aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--drivers/staging/android/ashmem.c2
-rw-r--r--fs/ioctl.c2
-rw-r--r--fs/lockd/mon.c2
-rw-r--r--fs/lockd/svc.c2
-rw-r--r--fs/nfsd/nfs4proc.c57
-rw-r--r--fs/nfsd/nfs4state.c68
-rw-r--r--fs/nfsd/nfs4xdr.c34
-rw-r--r--fs/nfsd/nfscache.c4
-rw-r--r--fs/nfsd/nfsctl.c6
-rw-r--r--fs/nfsd/nfsfh.c2
-rw-r--r--fs/nfsd/nfssvc.c2
-rw-r--r--fs/nfsd/state.h19
-rw-r--r--fs/nfsd/vfs.c37
-rw-r--r--fs/nfsd/vfs.h2
-rw-r--r--fs/nfsd/xdr4.h9
-rw-r--r--fs/open.c5
-rw-r--r--include/linux/fs.h2
-rw-r--r--include/linux/sunrpc/svc.h34
-rw-r--r--include/linux/sunrpc/svc_xprt.h7
-rw-r--r--include/trace/events/sunrpc.h120
-rw-r--r--mm/madvise.c2
-rw-r--r--net/sunrpc/auth_gss/svcauth_gss.c2
-rw-r--r--net/sunrpc/cache.c26
-rw-r--r--net/sunrpc/svc.c42
-rw-r--r--net/sunrpc/svc_xprt.c292
-rw-r--r--net/sunrpc/svcsock.c5
-rw-r--r--net/sunrpc/xdr.c9
27 files changed, 525 insertions, 269 deletions
diff --git a/drivers/staging/android/ashmem.c b/drivers/staging/android/ashmem.c
index 46f8ef42559e..8c7852742f4b 100644
--- a/drivers/staging/android/ashmem.c
+++ b/drivers/staging/android/ashmem.c
@@ -446,7 +446,7 @@ ashmem_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
446 loff_t start = range->pgstart * PAGE_SIZE; 446 loff_t start = range->pgstart * PAGE_SIZE;
447 loff_t end = (range->pgend + 1) * PAGE_SIZE; 447 loff_t end = (range->pgend + 1) * PAGE_SIZE;
448 448
449 do_fallocate(range->asma->file, 449 vfs_fallocate(range->asma->file,
450 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, 450 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
451 start, end - start); 451 start, end - start);
452 range->purged = ASHMEM_WAS_PURGED; 452 range->purged = ASHMEM_WAS_PURGED;
diff --git a/fs/ioctl.c b/fs/ioctl.c
index 77c9a7812542..214c3c11fbc2 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -443,7 +443,7 @@ int ioctl_preallocate(struct file *filp, void __user *argp)
443 return -EINVAL; 443 return -EINVAL;
444 } 444 }
445 445
446 return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len); 446 return vfs_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
447} 447}
448 448
449static int file_ioctl(struct file *filp, unsigned int cmd, 449static int file_ioctl(struct file *filp, unsigned int cmd,
diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c
index 9106f42c472c..1cc6ec51e6b1 100644
--- a/fs/lockd/mon.c
+++ b/fs/lockd/mon.c
@@ -214,7 +214,7 @@ int nsm_monitor(const struct nlm_host *host)
214 if (unlikely(res.status != 0)) 214 if (unlikely(res.status != 0))
215 status = -EIO; 215 status = -EIO;
216 if (unlikely(status < 0)) { 216 if (unlikely(status < 0)) {
217 printk(KERN_NOTICE "lockd: cannot monitor %s\n", nsm->sm_name); 217 pr_notice_ratelimited("lockd: cannot monitor %s\n", nsm->sm_name);
218 return status; 218 return status;
219 } 219 }
220 220
diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index d1bb7ecfd201..e94c887da2d7 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -350,7 +350,7 @@ static struct svc_serv *lockd_create_svc(void)
350 printk(KERN_WARNING 350 printk(KERN_WARNING
351 "lockd_up: no pid, %d users??\n", nlmsvc_users); 351 "lockd_up: no pid, %d users??\n", nlmsvc_users);
352 352
353 serv = svc_create(&nlmsvc_program, LOCKD_BUFSIZE, NULL); 353 serv = svc_create(&nlmsvc_program, LOCKD_BUFSIZE, svc_rpcb_cleanup);
354 if (!serv) { 354 if (!serv) {
355 printk(KERN_WARNING "lockd_up: create service failed\n"); 355 printk(KERN_WARNING "lockd_up: create service failed\n");
