nfsd: Initial implementation of NFSv4 Courteous Server
Subsystems (4)
Dai Ngo
reviewerv3 cover nfsd: Initial implementation of NFSv4 Courteous Server
Hi Bruce,
This series of patches implement the NFSv4 Courteous Server.
A server which does not immediately expunge the state on lease expiration is known as a Courteous Server. A Courteous Server continues to recognize previously generated state tokens as valid until conflict arises between the expired state and the requests from another client, or the server reboots.
The v2 patch includes the following:
. add new callback, lm_expire_lock, to lock_manager_operations to
allow the lock manager to take appropriate action with conflict lock.
. handle conflicts of NFSv4 locks with NFSv3/NLM and local locks.
. expire courtesy client after 24hr if client has not reconnected.
. do not allow expired client to become courtesy client if there are
waiters for client's locks.
. modify client_info_show to show courtesy client and seconds from
last renew.
. fix a problem with NFSv4.1 server where the it keeps returning
SEQ4_STATUS_CB_PATH_DOWN in the successful SEQUENCE reply, after the courtesy client re-connects, causing the client to keep sending BCTS requests to server.
The v3 patch includes the following:
. modify posix_test_lock to check and resolve conflict locks
for handling of NLM TEST and NFSv4 LOCKT requests.
. separate out fix for back channel stuck in SEQ4_STATUS_CB_PATH_DOWN.
. merge with 5.15-rc1
Dai Ngo
reviewerPATCH v3 1/3 fs/lock: add new callback, lm_expire_lock, to lock_manager_operations
Add new callback, lm_expire_lock, to lock_manager_operations to allow the lock manager to take appropriate action to resolve the lock conflict if possible. The callback takes 2 arguments, file_lock of the blocker and a testonly flag:
testonly = 1 check and return true if lock conflict can be resolved
else return false.
testonly = 0 resolve the conflict if possible, return true if conflict
was resolved esle return false.
Lock manager, such as NFSv4 courteous server, uses this callback to resolve conflict by destroying lock owner, or the NFSv4 courtesy client (client that has expired but allowed to maintains its states) that owns the lock.
--- fs/locks.c | 26 +++++++++++++++++++++++--- include/linux/fs.h | 1 + 2 files changed, 24 insertions(+), 3 deletions(-)
fs/locks.c +23 -3
@@ -954,6 +954,7 @@ posix_test_lock(struct file *filp, struct file_lock *fl)struct file_lock *cfl;struct file_lock_context *ctx;struct inode *inode =locks_inode (filp); +bool ret; ctx =smp_load_acquire (&inode->i_flctx);if (!ctx ||list_empty_careful (&ctx->flc_posix)) {
@@ -962,11 +963,20 @@ posix_test_lock(struct file *filp, struct file_lock *fl) }spin_lock (&ctx->flc_lock); +retry:list_for_each_entry (cfl, &ctx->flc_posix, fl_list) { -if (posix_locks_conflict (fl, cfl)) { -locks_copy_conflock (fl, cfl); - goto out; +if (!posix_locks_conflict (fl, cfl)) +continue ; +if (cfl->fl_lmops && cfl->fl_lmops->lm_expire_lock && + cfl->fl_lmops->lm_expire_lock (cfl,1 )) { +spin_unlock (&ctx->flc_lock); + ret = cfl->fl_lmops->lm_expire_lock (cfl,0 ); +spin_lock (&ctx->flc_lock); +if (ret) + goto retry; } +locks_copy_conflock (fl, cfl); + goto out; } fl->fl_type =F_UNLCK ; out:
@@ -1140,6 +1150,7 @@ static int posix_lock_inode(struct inode *inode, struct file_lock *request,int error;bool added = false;LIST_HEAD (dispose); +bool ret; ctx =locks_get_lock_context (inode, request->fl_type);if (!ctx)
@@ -1166,9 +1177,18 @@ static int posix_lock_inode(struct inode *inode, struct file_lock *request,* blocker's list of waiters and the global blocked_hash. */ if (request->fl_type !=F_UNLCK ) { +retry:list_for_each_entry (fl, &ctx->flc_posix, fl_list) {if (!posix_locks_conflict (request, fl))continue ; +if (fl->fl_lmops && fl->fl_lmops->lm_expire_lock && + fl->fl_lmops->lm_expire_lock (fl,1 )) { +spin_unlock (&ctx->flc_lock); + ret = fl->fl_lmops->lm_expire_lock (fl,0 ); +spin_lock (&ctx->flc_lock); +if (ret) + goto retry; + }if (conflock)locks_copy_conflock (conflock, fl); error = -EAGAIN ;
include/linux/fs.h +1 -0
@@ -1071,6 +1071,7 @@ struct lock_manager_operations {int (*lm_change)(struct file_lock *,int ,struct list_head *);void (*lm_setup)(struct file_lock *,void **);bool (*lm_breaker_owns_lease)(struct file_lock *); +bool (*lm_expire_lock)(struct file_lock *fl,bool testonly); };struct lock_manager {
Signature · 1 line
2.9.5
Dai Ngo
reviewerPATCH v3 2/3 nfsd: Initial implementation of NFSv4 Courteous Server
In reply to Dai NgoCurrently an NFSv4 client must maintain its lease by using the at least one of the state tokens or if nothing else, by issuing a RENEW (4.0), or a singleton SEQUENCE (4.1) at least once during each lease period. If the client fails to renew the lease, for any reason, the Linux server expunges the state tokens immediately upon detection of the "failure to renew the lease" condition and begins returning NFS4ERR_EXPIRED if the client should reconnect and attempt to use the (now) expired state.
The default lease period for the Linux server is 90 seconds. The typical client cuts that in half and will issue a lease renewing operation every 45 seconds. The 90 second lease period is very short considering the potential for moderately long term network partitions. A network partition refers to any loss of network connectivity between the NFS client and the NFS server, regardless of its root cause. This includes NIC failures, NIC driver bugs, network misconfigurations & administrative errors, routers & switches crashing and/or having software updates applied, even down to cables being physically pulled. In most cases, these network failures are transient, although the duration is unknown.
A server which does not immediately expunge the state on lease expiration is known as a Courteous Server. A Courteous Server continues to recognize previously generated state tokens as valid until conflict arises between the expired state and the requests from another client, or the server reboots.
The initial implementation of the Courteous Server will do the following:
. when the laundromat thread detects an expired client and if that client still has established states on the Linux server and there is no waiters for the client's locks then mark the client as a COURTESY_CLIENT and skip destroying the client and all its states, otherwise destroy the client as usual.
. detects conflict of OPEN request with COURTESY_CLIENT, destroys the expired client and all its states, skips the delegation recall then allows the conflicting request to succeed.
. detects conflict of LOCK/LOCKT, NLM LOCK and TEST, and local locks requests with COURTESY_CLIENT, destroys the expired client and all its states then allows the conflicting request to succeed.
