#include <linux/sched.h>
#include "internal.h"
+/*
+ * Handle a change to the volume creation time in the VolSync record.
+ */
+static void afs_update_volume_creation_time(struct afs_operation *op, struct afs_volume *volume)
+{
+ enum afs_cb_break_reason reason = afs_cb_break_for_vos_release;
+ unsigned int snap;
+ time64_t cur = READ_ONCE(volume->volsync.creation);
+ time64_t old = op->pre_volsync.creation;
+ time64_t new = op->volsync.creation;
+
+ _enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
+
+ if (!(op->volsync.mask & AFS_VOLSYNC_CREATION) ||
+ !(volume->volsync.mask & AFS_VOLSYNC_CREATION)) {
+ WRITE_ONCE(volume->volsync.creation, new);
+ volume->volsync.mask |= AFS_VOLSYNC_CREATION;
+ return;
+ }
+
+ if (likely(new == cur))
+ return;
+
+ /* If the creation time changes in an unexpected way, we need to scrub
+ * our caches. For a RW vol, this will only change if the volume is
+ * restored from a backup; for a RO/Backup vol, this will advance when
+ * the volume is updated to a new snapshot (eg. "vos release").
+ */
+ if (volume->type == AFSVL_RWVOL || new < old)
+ reason = afs_cb_break_for_creation_regress;
+
+ /* Try to advance the creation timestamp from what we had before the
+ * operation to what we got back from the server. This should
+ * hopefully ensure that in a race between multiple operations only one
+ * of them will do this.
+ */
+ if (try_cmpxchg(&volume->volsync.creation, &old, new)) {
+ if (reason == afs_cb_break_for_creation_regress)
+ atomic_inc(&volume->cb_scrub);
+ else if (volume->type != AFSVL_RWVOL)
+ snap = atomic_inc_return(&volume->cb_ro_snapshot);
+ trace_afs_cb_v_break(volume->vid, snap, reason);
+ }
+}
+
+/*
+ * Handle a change to the volume update time in the VolSync record.
+ */
+static void afs_update_volume_update_time(struct afs_operation *op, struct afs_volume *volume)
+{
+ enum afs_cb_break_reason reason = afs_cb_break_no_break;
+ time64_t cur = READ_ONCE(volume->volsync.update);
+ time64_t old = op->pre_volsync.update;
+ time64_t new = op->volsync.update;
+
+ _enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
+
+ if (!(op->volsync.mask & AFS_VOLSYNC_UPDATE) ||
+ !(volume->volsync.mask & AFS_VOLSYNC_UPDATE)) {
+ WRITE_ONCE(volume->volsync.update, new);
+ volume->volsync.mask |= AFS_VOLSYNC_UPDATE;
+ return;
+ }
+
+ if (likely(new == cur))
+ return;
+
+ /* If the volume update time changes in an unexpected way, we need to
+ * scrub our caches. For a RW vol, this will advance on every
+ * modification op; for a RO/Backup vol, this will advance when the
+ * volume is updated to a new snapshot (eg. "vos release").
+ */
+ if (new < old)
+ reason = afs_cb_break_for_update_regress;
+
+ /* Try to advance the update timestamp from what we had before the
+ * operation to what we got back from the server. This should
+ * hopefully ensure that in a race between multiple operations only one
+ * of them will do this.
+ */
+ if (try_cmpxchg(&volume->volsync.update, &old, new)) {
+ if (reason == afs_cb_break_for_update_regress) {
+ atomic_inc(&volume->cb_scrub);
+ trace_afs_cb_v_break(volume->vid, 0, reason);
+ }
+ }
+}
+
+/*
+ * Update the state of a volume.