356 return ERR_PTR(-ENOMEM); 356 return ERR_PTR(-ENOMEM);
diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 0beb023f25ac..ac71d13c69ef 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -33,6 +33,7 @@
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */ 34 */
35#include <linux/file.h> 35#include <linux/file.h>
36#include <linux/falloc.h>
36#include <linux/slab.h> 37#include <linux/slab.h>
37 38
38#include "idmap.h" 39#include "idmap.h"
@@ -772,7 +773,7 @@ nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
772 * the client wants us to do more in this compound: 773 * the client wants us to do more in this compound:
773 */ 774 */
774 if (!nfsd4_last_compound_op(rqstp)) 775 if (!nfsd4_last_compound_op(rqstp))
775 rqstp->rq_splice_ok = false; 776 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
776 777
777 /* check stateid */ 778 /* check stateid */
778 if ((status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), 779 if ((status = nfs4_preprocess_stateid_op(SVC_NET(rqstp),
@@ -1014,6 +1015,44 @@ nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1014} 1015}
1015 1016
1016static __be32 1017static __be32
1018nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1019 struct nfsd4_fallocate *fallocate, int flags)
1020{
1021 __be32 status = nfserr_notsupp;
1022 struct file *file;
1023
1024 status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), cstate,
1025 &fallocate->falloc_stateid,
1026 WR_STATE, &file);
1027 if (status != nfs_ok) {
1028 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1029 return status;
1030 }
1031
1032 status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, file,
1033 fallocate->falloc_offset,
1034 fallocate->falloc_length,
1035 flags);
1036 fput(file);
1037 return status;
1038}
1039
1040static __be32
1041nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1042 struct nfsd4_fallocate *fallocate)
1043{
1044 return nfsd4_fallocate(rqstp, cstate, fallocate, 0);
1045}
1046
1047static __be32
1048nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1049 struct nfsd4_fallocate *fallocate)
1050{
1051 return nfsd4_fallocate(rqstp, cstate, fallocate,
1052 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1053}
1054
1055static __be32
1017nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, 1056nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1018 struct nfsd4_seek *seek) 1057 struct nfsd4_seek *seek)
1019{ 1058{
@@ -1331,7 +1370,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
1331 * Don't use the deferral mechanism for NFSv4; compounds make it 1370 * Don't use the deferral mechanism for NFSv4; compounds make it
1332 * too hard to avoid non-idempotency problems. 1371 * too hard to avoid non-idempotency problems.
1333 */ 1372 */
1334 rqstp->rq_usedeferral = false; 1373 clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1335 1374
1336 /* 1375 /*
1337 * According to RFC3010, this takes precedence over all other errors. 1376 * According to RFC3010, this takes precedence over all other errors.
@@ -1447,7 +1486,7 @@ encode_op:
1447 BUG_ON(cstate->replay_owner); 1486 BUG_ON(cstate->replay_owner);
1448out: 1487out:
1449 /* Reset deferral mechanism for RPC deferrals */ 1488 /* Reset deferral mechanism for RPC deferrals */
1450 rqstp->rq_usedeferral = true; 1489 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1451 dprintk("nfsv4 compound returned %d\n", ntohl(status)); 1490 dprintk("nfsv4 compound returned %d\n", ntohl(status));
1452 return status; 1491 return status;