--- fs/nfsd/nfs4state.c | 155 ++++++++++++++++++++++++++++++++++++++++++++- fs/nfsd/state.h | 3 + include/linux/sunrpc/svc.h | 1 + 3 files changed, 156 insertions(+), 3 deletions(-)
fs/nfsd/nfs4state.c +152 -3
@@ -125,6 +125,11 @@ static void free_session(struct nfsd4_session *);static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops; +static struct workqueue_struct *laundry_wq; +static void laundromat_main (struct work_struct *); + +static int courtesy_client_expiry = (24 *60 *60 );/* in secs */ +static bool is_session_dead (struct nfsd4_session *ses) {return ses->se_flags &NFS4_SESSION_DEAD ;
@@ -172,6 +177,7 @@ renew_client_locked(struct nfs4_client *clp)list_move_tail (&clp->cl_lru, &nn->client_lru); clp->cl_time =ktime_get_boottime_seconds (); +clear_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); }static void put_client_renew_locked (struct nfs4_client *clp)
@@ -2389,6 +2395,10 @@ static int client_info_show(struct seq_file *m, void *v)seq_puts (m,"status: confirmed\n" ); else seq_puts(m, "status: unconfirmed\n"); +seq_printf (m,"courtesy client: %s\n" , +test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ?"yes" :"no" ); +seq_printf (m,"last renew: %lld secs\n" , +ktime_get_boottime_seconds () - clp->cl_time);seq_printf (m,"name: " );seq_quote_mem (m, clp->cl_name.data, clp->cl_name.len);seq_printf (m,"\nminor version: %d\n" , clp->cl_minorversion);
@@ -4652,6 +4662,42 @@ static void nfsd_break_one_deleg(struct nfs4_delegation *dp)nfsd4_run_cb (&dp->dl_recall); } +/* +* If the conflict happens due to a NFSv4 request then check for +* courtesy client and set rq_conflict_client so that upper layer +* can destroy the conflict client and retry the call. +*/ +static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 ) +return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock); +return ret; +} +/* Called from break_lease() with i_lock held. */ static bool nfsd_break_deleg_cb (struct file_lock *fl)
@@ -4660,6 +4706,8 @@ nfsd_break_deleg_cb(struct file_lock *fl)struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;struct nfs4_file *fp = dp->dl_stid.sc_file; +if (nfsd_check_courtesy_client (dp)) +return false;trace_nfsd_cb_recall (&dp->dl_stid);/*
@@ -5279,6 +5327,22 @@ static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,*/ } +static bool +nfs4_destroy_courtesy_client (struct nfs4_client *clp) +{ +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); + +spin_lock (&nn->client_lock); +if (!test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) || +mark_client_expired_locked (clp)) { +spin_unlock (&nn->client_lock); +return false; + } +spin_unlock (&nn->client_lock); +expire_client (clp); +return true; +} +__be32 nfsd4_process_open2 (struct svc_rqst *rqstp,struct svc_fh *current_fh,struct nfsd4_open *open) {
@@ -5328,7 +5392,13 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf goto out; } }else { + rqstp->rq_conflict_client =NULL ; status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); +if (status == nfserr_jukebox && rqstp->rq_conflict_client) { +if (nfs4_destroy_courtesy_client (rqstp->rq_conflict_client)) + status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); + } +if (status) { stp->st_stid.sc_type =NFS4_CLOSED_STID ;release_open_stateid (stp);
@@ -5562,6 +5632,47 @@ static void nfsd4_ssc_expire_umount(struct nfsd_net *nn) }#endif +static +bool nfs4_anylock_conflict (struct nfs4_client *clp) +{ +int i; +struct nfs4_stateowner *so, *tmp; +struct nfs4_lockowner *lo; +struct nfs4_ol_stateid *stp; +struct nfs4_file *nf; +struct inode *ino; +struct file_lock_context *ctx; +struct file_lock *fl; + +for (i =0 ; i <OWNER_HASH_SIZE ; i++) { +/* scan each lock owner */ +list_for_each_entry_safe (so, tmp, &clp->cl_ownerstr_hashtbl[i], + so_strhash) { +if (so->so_is_open_owner) +continue ; + +/* scan lock states of this lock owner */ + lo =lockowner (so); +list_for_each_entry (stp, &lo->lo_owner.so_stateids, + st_perstateowner) { + nf = stp->st_stid.sc_file; + ino = nf->fi_inode; + ctx = ino->i_flctx; +if (!ctx) +continue ; +/* check each lock belongs to this lock state */ +list_for_each_entry (fl, &ctx->flc_posix, fl_list) { +if (fl->fl_owner != lo) +continue ; +if (!list_empty (&fl->fl_blocked_requests)) +return true; + } + } + } + } +return false; +} +static time64_t nfs4_laundromat (struct nfsd_net *nn) {
@@ -5577,7 +5688,9 @@ nfs4_laundromat(struct nfsd_net *nn) };struct nfs4_cpntf_state *cps;copy_stateid_t *cps_t; +struct nfs4_stid *stid;int i; +int id =0 ;if (clients_still_reclaiming (nn)) { lt.new_timeo =0 ;
@@ -5598,8 +5711,33 @@ nfs4_laundromat(struct nfsd_net *nn)spin_lock (&nn->client_lock);list_for_each_safe (pos, next, &nn->client_lru) { clp = list_entry(pos, struct nfs4_client, cl_lru); +if (test_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags)) { +clear_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); + goto exp_client; + } +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +if (ktime_get_boottime_seconds () >= clp->courtesy_client_expiry) + goto exp_client; +/* +* after umount, v4.0 client is still +* around waiting to be expired +*/ +if (clp->cl_minorversion) +continue ; + }if (!state_expired (<, clp->cl_time))break ; +spin_lock (&clp->cl_lock); + stid =idr_get_next (&clp->cl_stateids, &id); +spin_unlock (&clp->cl_lock); +if (stid && !nfs4_anylock_conflict (clp)) { +/* client still has states */ + clp->courtesy_client_expiry = +ktime_get_boottime_seconds () + courtesy_client_expiry; +set_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); +continue ; + } +exp_client:if (mark_client_expired_locked (clp))continue ;list_add (&clp->cl_lru, &reaplist);
@@ -5679,9 +5817,6 @@ nfs4_laundromat(struct nfsd_net *nn)return max_t (time64_t, lt.new_timeo,NFSD_LAUNDROMAT_MINTIMEOUT ); } -static struct workqueue_struct *laundry_wq; -static void laundromat_main (struct work_struct *); -static void laundromat_main (struct work_struct *laundry) {
@@ -6486,6 +6621,19 @@ nfs4_transform_lock_offset(struct file_lock *lock) lock->fl_end =OFFSET_MAX ; } +/* return true if lock was expired else return false */ +static bool +nfsd4_fl_expire_lock (struct file_lock *fl,bool testonly) +{ +struct nfs4_lockowner *lo = (struct nfs4_lockowner *)fl->fl_owner; +struct nfs4_client *clp = lo->lo_owner.so_client; + +if (testonly) +return test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ? + true : false; +return nfs4_destroy_courtesy_client (clp); +} +static fl_owner_t nfsd4_fl_get_owner (fl_owner_t owner) {
@@ -6533,6 +6681,7 @@ static const struct lock_manager_operations nfsd_posix_mng_ops = { .lm_notify = nfsd4_lm_notify, .lm_get_owner = nfsd4_fl_get_owner, .lm_put_owner = nfsd4_fl_put_owner, + .lm_expire_lock = nfsd4_fl_expire_lock, }; static inline void
fs/nfsd/state.h +3 -0
@@ -345,6 +345,8 @@ struct nfs4_client {#define NFSD4_CLIENT_UPCALL_LOCK (5)/* upcall serialization */ #define NFSD4_CLIENT_CB_FLAG_MASK (1 << NFSD4_CLIENT_CB_UPDATE | \ 1 << NFSD4_CLIENT_CB_KILL) +#define NFSD4_COURTESY_CLIENT (6)/* be nice to expired client */ +#define NFSD4_DESTROY_COURTESY_CLIENT (7)unsigned long cl_flags;const struct cred *cl_cb_cred;struct rpc_clnt *cl_cb_client;
@@ -385,6 +387,7 @@ struct nfs4_client {struct list_head async_copies;/* list of async copies */ spinlock_t async_lock;/* lock for async copies */ atomic_t cl_cb_inflight;/* Outstanding callbacks */ +int courtesy_client_expiry; };/* struct nfs4_client_reset
include/linux/sunrpc/svc.h +1 -0
@@ -306,6 +306,7 @@ struct svc_rqst {* net namespace */ void ** rq_lease_breaker;/* The v4 client breaking a lease */ +void *rq_conflict_client; };#define SVC_NET (rqst) (rqst->rq_xprt ? rqst->rq_xprt->xpt_net : rqst->rq_bc_net)
Signature · 1 line
2.9.5
fs/nfsd/nfs4state.c static int client_info_show(struct seq_file *m, void *v) · line 2400
Quoted 7 lines · Dai Ngo
fs/nfsd/nfs4state.c +3
-0@@ -2389,6 +2395,10 @@ static int client_info_show(struct seq_file *m, void *v)seq_puts (m,"status: confirmed\n" ); else seq_puts(m, "status: unconfirmed\n"); +seq_printf (m,"courtesy client: %s\n" , +test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ?"yes" :"no" ); + seq_printf(m, "last renew: %lld secs\n",
I'd rather keep any units to the left of the colon. Also, "last renew" suggests to me that it's the absolute time of the last renew. Maybe "seconds since last renew:" ?