+ */
+void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp)
+{
+ struct afs_volume *volume = op->volume;
+
+ _enter("%llx", op->volume->vid);
+
+ if (op->volsync.mask & AFS_VOLSYNC_CREATION)
+ afs_update_volume_creation_time(op, volume);
+ if (op->volsync.mask & AFS_VOLSYNC_UPDATE)
+ afs_update_volume_update_time(op, volume);
+}
+
/*
* mark the data attached to an inode as obsolete due to a write on the server
* - might also want to ditch all the outstanding writes and dirty pages
@@ -74,7 +177,7 @@ bool afs_check_validity(struct afs_vnode *vnode)
cb_break = vnode->cb_break;
if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- if (vnode->cb_v_break != vnode->volume->cb_v_break)
+ if (vnode->cb_v_break != atomic_read(&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;
@@ -97,7 +200,7 @@ bool afs_check_validity(struct afs_vnode *vnode)
write_seqlock(&vnode->cb_lock);
if (need_clear == afs_cb_break_no_promise)
- vnode->cb_v_break = vnode->volume->cb_v_break;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
else if (cb_break == vnode->cb_break)
__afs_break_callback(vnode, need_clear);
else
@@ -245,12 +245,16 @@ static void xdr_decode_YFSVolSync(const __be32 **_bp,
struct afs_volsync *volsync)
{
struct yfs_xdr_YFSVolSync *x = (void *)*_bp;
- u64 creation;
+ u64 creation, update;
if (volsync) {
creation = xdr_to_u64(x->vol_creation_date);
do_div(creation, 10 * 1000 * 1000);
volsync->creation = creation;
+ update = xdr_to_u64(x->vol_update_date);
+ do_div(update, 10 * 1000 * 1000);
+ volsync->update = update;
+ volsync->mask |= AFS_VOLSYNC_CREATION | AFS_VOLSYNC_UPDATE;
}
*_bp += xdr_size(x);
@@ -440,12 +440,15 @@ enum yfs_cm_operation {
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_creation_regress, "creation-regress") \
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_update_regress, "update-regress") \
EM(afs_cb_break_for_v_break, "break-v") \
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
+ EM(afs_cb_break_for_vos_release, "break-vos-release") \
E_(afs_cb_break_for_zap, "break-zap")
/*
@@ -1249,6 +1252,30 @@ TRACE_EVENT(afs_get_tree,
__entry->cell, __entry->volume, __entry->vid)
);
+TRACE_EVENT(afs_cb_v_break,
+ TP_PROTO(afs_volid_t vid, unsigned int cb_v_break,
+ enum afs_cb_break_reason reason),
+
+ TP_ARGS(vid, cb_v_break, reason),
+
+ TP_STRUCT__entry(
+ __field(afs_volid_t, vid)
+ __field(unsigned int, cb_v_break)
+ __field(enum afs_cb_break_reason, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->vid = vid;
+ __entry->cb_v_break = cb_v_break;
+ __entry->reason = reason;
+ ),
+
+ TP_printk("%llx vb=%x %s",
+ __entry->vid,
+ __entry->cb_v_break,
+ __print_symbolic(__entry->reason, afs_cb_break_reasons))
+ );
+
TRACE_EVENT(afs_cb_break,
TP_PROTO(struct afs_fid *fid, unsigned int cb_break,
enum afs_cb_break_reason reason, bool skipped),
A number of fileserver RPC operations return a VolSync record as part of their reply that gives some information about the state of the volume being accessed, including: (1) A volume Creation timestamp. For an RW volume, this is the time at which the volume was created; if it changes, the RW volume was presumably restored from a backup and all cached data should be scrubbed as Data Version numbers could regress on the files in the volume. For an RO volume, this is the time it was last snapshotted from the RW volume. It is expected to advance each time this happens; if it regresses, cached data should be scrubbed. (2) A volume Update timestamp (Auristor only). For an RW volume, this is updated any time any change is made to a volume or its contents. If it regresses, all cached data must be scrubbed. For an RO volume, this is a copy of the RW volume's Update timestamp at the point of snapshotting. It can be used as a version number when checking to see if a callback on a RO volume was due to a snapshot. If it regresses, all cached data must be scrubbed. but this is currently not made use of by the in-kernel afs filesystem. Make the afs filesystem use this by: (1) Add a mask to the afs_volsync struct to record what fields are set there and add an update time. (2) Add an afs_volsync struct to the afs_volume struct and use this to track the two timestamps in the afs_volume struct. (3) Add a 'pre-op volsync' struct to the afs_operation struct to record the state of the volume tracking before the op. (4) Add a new counter, cb_scrub, to the afs_volume struct to count events that require all data to be scrubbed. A copy is placed in the afs_vnode struct (inode) and if they no longer match, a scrub takes place. (5) When the result of an operation is being parsed, parse the VolSync data too, if it is provided. Note that the two timestamps are handled separately, since they don't work in quite the same way. - If the afs_volume tracking is unset, just set it and do nothing else. - If the result timestamps are the same as the ones in afs_volume, do nothing. - If the timestamps regress, propose incrementing cb_scrub. - If the RW creation timestamp changes, propose incrementing cb_scrub. - If any timestamp changes, propose incrementing cb_v_break, indicating updates may be required. - If an update cb_scrub or cb_v_break is proposed, try to cmpxchg the copy of the timestamp in afs_volume from the pre-op version to the result version and only update if successful. This should help mitigate the case where a bunch of parallel ops all see the change and all try crank the counters. Signed-off-by: David Howells <dhowells@redhat.com> cc: Marc Dionne <marc.dionne@auristor.com> cc: linux-afs@lists.infradead.org --- fs/afs/afs.h | 4 ++ fs/afs/callback.c | 7 +-- fs/afs/fs_operation.c | 15 +++--- fs/afs/fsclient.c | 9 +++- fs/afs/inode.c | 4 +- fs/afs/internal.h | 12 +++-- fs/afs/rotate.c | 2 +- fs/afs/validation.c | 107 ++++++++++++++++++++++++++++++++++++- fs/afs/yfsclient.c | 6 ++- include/trace/events/afs.h | 27 ++++++++++ 10 files changed, 174 insertions(+), 19 deletions(-)