1453} 1492}
@@ -1929,6 +1968,18 @@ static struct nfsd4_operation nfsd4_ops[] = {
1929 }, 1968 },
1930 1969
1931 /* NFSv4.2 operations */ 1970 /* NFSv4.2 operations */
1971 [OP_ALLOCATE] = {
1972 .op_func = (nfsd4op_func)nfsd4_allocate,
1973 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1974 .op_name = "OP_ALLOCATE",
1975 .op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1976 },
1977 [OP_DEALLOCATE] = {
1978 .op_func = (nfsd4op_func)nfsd4_deallocate,
1979 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1980 .op_name = "OP_DEALLOCATE",
1981 .op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1982 },
1932 [OP_SEEK] = { 1983 [OP_SEEK] = {
1933 .op_func = (nfsd4op_func)nfsd4_seek, 1984 .op_func = (nfsd4op_func)nfsd4_seek,
1934 .op_name = "OP_SEEK", 1985 .op_name = "OP_SEEK",
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 4e1d7268b004..3550a9c87616 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -275,9 +275,11 @@ opaque_hashval(const void *ptr, int nbytes)
275 return x; 275 return x;
276} 276}
277 277
278static void nfsd4_free_file(struct nfs4_file *f) 278static void nfsd4_free_file_rcu(struct rcu_head *rcu)
279{ 279{
280 kmem_cache_free(file_slab, f); 280 struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
281
282 kmem_cache_free(file_slab, fp);
281} 283}
282 284
283static inline void 285static inline void
@@ -286,9 +288,10 @@ put_nfs4_file(struct nfs4_file *fi)
286 might_lock(&state_lock); 288 might_lock(&state_lock);
287 289
288 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) { 290 if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
289 hlist_del(&fi->fi_hash); 291 hlist_del_rcu(&fi->fi_hash);
290 spin_unlock(&state_lock); 292 spin_unlock(&state_lock);
291 nfsd4_free_file(fi); 293 WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
294 call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
292 } 295 }
293} 296}
294 297
@@ -1440,7 +1443,7 @@ static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, stru
1440 list_add(&new->se_perclnt, &clp->cl_sessions); 1443 list_add(&new->se_perclnt, &clp->cl_sessions);
1441 spin_unlock(&clp->cl_lock); 1444 spin_unlock(&clp->cl_lock);
1442 1445
1443 if (cses->flags & SESSION4_BACK_CHAN) { 1446 {
1444 struct sockaddr *sa = svc_addr(rqstp); 1447 struct sockaddr *sa = svc_addr(rqstp);
1445 /* 1448 /*
1446 * This is a little silly; with sessions there's no real 1449 * This is a little silly; with sessions there's no real
@@ -1711,15 +1714,14 @@ static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1711 return 0; 1714 return 0;
1712} 1715}
1713 1716
1714static long long 1717static int
1715compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2) 1718compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1716{ 1719{
1717 long long res; 1720 if (o1->len < o2->len)
1718 1721 return -1;
1719 res = o1->len - o2->len; 1722 if (o1->len > o2->len)
1720 if (res) 1723 return 1;
1721 return res; 1724 return memcmp(o1->data, o2->data, o1->len);
1722 return (long long)memcmp(o1->data, o2->data, o1->len);
1723} 1725}
1724 1726
1725static int same_name(const char *n1, const char *n2) 1727static int same_name(const char *n1, const char *n2)
@@ -1907,7 +1909,7 @@ add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1907static struct nfs4_client * 1909static struct nfs4_client *
1908find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root) 1910find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1909{ 1911{