Quoted 31 lines
fs/nfsd/nfs4state.c +24
-0+ ktime_get_boottime_seconds() - clp->cl_time);seq_printf (m,"name: " );seq_quote_mem (m, clp->cl_name.data, clp->cl_name.len);seq_printf (m,"\nminor version: %d\n" , clp->cl_minorversion);@@ -4652,6 +4662,42 @@ static void nfsd_break_one_deleg(struct nfs4_delegation *dp)nfsd4_run_cb (&dp->dl_recall); } +/* +* If the conflict happens due to a NFSv4 request then check for +* courtesy client and set rq_conflict_client so that upper layer +* can destroy the conflict client and retry the call. +*/ +static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 )
fs/nfsd/nfs4state.c static void nfsd_break_one_deleg(struct nfs4_delegation *dp) · line 4697
Quoted 10 lines
fs/nfsd/nfs4state.c +10
-0+return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock);
Check whether this is safe; I think the flc_lock may be taken inside of this lock elsewhere, resulting in a potential deadlock?
rqst doesn't need any locking as it's only being used by this thread, so it's the client expiration stuff that's the problem, I guess.
--b.
[email protected]
reviewerQuoted 11 lines · J. Bruce Fields
fs/nfsd/nfs4state.c static int client_info_show(struct seq_file *m, void *v) · line 2400
Quoted 7 lines · Dai Ngo
fs/nfsd/nfs4state.c +3
-0@@ -2389,6 +2395,10 @@ static int client_info_show(struct seq_file *m, void *v)seq_puts (m,"status: confirmed\n" ); else seq_puts(m, "status: unconfirmed\n"); +seq_printf (m,"courtesy client: %s\n" , +test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ?"yes" :"no" ); + seq_printf(m, "last renew: %lld secs\n",I'd rather keep any units to the left of the colon. Also, "last renew" suggests to me that it's the absolute time of the last renew. Maybe "seconds since last renew:" ?
will fix in v4.
Quoted 45 lines
Quoted 32 lines
fs/nfsd/nfs4state.c +24
-0+ ktime_get_boottime_seconds() - clp->cl_time);seq_printf (m,"name: " );seq_quote_mem (m, clp->cl_name.data, clp->cl_name.len);seq_printf (m,"\nminor version: %d\n" , clp->cl_minorversion);@@ -4652,6 +4662,42 @@ static void nfsd_break_one_deleg(struct nfs4_delegation *dp)nfsd4_run_cb (&dp->dl_recall); } +/* +* If the conflict happens due to a NFSv4 request then check for +* courtesy client and set rq_conflict_client so that upper layer +* can destroy the conflict client and retry the call. +*/ +static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 )fs/nfsd/nfs4state.c static void nfsd_break_one_deleg(struct nfs4_delegation *dp) · line 4697
Quoted 10 lines
fs/nfsd/nfs4state.c +10
-0+return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock);Check whether this is safe; I think the
flc_lockmay be taken inside of this lock elsewhere, resulting in a potential deadlock?
Quoted 2 lines
rqst doesn't need any locking as it's only being used by this thread, so it's the client expiration stuff that's the problem, I guess.
mark_client_expired_locked needs to acquire cl_lock. I think the lock ordering is ok, client_lock -> cl_lock. nfsd4_exchange_id uses this lock ordering.
I will submit v4 patch with the fix in client_info_show and also new code for handling NFSv4 share reservation conflicts with courtesy clients.
Thanks Bruce,
-Dai
Quoted 2 lines
--b.
Quoted 39 lines · On Wed, Sep 22, 2021 at 03:16:34PM -0700, [email protected] wrote:
Quoted 37 lines · J. Bruce Fields
Quoted 23 lines · Dai Ngo
fs/nfsd/nfs4state.c +23
-0+/* +* If the conflict happens due to a NFSv4 request then check for +* courtesy client and set rq_conflict_client so that upper layer +* can destroy the conflict client and retry the call. +*/ +static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 )fs/nfsd/nfs4state.c static void nfsd_break_one_deleg(struct nfs4_delegation *dp) · line 4697
Quoted 10 lines
fs/nfsd/nfs4state.c +10
-0+return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock);Check whether this is safe; I think the
flc_lockmay be taken inside of this lock elsewhere, resulting in a potential deadlock?
Quoted 6 lines
Quoted 2 lines
rqst doesn't need any locking as it's only being used by this thread, so it's the client expiration stuff that's the problem, I guess.
mark_client_expired_lockedneeds to acquirecl_lock. I think the lock ordering is ok,client_lock->cl_lock.nfsd4_exchange_iduses this lock ordering.
It's flc_lock (see locks.c) that I'm worried about. I've got a lockdep warning here, taking a closer look....
nfsd4_release_lockowner takes clp->cl_lock and then fcl_lock.
Here we're taking fcl_lock and then client_lock.
As you say, elsewhere client_lock is taken and then cl_lock.
So that's the loop, I think.
--b.
[email protected]
reviewerQuoted 47 lines · J. Bruce Fields
Quoted 38 lines · On Wed, Sep 22, 2021 at 03:16:34PM -0700, [email protected] wrote:
Quoted 37 lines · J. Bruce Fields
Quoted 23 lines · Dai Ngo
fs/nfsd/nfs4state.c +23
-0+/* +* If the conflict happens due to a NFSv4 request then check for +* courtesy client and set rq_conflict_client so that upper layer +* can destroy the conflict client and retry the call. +*/ +static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 )fs/nfsd/nfs4state.c static void nfsd_break_one_deleg(struct nfs4_delegation *dp) · line 4697
Quoted 10 lines
fs/nfsd/nfs4state.c +10
-0+return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock);Check whether this is safe; I think the
flc_lockmay be taken inside of this lock elsewhere, resulting in a potential deadlock?Quoted 5 lines
Quoted 2 lines
rqst doesn't need any locking as it's only being used by this thread, so it's the client expiration stuff that's the problem, I guess.
mark_client_expired_lockedneeds to acquirecl_lock. I think the lock ordering is ok,client_lock->cl_lock.nfsd4_exchange_iduses this lock ordering.It's
flc_lock(seelocks.c) that I'm worried about. I've got a lockdep warning here, taking a closer look....
Quoted 7 lines
nfsd4_release_lockownertakesclp->cl_lockand thenfcl_lock.Here we're taking
fcl_lockand thenclient_lock.As you say, elsewhere
client_lockis taken and thencl_lock.So that's the loop, I think.
Thanks Bruce, I see the deadlock. We will need a new approach for this.
-Dai
Quoted 2 lines
--b.
Quoted 5 lines · Dai Ngo
+/*
+ * If the conflict happens due to a NFSv4 request then check for
+ * courtesy client and setrq_conflict_clientso that upper layer
+ * can destroy the conflict client and retry the call.
+ */
I think we need a different approach. Wouldn't we need to take a reference on clp when we assign to rq_conflict_client?
What happens if there are multiple leases found in the loop in __break_lease?
It doesn't seem right that we'd need to treat the case where nfsd is the breaker differently the case where it's another process.
I'm not sure what to suggest instead, though.... Maybe as with locks we need two separate callbacks, one that tests whether there's a courtesy client that needs removing, one that does it after we've dropped the lock.
--b.
Quoted 245 lines
fs/nfsd/nfs4state.c +137
-3+static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 ) +return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock); +return ret; +} +/* Called from break_lease() with i_lock held. */ static bool nfsd_break_deleg_cb (struct file_lock *fl)@@ -4660,6 +4706,8 @@ nfsd_break_deleg_cb(struct file_lock *fl)struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;struct nfs4_file *fp = dp->dl_stid.sc_file; +if (nfsd_check_courtesy_client (dp)) +return false;trace_nfsd_cb_recall (&dp->dl_stid);/* @@ -5279,6 +5327,22 @@ static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,*/ } +static bool +nfs4_destroy_courtesy_client (struct nfs4_client *clp) +{ +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); + +spin_lock (&nn->client_lock); +if (!test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) || +mark_client_expired_locked (clp)) { +spin_unlock (&nn->client_lock); +return false; + } +spin_unlock (&nn->client_lock); +expire_client (clp); +return true; +} +__be32 nfsd4_process_open2 (struct svc_rqst *rqstp,struct svc_fh *current_fh,struct nfsd4_open *open) {@@ -5328,7 +5392,13 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf goto out; } }else { + rqstp->rq_conflict_client =NULL ; status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); +if (status == nfserr_jukebox && rqstp->rq_conflict_client) { +if (nfs4_destroy_courtesy_client (rqstp->rq_conflict_client)) + status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); + } +if (status) { stp->st_stid.sc_type =NFS4_CLOSED_STID ;release_open_stateid (stp);@@ -5562,6 +5632,47 @@ static void nfsd4_ssc_expire_umount(struct nfsd_net *nn) }#endif +static +bool nfs4_anylock_conflict (struct nfs4_client *clp) +{ +int i; +struct nfs4_stateowner *so, *tmp; +struct nfs4_lockowner *lo; +struct nfs4_ol_stateid *stp; +struct nfs4_file *nf; +struct inode *ino; +struct file_lock_context *ctx; +struct file_lock *fl; + +for (i =0 ; i <OWNER_HASH_SIZE ; i++) { +/* scan each lock owner */ +list_for_each_entry_safe (so, tmp, &clp->cl_ownerstr_hashtbl[i], + so_strhash) { +if (so->so_is_open_owner) +continue ; + +/* scan lock states of this lock owner */ + lo =lockowner (so); +list_for_each_entry (stp, &lo->lo_owner.so_stateids, + st_perstateowner) { + nf = stp->st_stid.sc_file; + ino = nf->fi_inode; + ctx = ino->i_flctx; +if (!ctx) +continue ; +/* check each lock belongs to this lock state */ +list_for_each_entry (fl, &ctx->flc_posix, fl_list) { +if (fl->fl_owner != lo) +continue ; +if (!list_empty (&fl->fl_blocked_requests)) +return true; + } + } + } + } +return false; +} +static time64_t nfs4_laundromat (struct nfsd_net *nn) {@@ -5577,7 +5688,9 @@ nfs4_laundromat(struct nfsd_net *nn) };struct nfs4_cpntf_state *cps;copy_stateid_t *cps_t; +struct nfs4_stid *stid;int i; +int id =0 ;if (clients_still_reclaiming (nn)) { lt.new_timeo =0 ;@@ -5598,8 +5711,33 @@ nfs4_laundromat(struct nfsd_net *nn)spin_lock (&nn->client_lock);list_for_each_safe (pos, next, &nn->client_lru) { clp = list_entry(pos, struct nfs4_client, cl_lru); +if (test_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags)) { +clear_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); + goto exp_client; + } +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +if (ktime_get_boottime_seconds () >= clp->courtesy_client_expiry) + goto exp_client; +/* +* after umount, v4.0 client is still +* around waiting to be expired +*/ +if (clp->cl_minorversion) +continue ; + }if (!state_expired (<, clp->cl_time))break ; +spin_lock (&clp->cl_lock); + stid =idr_get_next (&clp->cl_stateids, &id); +spin_unlock (&clp->cl_lock); +if (stid && !nfs4_anylock_conflict (clp)) { +/* client still has states */ + clp->courtesy_client_expiry = +ktime_get_boottime_seconds () + courtesy_client_expiry; +set_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); +continue ; + } +exp_client:if (mark_client_expired_locked (clp))continue ;list_add (&clp->cl_lru, &reaplist);@@ -5679,9 +5817,6 @@ nfs4_laundromat(struct nfsd_net *nn)return max_t (time64_t, lt.new_timeo,NFSD_LAUNDROMAT_MINTIMEOUT ); } -static struct workqueue_struct *laundry_wq; -static void laundromat_main (struct work_struct *); -static void laundromat_main (struct work_struct *laundry) {@@ -6486,6 +6621,19 @@ nfs4_transform_lock_offset(struct file_lock *lock) lock->fl_end =OFFSET_MAX ; } +/* return true if lock was expired else return false */ +static bool +nfsd4_fl_expire_lock (struct file_lock *fl,bool testonly) +{ +struct nfs4_lockowner *lo = (struct nfs4_lockowner *)fl->fl_owner; +struct nfs4_client *clp = lo->lo_owner.so_client; + +if (testonly) +return test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ? + true : false; +return nfs4_destroy_courtesy_client (clp); +} +static fl_owner_t nfsd4_fl_get_owner (fl_owner_t owner) {@@ -6533,6 +6681,7 @@ static const struct lock_manager_operations nfsd_posix_mng_ops = { .lm_notify = nfsd4_lm_notify, .lm_get_owner = nfsd4_fl_get_owner, .lm_put_owner = nfsd4_fl_put_owner, + .lm_expire_lock = nfsd4_fl_expire_lock, }; static inline voidfs/nfsd/state.h +3
-0@@ -345,6 +345,8 @@ struct nfs4_client {#define NFSD4_CLIENT_UPCALL_LOCK (5)/* upcall serialization */ #define NFSD4_CLIENT_CB_FLAG_MASK (1 << NFSD4_CLIENT_CB_UPDATE | \ 1 << NFSD4_CLIENT_CB_KILL) +#define NFSD4_COURTESY_CLIENT (6)/* be nice to expired client */ +#define NFSD4_DESTROY_COURTESY_CLIENT (7)unsigned long cl_flags;const struct cred *cl_cb_cred;struct rpc_clnt *cl_cb_client;@@ -385,6 +387,7 @@ struct nfs4_client {struct list_head async_copies;/* list of async copies */ spinlock_t async_lock;/* lock for async copies */ atomic_t cl_cb_inflight;/* Outstanding callbacks */ +int courtesy_client_expiry; };/* struct nfs4_client_reset include/linux/sunrpc/svc.h +1
-0@@ -306,6 +306,7 @@ struct svc_rqst {* net namespace */ void ** rq_lease_breaker;/* The v4 client breaking a lease */ +void *rq_conflict_client; };#define SVC_NET (rqst) (rqst->rq_xprt ? rqst->rq_xprt->xpt_net : rqst->rq_bc_net)Signature · 1 line
2.9.5
[email protected]
reviewerQuoted 7 lines · J. Bruce Fields
Quoted 5 lines · Dai Ngo
+/*
+ * If the conflict happens due to a NFSv4 request then check for
+ * courtesy client and setrq_conflict_clientso that upper layer
+ * can destroy the conflict client and retry the call.