1910 long long cmp; 1912 int cmp;
1911 struct rb_node *node = root->rb_node; 1913 struct rb_node *node = root->rb_node;
1912 struct nfs4_client *clp; 1914 struct nfs4_client *clp;
1913 1915
@@ -3057,10 +3059,9 @@ static struct nfs4_file *nfsd4_alloc_file(void)
3057} 3059}
3058 3060
3059/* OPEN Share state helper functions */ 3061/* OPEN Share state helper functions */
3060static void nfsd4_init_file(struct nfs4_file *fp, struct knfsd_fh *fh) 3062static void nfsd4_init_file(struct knfsd_fh *fh, unsigned int hashval,
3063 struct nfs4_file *fp)
3061{ 3064{
3062 unsigned int hashval = file_hashval(fh);
3063
3064 lockdep_assert_held(&state_lock); 3065 lockdep_assert_held(&state_lock);
3065 3066
3066 atomic_set(&fp->fi_ref, 1); 3067 atomic_set(&fp->fi_ref, 1);
@@ -3073,7 +3074,7 @@ static void nfsd4_init_file(struct nfs4_file *fp, struct knfsd_fh *fh)
3073 fp->fi_share_deny = 0; 3074 fp->fi_share_deny = 0;
3074 memset(fp->fi_fds, 0, sizeof(fp->fi_fds)); 3075 memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3075 memset(fp->fi_access, 0, sizeof(fp->fi_access)); 3076 memset(fp->fi_access, 0, sizeof(fp->fi_access));
3076 hlist_add_head(&fp->fi_hash, &file_hashtbl[hashval]); 3077 hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
3077} 3078}
3078 3079
3079void 3080void
@@ -3294,17 +3295,14 @@ move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3294 3295
3295/* search file_hashtbl[] for file */ 3296/* search file_hashtbl[] for file */
3296static struct nfs4_file * 3297static struct nfs4_file *
3297find_file_locked(struct knfsd_fh *fh) 3298find_file_locked(struct knfsd_fh *fh, unsigned int hashval)
3298{ 3299{
3299 unsigned int hashval = file_hashval(fh);
3300 struct nfs4_file *fp; 3300 struct nfs4_file *fp;
3301 3301
3302 lockdep_assert_held(&state_lock); 3302 hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash) {
3303
3304 hlist_for_each_entry(fp, &file_hashtbl[hashval], fi_hash) {
3305 if (nfsd_fh_match(&fp->fi_fhandle, fh)) { 3303 if (nfsd_fh_match(&fp->fi_fhandle, fh)) {
3306 get_nfs4_file(fp); 3304 if (atomic_inc_not_zero(&fp->fi_ref))
3307 return fp; 3305 return fp;
3308 } 3306 }
3309 } 3307 }
3310 return NULL; 3308 return NULL;
@@ -3314,10 +3312,11 @@ static struct nfs4_file *
3314find_file(struct knfsd_fh *fh) 3312find_file(struct knfsd_fh *fh)
3315{ 3313{
3316 struct nfs4_file *fp; 3314 struct nfs4_file *fp;
3315 unsigned int hashval = file_hashval(fh);
3317 3316
3318 spin_lock(&state_lock); 3317 rcu_read_lock();
3319 fp = find_file_locked(fh); 3318 fp = find_file_locked(fh, hashval);
3320 spin_unlock(&state_lock); 3319 rcu_read_unlock();
3321 return fp; 3320 return fp;
3322} 3321}
3323 3322
@@ -3325,11 +3324,18 @@ static struct nfs4_file *
3325find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh) 3324find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3326{ 3325{
3327 struct nfs4_file *fp; 3326 struct nfs4_file *fp;
3327 unsigned int hashval = file_hashval(fh);
3328
3329 rcu_read_lock();
3330 fp = find_file_locked(fh, hashval);
3331 rcu_read_unlock();
3332 if (fp)
3333 return fp;
3328 3334
3329 spin_lock(&state_lock); 3335 spin_lock(&state_lock);
3330 fp = find_file_locked(fh); 3336 fp = find_file_locked(fh, hashval);
3331 if (fp == NULL) { 3337 if (likely(fp == NULL)) {
3332 nfsd4_init_file(new, fh); 3338 nfsd4_init_file(fh, hashval, new);
3333 fp = new; 3339 fp = new;
3334 } 3340 }
3335 spin_unlock(&state_lock); 3341 spin_unlock(&state_lock);