+ */I think we need a different approach.
I think nfsd_check_courtesy_client is used to handle conflict with delegation. So instead of using rq_conflict_client to let the caller knows and destroy the courtesy client as the current patch does, we can ask the laundromat thread to do the destroy. In that case, nfs4_get_vfs_file in nfsd4_process_open2 will either return no error since the the laufromat destroyed the courtesy client or it gets get nfserr_jukebox which causes the NFS client to retry. By the time the retry comes the courtesy client should already be destroyed.
Quoted 2 lines
Wouldn't we need to take a
reference on clp when we assign torq_conflict_client?
we won't need rq_conflict_client with the new approach.
Quoted 3 lines
What happens if there are multiple leases found in the loop in
__break_lease?
this should no longer be a problem also.
Quoted 7 lines
It doesn't seem right that we'd need to treat the case where nfsd is the breaker differently the case where it's another process.
I'm not sure what to suggest instead, though.... Maybe as with locks we need two separate callbacks, one that tests whether there's a courtesy client that needs removing, one that does it after we've dropped the
I will try the new approach if you don't see any obvious problems with it.
-Dai
Quoted 249 lines
lock.
--b.
Quoted 245 lines
fs/nfsd/nfs4state.c +137
-3+static bool +nfsd_check_courtesy_client (struct nfs4_delegation *dp) +{ +struct svc_rqst *rqst; +struct nfs4_client *clp = dp->dl_recall.cb_clp; +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); +bool ret = false; + +if (!i_am_nfsd ()) { +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +set_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags); +mod_delayed_work (laundry_wq, &nn->laundromat_work,0 ); +return true; + } +return false; + } + rqst =kthread_data (current); +if (rqst->rq_prog !=NFS_PROGRAM || rqst->rq_vers <4 ) +return false; + rqst->rq_conflict_client =NULL ; + +spin_lock (&nn->client_lock); +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) && + !mark_client_expired_locked (clp)) { + rqst->rq_conflict_client = clp; + ret = true; + } +spin_unlock (&nn->client_lock); +return ret; +} +/* Called from break_lease() with i_lock held. */ static bool nfsd_break_deleg_cb (struct file_lock *fl)@@ -4660,6 +4706,8 @@ nfsd_break_deleg_cb(struct file_lock *fl)struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;struct nfs4_file *fp = dp->dl_stid.sc_file; +if (nfsd_check_courtesy_client (dp)) +return false;trace_nfsd_cb_recall (&dp->dl_stid);/* @@ -5279,6 +5327,22 @@ static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,*/ } +static bool +nfs4_destroy_courtesy_client (struct nfs4_client *clp) +{ +struct nfsd_net *nn =net_generic (clp->net, nfsd_net_id); + +spin_lock (&nn->client_lock); +if (!test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) || +mark_client_expired_locked (clp)) { +spin_unlock (&nn->client_lock); +return false; + } +spin_unlock (&nn->client_lock); +expire_client (clp); +return true; +} +__be32 nfsd4_process_open2 (struct svc_rqst *rqstp,struct svc_fh *current_fh,struct nfsd4_open *open) {@@ -5328,7 +5392,13 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf goto out; } }else { + rqstp->rq_conflict_client =NULL ; status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); +if (status == nfserr_jukebox && rqstp->rq_conflict_client) { +if (nfs4_destroy_courtesy_client (rqstp->rq_conflict_client)) + status =nfs4_get_vfs_file (rqstp, fp, current_fh, stp, open); + } +if (status) { stp->st_stid.sc_type =NFS4_CLOSED_STID ;release_open_stateid (stp);@@ -5562,6 +5632,47 @@ static void nfsd4_ssc_expire_umount(struct nfsd_net *nn) }#endif +static +bool nfs4_anylock_conflict (struct nfs4_client *clp) +{ +int i; +struct nfs4_stateowner *so, *tmp; +struct nfs4_lockowner *lo; +struct nfs4_ol_stateid *stp; +struct nfs4_file *nf; +struct inode *ino; +struct file_lock_context *ctx; +struct file_lock *fl; + +for (i =0 ; i <OWNER_HASH_SIZE ; i++) { +/* scan each lock owner */ +list_for_each_entry_safe (so, tmp, &clp->cl_ownerstr_hashtbl[i], + so_strhash) { +if (so->so_is_open_owner) +continue ; + +/* scan lock states of this lock owner */ + lo =lockowner (so); +list_for_each_entry (stp, &lo->lo_owner.so_stateids, + st_perstateowner) { + nf = stp->st_stid.sc_file; + ino = nf->fi_inode; + ctx = ino->i_flctx; +if (!ctx) +continue ; +/* check each lock belongs to this lock state */ +list_for_each_entry (fl, &ctx->flc_posix, fl_list) { +if (fl->fl_owner != lo) +continue ; +if (!list_empty (&fl->fl_blocked_requests)) +return true; + } + } + } + } +return false; +} +static time64_t nfs4_laundromat (struct nfsd_net *nn) {@@ -5577,7 +5688,9 @@ nfs4_laundromat(struct nfsd_net *nn) };struct nfs4_cpntf_state *cps;copy_stateid_t *cps_t; +struct nfs4_stid *stid;int i; +int id =0 ;if (clients_still_reclaiming (nn)) { lt.new_timeo =0 ;@@ -5598,8 +5711,33 @@ nfs4_laundromat(struct nfsd_net *nn)spin_lock (&nn->client_lock);list_for_each_safe (pos, next, &nn->client_lru) { clp = list_entry(pos, struct nfs4_client, cl_lru); +if (test_bit (NFSD4_DESTROY_COURTESY_CLIENT , &clp->cl_flags)) { +clear_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); + goto exp_client; + } +if (test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags)) { +if (ktime_get_boottime_seconds () >= clp->courtesy_client_expiry) + goto exp_client; +/* +* after umount, v4.0 client is still +* around waiting to be expired +*/ +if (clp->cl_minorversion) +continue ; + }if (!state_expired (<, clp->cl_time))break ; +spin_lock (&clp->cl_lock); + stid =idr_get_next (&clp->cl_stateids, &id); +spin_unlock (&clp->cl_lock); +if (stid && !nfs4_anylock_conflict (clp)) { +/* client still has states */ + clp->courtesy_client_expiry = +ktime_get_boottime_seconds () + courtesy_client_expiry; +set_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags); +continue ; + } +exp_client:if (mark_client_expired_locked (clp))continue ;list_add (&clp->cl_lru, &reaplist);@@ -5679,9 +5817,6 @@ nfs4_laundromat(struct nfsd_net *nn)return max_t (time64_t, lt.new_timeo,NFSD_LAUNDROMAT_MINTIMEOUT ); } -static struct workqueue_struct *laundry_wq; -static void laundromat_main (struct work_struct *); -static void laundromat_main (struct work_struct *laundry) {@@ -6486,6 +6621,19 @@ nfs4_transform_lock_offset(struct file_lock *lock) lock->fl_end =OFFSET_MAX ; } +/* return true if lock was expired else return false */ +static bool +nfsd4_fl_expire_lock (struct file_lock *fl,bool testonly) +{ +struct nfs4_lockowner *lo = (struct nfs4_lockowner *)fl->fl_owner; +struct nfs4_client *clp = lo->lo_owner.so_client; + +if (testonly) +return test_bit (NFSD4_COURTESY_CLIENT , &clp->cl_flags) ? + true : false; +return nfs4_destroy_courtesy_client (clp); +} +static fl_owner_t nfsd4_fl_get_owner (fl_owner_t owner) {@@ -6533,6 +6681,7 @@ static const struct lock_manager_operations nfsd_posix_mng_ops = { .lm_notify = nfsd4_lm_notify, .lm_get_owner = nfsd4_fl_get_owner, .lm_put_owner = nfsd4_fl_put_owner, + .lm_expire_lock = nfsd4_fl_expire_lock, }; static inline voidfs/nfsd/state.h +3
-0@@ -345,6 +345,8 @@ struct nfs4_client {#define NFSD4_CLIENT_UPCALL_LOCK (5)/* upcall serialization */ #define NFSD4_CLIENT_CB_FLAG_MASK (1 << NFSD4_CLIENT_CB_UPDATE | \ 1 << NFSD4_CLIENT_CB_KILL) +#define NFSD4_COURTESY_CLIENT (6)/* be nice to expired client */ +#define NFSD4_DESTROY_COURTESY_CLIENT (7)unsigned long cl_flags;const struct cred *cl_cb_cred;struct rpc_clnt *cl_cb_client;@@ -385,6 +387,7 @@ struct nfs4_client {struct list_head async_copies;/* list of async copies */ spinlock_t async_lock;/* lock for async copies */ atomic_t cl_cb_inflight;/* Outstanding callbacks */ +int courtesy_client_expiry; };/* struct nfs4_client_reset include/linux/sunrpc/svc.h +1
-0@@ -306,6 +306,7 @@ struct svc_rqst {* net namespace */ void ** rq_lease_breaker;/* The v4 client breaking a lease */ +void *rq_conflict_client; };#define SVC_NET (rqst) (rqst->rq_xprt ? rqst->rq_xprt->xpt_net : rqst->rq_bc_net)Signature · 1 line
2.9.5
Quoted 7 lines · On Thu, Sep 23, 2021 at 10:09:35AM -0700, [email protected] wrote:
Quoted 6 lines · J. Bruce Fields
Quoted 5 lines · Dai Ngo
+/*
+ * If the conflict happens due to a NFSv4 request then check for
+ * courtesy client and setrq_conflict_clientso that upper layer
+ * can destroy the conflict client and retry the call.
+ */
Quoted 6 lines
Quoted 1 line
I think we need a different approach.
I think
nfsd_check_courtesy_clientis used to handle conflict with delegation. So instead of usingrq_conflict_clientto let the caller knows and destroy the courtesy client as the current patch does, we can ask the laundromat thread to do the destroy.
I can't see right now why that wouldn't work.
Quoted 5 lines
In that case,
nfs4_get_vfs_fileinnfsd4_process_open2will either return no error since the the laufromat destroyed the courtesy client or it gets getnfserr_jukeboxwhich causes the NFS client to retry. By the time the retry comes the courtesy client should already be destroyed.
Make sure this works for local (non-NFS) lease breakers as well. I think that mainly means making sure the !O_NONBLOCK case of __break_lease works.
--b.
[email protected]
reviewerQuoted 15 lines · J. Bruce Fields
Quoted 7 lines · On Thu, Sep 23, 2021 at 10:09:35AM -0700, [email protected] wrote:
Quoted 6 lines · J. Bruce Fields
Quoted 5 lines · Dai Ngo
+/*
+ * If the conflict happens due to a NFSv4 request then check for
+ * courtesy client and setrq_conflict_clientso that upper layer
+ * can destroy the conflict client and retry the call.
+ */Quoted 5 lines
Quoted 1 line
I think we need a different approach.
I think
nfsd_check_courtesy_clientis used to handle conflict with delegation. So instead of usingrq_conflict_clientto let the caller knows and destroy the courtesy client as the current patch does, we can ask the laundromat thread to do the destroy.I can't see right now why that wouldn't work.
Quoted 8 lines
Quoted 5 lines
In that case,
nfs4_get_vfs_fileinnfsd4_process_open2will either return no error since the the laufromat destroyed the courtesy client or it gets getnfserr_jukeboxwhich causes the NFS client to retry. By the time the retry comes the courtesy client should already be destroyed.Make sure this works for local (non-NFS) lease breakers as well. I think that mainly means making sure the !O_NONBLOCK case of
__break_leaseworks.
Yes, local lease breakers use (!O_NONBLOCK). In this case __break_lease will call lm_break then wait for all lease conflicts to be resolved before returning to caller.
-Dai
Quoted 2 lines
--b.
Dai Ngo
reviewerPATCH v3 3/3 nfsd: back channel stuck in SEQ4_STATUS_CB_PATH_DOWN
In reply to Dai NgoWhen the back channel enters SEQ4_STATUS_CB_PATH_DOWN state, the client recovers by sending BIND_CONN_TO_SESSION but the server fails to recover the back channel and leaves it as NFSD4_CB_DOWN.
Fix by enhancing nfsd4_bind_conn_to_session to probe the back channel by calling nfsd4_probe_callback.