@@ -4127,7 +4133,7 @@ void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4127 nfs4_put_stateowner(so); 4133 nfs4_put_stateowner(so);
4128 } 4134 }
4129 if (open->op_file) 4135 if (open->op_file)
4130 nfsd4_free_file(open->op_file); 4136 kmem_cache_free(file_slab, open->op_file);
4131 if (open->op_stp) 4137 if (open->op_stp)
4132 nfs4_put_stid(&open->op_stp->st_stid); 4138 nfs4_put_stid(&open->op_stp->st_stid);
4133} 4139}
diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index b1eed4dd2eab..15f7b73e0c0f 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -1514,6 +1514,23 @@ static __be32 nfsd4_decode_reclaim_complete(struct nfsd4_compoundargs *argp, str
1514} 1514}
1515 1515
1516static __be32 1516static __be32
1517nfsd4_decode_fallocate(struct nfsd4_compoundargs *argp,
1518 struct nfsd4_fallocate *fallocate)
1519{
1520 DECODE_HEAD;
1521
1522 status = nfsd4_decode_stateid(argp, &fallocate->falloc_stateid);
1523 if (status)
1524 return status;
1525
1526 READ_BUF(16);
1527 p = xdr_decode_hyper(p, &fallocate->falloc_offset);
1528 xdr_decode_hyper(p, &fallocate->falloc_length);
1529
1530 DECODE_TAIL;
1531}
1532
1533static __be32
1517nfsd4_decode_seek(struct nfsd4_compoundargs *argp, struct nfsd4_seek *seek) 1534nfsd4_decode_seek(struct nfsd4_compoundargs *argp, struct nfsd4_seek *seek)
1518{ 1535{
1519 DECODE_HEAD; 1536 DECODE_HEAD;
@@ -1604,10 +1621,10 @@ static nfsd4_dec nfsd4_dec_ops[] = {
1604 [OP_RECLAIM_COMPLETE] = (nfsd4_dec)nfsd4_decode_reclaim_complete, 1621 [OP_RECLAIM_COMPLETE] = (nfsd4_dec)nfsd4_decode_reclaim_complete,
1605 1622
1606 /* new operations for NFSv4.2 */ 1623 /* new operations for NFSv4.2 */
1607 [OP_ALLOCATE] = (nfsd4_dec)nfsd4_decode_notsupp, 1624 [OP_ALLOCATE] = (nfsd4_dec)nfsd4_decode_fallocate,
1608 [OP_COPY] = (nfsd4_dec)nfsd4_decode_notsupp, 1625 [OP_COPY] = (nfsd4_dec)nfsd4_decode_notsupp,
1609 [OP_COPY_NOTIFY] = (nfsd4_dec)nfsd4_decode_notsupp, 1626 [OP_COPY_NOTIFY] = (nfsd4_dec)nfsd4_decode_notsupp,
1610 [OP_DEALLOCATE] = (nfsd4_dec)nfsd4_decode_notsupp, 1627 [OP_DEALLOCATE] = (nfsd4_dec)nfsd4_decode_fallocate,
1611 [OP_IO_ADVISE] = (nfsd4_dec)nfsd4_decode_notsupp, 1628 [OP_IO_ADVISE] = (nfsd4_dec)nfsd4_decode_notsupp,
1612 [OP_LAYOUTERROR] = (nfsd4_dec)nfsd4_decode_notsupp, 1629 [OP_LAYOUTERROR] = (nfsd4_dec)nfsd4_decode_notsupp,
1613 [OP_LAYOUTSTATS] = (nfsd4_dec)nfsd4_decode_notsupp, 1630 [OP_LAYOUTSTATS] = (nfsd4_dec)nfsd4_decode_notsupp,
@@ -1714,7 +1731,7 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
1714 argp->rqstp->rq_cachetype = cachethis ? RC_REPLBUFF : RC_NOCACHE; 1731 argp->rqstp->rq_cachetype = cachethis ? RC_REPLBUFF : RC_NOCACHE;
1715 1732
1716 if (readcount > 1 || max_reply > PAGE_SIZE - auth_slack) 1733 if (readcount > 1 || max_reply > PAGE_SIZE - auth_slack)
1717 argp->rqstp->rq_splice_ok = false; 1734 clear_bit(RQ_SPLICE_OK, &argp->rqstp->rq_flags);
1718 1735
1719 DECODE_TAIL; 1736 DECODE_TAIL;
1720} 1737}
@@ -1795,9 +1812,12 @@ static __be32 nfsd4_encode_components_esc(struct xdr_stream *xdr, char sep,
1795 } 1812 }
1796 else 1813 else
1797 end++; 1814 end++;
1815 if (found_esc)
1816 end = next;
1817
1798 str = end; 1818 str = end;
1799 } 1819 }
1800 pathlen = htonl(xdr->buf->len - pathlen_offset); 1820 pathlen = htonl(count);
1801 write_bytes_to_xdr_buf(xdr->buf, pathlen_offset, &pathlen, 4); 1821 write_bytes_to_xdr_buf(xdr->buf, pathlen_offset, &pathlen, 4);
1802 return 0; 1822 return 0;
1803} 1823}
@@ -3236,10 +3256,10 @@ nfsd4_encode_read(struct nfsd4_compoundres *resp, __be32 nfserr,
3236 3256