--- fs/nfsd/nfs4state.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-)
fs/nfsd/nfs4state.c +13 -3
@@ -3580,7 +3580,7 @@ static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_s }static __be32 nfsd4_match_existing_connection (struct svc_rqst *rqst, -struct nfsd4_session *session,u32 req) +struct nfsd4_session *session,u32 req,struct nfsd4_conn **conn) {struct nfs4_client *clp = session->se_client;struct svc_xprt *xpt = rqst->rq_xprt;
@@ -3603,6 +3603,8 @@ static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,else status = nfserr_inval;spin_unlock (&clp->cl_lock); +if (status == nfs_ok && conn) + *conn = c;return status; }
@@ -3627,8 +3629,16 @@ __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp, status = nfserr_wrong_cred;if (!nfsd4_mach_creds_match (session->se_client, rqstp)) goto out; - status =nfsd4_match_existing_connection (rqstp, session, bcts->dir); -if (status == nfs_ok || status == nfserr_inval) + status =nfsd4_match_existing_connection (rqstp, session, + bcts->dir, &conn); +if (status == nfs_ok) { +if (bcts->dir ==NFS4_CDFC4_FORE_OR_BOTH || + bcts->dir ==NFS4_CDFC4_BACK ) + conn->cn_flags |=NFS4_CDFC4_BACK ; +nfsd4_probe_callback (session->se_client); + goto out; + } +if (status == nfserr_inval) goto out; status =nfsd4_map_bcts_dir (&bcts->dir);if (status)
Signature · 1 line
2.9.5
Re: [PATCH v3 3/3] nfsd: back channel stuck in SEQ4_STATUS_CB_PATH_DOWN
Bruce, Dai -
Quoted 10 lines · Dai Ngo
When the back channel enters
SEQ4_STATUS_CB_PATH_DOWNstate, the client recovers by sendingBIND_CONN_TO_SESSIONbut the server fails to recover the back channel and leaves it asNFSD4_CB_DOWN.Fix by enhancing
nfsd4_bind_conn_to_sessionto probe the back channel by callingnfsd4_probe_callback.Signed-off-by Dai Ngo <[email protected]>
I'm wondering if this one is appropriate to pull into v5.15-rc.
Quoted 48 lines
---
fs/nfsd/nfs4state.c| 16 +++++++++++++---1 file changed, 13 insertions(+), 3 deletions(-)fs/nfsd/nfs4state.c +13
-3@@ -3580,7 +3580,7 @@ static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_s }static __be32 nfsd4_match_existing_connection (struct svc_rqst *rqst, -struct nfsd4_session *session,u32 req) +struct nfsd4_session *session,u32 req,struct nfsd4_conn **conn) {struct nfs4_client *clp = session->se_client;struct svc_xprt *xpt = rqst->rq_xprt;@@ -3603,6 +3603,8 @@ static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,else status = nfserr_inval;spin_unlock (&clp->cl_lock); +if (status == nfs_ok && conn) + *conn = c;return status; }@@ -3627,8 +3629,16 @@ __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp, status = nfserr_wrong_cred;if (!nfsd4_mach_creds_match (session->se_client, rqstp)) goto out; - status =nfsd4_match_existing_connection (rqstp, session, bcts->dir); -if (status == nfs_ok || status == nfserr_inval) + status =nfsd4_match_existing_connection (rqstp, session, + bcts->dir, &conn); +if (status == nfs_ok) { +if (bcts->dir ==NFS4_CDFC4_FORE_OR_BOTH || + bcts->dir ==NFS4_CDFC4_BACK ) + conn->cn_flags |=NFS4_CDFC4_BACK ; +nfsd4_probe_callback (session->se_client); + goto out; + } +if (status == nfserr_inval) goto out; status =nfsd4_map_bcts_dir (&bcts->dir);if (status)Signature · 1 line
2.9.5
--
Chuck Lever
Re: [PATCH v3 3/3] nfsd: back channel stuck in SEQ4_STATUS_CB_PATH_DOWN
Quoted 8 lines · Chuck Lever III
Bruce, Dai -
Quoted 6 lines · Dai Ngo
When the back channel enters
SEQ4_STATUS_CB_PATH_DOWNstate, the client recovers by sendingBIND_CONN_TO_SESSIONbut the server fails to recover the back channel and leaves it asNFSD4_CB_DOWN.
Quoted 6 lines
Quoted 4 lines
Fix by enhancing
nfsd4_bind_conn_to_sessionto probe the back channel by callingnfsd4_probe_callback.Signed-off-by Dai Ngo <[email protected]>I'm wondering if this one is appropriate to pull into v5.15-rc.
I think so.
Dai, do you have a pynfs test for this case?
--b.
Quoted 53 lines
Quoted 48 lines
---
fs/nfsd/nfs4state.c| 16 +++++++++++++---1 file changed, 13 insertions(+), 3 deletions(-)fs/nfsd/nfs4state.c +13
-3@@ -3580,7 +3580,7 @@ static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_s }static __be32 nfsd4_match_existing_connection (struct svc_rqst *rqst, -struct nfsd4_session *session,u32 req) +struct nfsd4_session *session,u32 req,struct nfsd4_conn **conn) {struct nfs4_client *clp = session->se_client;struct svc_xprt *xpt = rqst->rq_xprt;@@ -3603,6 +3603,8 @@ static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,else status = nfserr_inval;spin_unlock (&clp->cl_lock); +if (status == nfs_ok && conn) + *conn = c;return status; }@@ -3627,8 +3629,16 @@ __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp, status = nfserr_wrong_cred;if (!nfsd4_mach_creds_match (session->se_client, rqstp)) goto out; - status =nfsd4_match_existing_connection (rqstp, session, bcts->dir); -if (status == nfs_ok || status == nfserr_inval) + status =nfsd4_match_existing_connection (rqstp, session, + bcts->dir, &conn); +if (status == nfs_ok) { +if (bcts->dir ==NFS4_CDFC4_FORE_OR_BOTH || + bcts->dir ==NFS4_CDFC4_BACK ) + conn->cn_flags |=NFS4_CDFC4_BACK ; +nfsd4_probe_callback (session->se_client); + goto out; + } +if (status == nfserr_inval) goto out; status =nfsd4_map_bcts_dir (&bcts->dir);if (status)Signature · 1 line
2.9.5--
Chuck Lever
[email protected]
reviewerRe: [PATCH v3 3/3] nfsd: back channel stuck in SEQ4_STATUS_CB_PATH_DOWN
Quoted 9 lines · Bruce Fields
Quoted 8 lines · Chuck Lever III
Bruce, Dai -
Quoted 6 lines · Dai Ngo
When the back channel enters
SEQ4_STATUS_CB_PATH_DOWNstate, the client recovers by sendingBIND_CONN_TO_SESSIONbut the server fails to recover the back channel and leaves it asNFSD4_CB_DOWN.
Quoted 8 lines
Quoted 5 lines
Quoted 4 lines
Fix by enhancing
nfsd4_bind_conn_to_sessionto probe the back channel by callingnfsd4_probe_callback.Signed-off-by Dai Ngo <[email protected]>I'm wondering if this one is appropriate to pull into v5.15-rc.
I think so.
Dai, do you have a pynfs test for this case?
I don't, but I can create a pynfs test for reproduce the problem.
-Dai
Quoted 55 lines
--b.
Quoted 52 lines
Quoted 48 lines
---
fs/nfsd/nfs4state.c| 16 +++++++++++++---1 file changed, 13 insertions(+), 3 deletions(-)fs/nfsd/nfs4state.c +13
-3@@ -3580,7 +3580,7 @@ static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_s }static __be32 nfsd4_match_existing_connection (struct svc_rqst *rqst, -struct nfsd4_session *session,u32 req) +struct nfsd4_session *session,u32 req,struct nfsd4_conn **conn) {struct nfs4_client *clp = session->se_client;struct svc_xprt *xpt = rqst->rq_xprt;@@ -3603,6 +3603,8 @@ static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,else status = nfserr_inval;spin_unlock (&clp->cl_lock); +if (status == nfs_ok && conn) + *conn = c;return status; }@@ -3627,8 +3629,16 @@ __be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp, status = nfserr_wrong_cred;if (!nfsd4_mach_creds_match (session->se_client, rqstp)) goto out; - status =nfsd4_match_existing_connection (rqstp, session, bcts->dir); -if (status == nfs_ok || status == nfserr_inval) + status =nfsd4_match_existing_connection (rqstp, session, + bcts->dir, &conn); +if (status == nfs_ok) { +if (bcts->dir ==NFS4_CDFC4_FORE_OR_BOTH || + bcts->dir ==NFS4_CDFC4_BACK ) + conn->cn_flags |=NFS4_CDFC4_BACK ; +nfsd4_probe_callback (session->se_client); + goto out; + } +if (status == nfserr_inval) goto out; status =nfsd4_map_bcts_dir (&bcts->dir);if (status)Signature · 1 line
2.9.5--
Chuck Lever
[email protected]
reviewerRe: [PATCH v3 3/3] nfsd: back channel stuck in SEQ4_STATUS_CB_PATH_DOWN
Quoted 14 lines · On 9/16/21 1:15 PM, [email protected] wrote:
Quoted 12 lines · Bruce Fields
Quoted 11 lines · Chuck Lever III
Bruce, Dai -
Quoted 9 lines · Dai Ngo
When the back channel enters
SEQ4_STATUS_CB_PATH_DOWNstate, the client recovers by sendingBIND_CONN_TO_SESSIONbut the server fails to recover the back channel and leaves it asNFSD4_CB_DOWN.Fix by enhancing
nfsd4_bind_conn_to_sessionto probe the back channel by callingnfsd4_probe_callback.
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
Quoted 1 line
Signed-off-by Dai Ngo <[email protected]>I'm wondering if this one is appropriate to pull into v5.15-rc.
I think so.
Dai, do you have a pynfs test for this case?
I don't, but I can create a pynfs test for reproduce the problem.
Here are the steps to reproduce the stuck SEQ4_STATUS_CB_PATH_DOWN problem using 'tcpkill':
Client: 5.13.0-rc2
Server: 5.15.0-rc1
1. [root@nfsvmd07 ~]# mount -o vers=4.1 nfsvme14:/root/xfs /tmp/mnt
2. [root@nfsvmd07 ~]# tcpkill host nfsvme14 and port 2049
3. [root@nfsvmd07 ~]# ls /tmp/mnt
4. CTRL-C to stop tcpkill
5. [root@nfsvmd07 ~]# ls /tmp/mnt
The problem can be observed in the wire trace where the back channel in stuck in SEQ4_STATUS_CB_PATH_DOWN causing the client to keep sending BCTS.
Note: this problem can only be reproduced with client running 5.13 or older. Client with 5.14 or newer does not have this problem. The reason is in 5.13, when the client re-establishes the TCP connection it re-uses the previous port number which was destroyed by tcpkill (client sends RST to server). This causes the server to set the state of the back channel to SEQ4_STATUS_CB_PATH_DOWN. In 5.14, the client uses a new port number when re-establish the connection this results in server returning NFS4ERR_CONN_NOT_BOUND_TO_SESSION in the reply of the stand-alone SEQUENCE which the causes the client to send BCTS once re-establish the back channel successfully.
I can provide the pcap files of a good and bad run of the test if interested.
I don't have pynfs test for this case.
-Dai
Quoted 63 lines
-Dai
Quoted 60 lines
--b.
Quoted 57 lines
Quoted 53 lines
---
fs/nfsd/nfs4state.c| 16 +++++++++++++---1 file changed, 13 insertions(+), 3 deletions(-)fs/nfsd/nfs4state.c +1
-1@@ -3580,7 +3580,7 @@ static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_s } static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst, - struct nfsd4_session *session, u32 req) + struct nfsd4_session *session, u32 req, struct nfsd4_conn **conn) {struct
nfs4_client*clp =session->se_client;
structsvc_xprt*xpt =rqst->rq_xprt;fs/nfsd/nfs4state.c +0
-0@@ -3603,6 +3603,8 @@ static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,else
status =nfserr_inval;
spin_unlock(&clp->cl_lock);
+ if (status ==nfs_ok&& conn)
+ *conn = c;
return status;
}fs/nfsd/nfs4state.c +10
-2@@ -3627,8 +3629,16 @@ __be32 nfsd4_bind_conn_to_session(structsvc_rqst *rqstp, status = nfserr_wrong_cred; if (!nfsd4_mach_creds_match(session->se_client, rqstp)) goto out; - status = nfsd4_match_existing_connection(rqstp, session, bcts->dir); - if (status == nfs_ok || status == nfserr_inval) + status = nfsd4_match_existing_connection(rqstp, session, + bcts->dir, &conn); + if (status == nfs_ok) { + if (bcts->dir == NFS4_CDFC4_FORE_OR_BOTH || + bcts->dir == NFS4_CDFC4_BACK) + conn->cn_flags |= NFS4_CDFC4_BACK; + nfsd4_probe_callback(session->se_client); + goto out; + } + if (status == nfserr_inval) goto out;status =
nfsd4_map_bcts_dir(&bcts->dir);
if (status)Signature · 1 line
2.9.5Signature · 1 line
Chuck Lever
I haven't tried to figure out why, but I notice after these patches that pynfs tests RENEW3, LKU10, CLOSE9, and CLOSE8 are failing with unexpected share denied errors.
--b.
[email protected]
reviewerQuoted 2 lines · J. Bruce Fields
I haven't tried to figure out why, but I notice after these patches that pynfs tests RENEW3
The failure is related to share reservation, will be fixed when we have code that handles share reservation with courtesy client. However, with courtesy client support, the test will need to be modified since the expected result will be NFS4_OK instead of NFS4ERR_EXPIRE.
Quoted 2 lines
, LKU10, CLOSE9, and CLOSE8 are failing with
unexpected share denied errors.
I suspected these tests are also related to share reservation. However, I had problems running these tests, they are skipped. For example:
[root@nfsvmf25 nfs4.0]# ./testserver.py $server -v CLOSE9
**************************************************
**************************************************
Command line asked for 1 of 673 tests
Of those: 1 Skipped, 0 Failed, 0 Warned, 0 Passed [root@nfsvmf25 nfs4.0]#
Do I need to do any special setup to run these tests?
Thanks,
-Dai
Quoted 2 lines
--b.
[email protected]
reviewerQuoted 9 lines · On 9/23/21 10:15 AM, [email protected] wrote:
Quoted 2 lines · J. Bruce Fields
I haven't tried to figure out why, but I notice after these patches that pynfs tests RENEW3
The failure is related to share reservation, will be fixed when we have code that handles share reservation with courtesy client. However, with courtesy client support, the test will need to be modified since the expected result will be
NFS4_OKinstead ofNFS4ERR_EXPIRE.
correction, with the patch for handling share reservation conflict, this test now passes with NFS4ERR_EXPIRE as expected since the courtesy client was destroyed.
-Dai
Quoted 7 lines
Quoted 3 lines
, LKU10, CLOSE9, and CLOSE8 are failing with
unexpected share denied errors.I suspected these tests are also related to share reservation. However, I had problems running these tests, they are skipped. For example:
Quoted 8 lines
[root@nfsvmf25 nfs4.0]# ./testserver.py $server -v CLOSE9
**************************************************
**************************************************
Command line asked for 1 of 673 tests
Of those: 1 Skipped, 0 Failed, 0 Warned, 0 Passed [root@nfsvmf25 nfs4.0]#Do I need to do any special setup to run these tests?
still trying to figure out why these tests are skipped on my setup.
-Dai
Quoted 6 lines
Thanks,
-DaiQuoted 2 lines
--b.