3237 p = xdr_reserve_space(xdr, 8); /* eof flag and byte count */ 3257 p = xdr_reserve_space(xdr, 8); /* eof flag and byte count */
3238 if (!p) { 3258 if (!p) {
3239 WARN_ON_ONCE(resp->rqstp->rq_splice_ok); 3259 WARN_ON_ONCE(test_bit(RQ_SPLICE_OK, &resp->rqstp->rq_flags));
3240 return nfserr_resource; 3260 return nfserr_resource;
3241 } 3261 }
3242 if (resp->xdr.buf->page_len && resp->rqstp->rq_splice_ok) { 3262 if (resp->xdr.buf->page_len && test_bit(RQ_SPLICE_OK, &resp->rqstp->rq_flags)) {
3243 WARN_ON_ONCE(1); 3263 WARN_ON_ONCE(1);
3244 return nfserr_resource; 3264 return nfserr_resource;
3245 } 3265 }
@@ -3256,7 +3276,7 @@ nfsd4_encode_read(struct nfsd4_compoundres *resp, __be32 nfserr,
3256 goto err_truncate; 3276 goto err_truncate;
3257 } 3277 }
3258 3278
3259 if (file->f_op->splice_read && resp->rqstp->rq_splice_ok) 3279 if (file->f_op->splice_read && test_bit(RQ_SPLICE_OK, &resp->rqstp->rq_flags))
3260 err = nfsd4_encode_splice_read(resp, read, file, maxcount); 3280 err = nfsd4_encode_splice_read(resp, read, file, maxcount);
3261 else 3281 else
3262 err = nfsd4_encode_readv(resp, read, file, maxcount); 3282 err = nfsd4_encode_readv(resp, read, file, maxcount);
diff --git a/fs/nfsd/nfscache.c b/fs/nfsd/nfscache.c
index 122f69185ef5..83a9694ec485 100644
--- a/fs/nfsd/nfscache.c
+++ b/fs/nfsd/nfscache.c
@@ -490,7 +490,7 @@ found_entry:
490 /* From the hall of fame of impractical attacks: 490 /* From the hall of fame of impractical attacks:
491 * Is this a user who tries to snoop on the cache? */ 491 * Is this a user who tries to snoop on the cache? */
492 rtn = RC_DOIT; 492 rtn = RC_DOIT;
493 if (!rqstp->rq_secure && rp->c_secure) 493 if (!test_bit(RQ_SECURE, &rqstp->rq_flags) && rp->c_secure)
494 goto out; 494 goto out;
495 495
496 /* Compose RPC reply header */ 496 /* Compose RPC reply header */
@@ -579,7 +579,7 @@ nfsd_cache_update(struct svc_rqst *rqstp, int cachetype, __be32 *statp)
579 spin_lock(&b->cache_lock); 579 spin_lock(&b->cache_lock);
580 drc_mem_usage += bufsize; 580 drc_mem_usage += bufsize;
581 lru_put_end(b, rp); 581 lru_put_end(b, rp);
582 rp->c_secure = rqstp->rq_secure; 582 rp->c_secure = test_bit(RQ_SECURE, &rqstp->rq_flags);
583 rp->c_type = cachetype; 583 rp->c_type = cachetype;
584 rp->c_state = RC_DONE; 584 rp->c_state = RC_DONE;
585 spin_unlock(&b->cache_lock); 585 spin_unlock(&b->cache_lock);
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index 9506ea565610..19ace74d35f6 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -608,7 +608,7 @@ static ssize_t __write_versions(struct file *file, char *buf, size_t size)
608 num); 608 num);
609 sep = " "; 609 sep = " ";
610 610
611 if (len > remaining) 611 if (len >= remaining)
612 break; 612 break;
613 remaining -= len; 613 remaining -= len;
614 buf += len; 614 buf += len;
@@ -623,7 +623,7 @@ static ssize_t __write_versions(struct file *file, char *buf, size_t size)
623 '+' : '-', 623 '+' : '-',
624 minor); 624 minor);
625 625
626 if (len > remaining) 626 if (len >= remaining)
627 break; 627 break;
628 remaining -= len; 628 remaining -= len;
629 buf += len; 629 buf += len;
@@ -631,7 +631,7 @@ static ssize_t __write_versions(struct file *file, char *buf, size_t size)
631 } 631 }
632 632
633 len = snprintf(buf, remaining, "\n"); 633 len = snprintf(buf, remaining, "\n");
634 if (len > remaining) 634 if (len >= remaining)
635 return -EINVAL; 635 return -EINVAL;
636 return tlen + len; 636 return tlen + len;
637} 637}
diff --git a/fs/nfsd/nfsfh.c b/fs/nfsd/nfsfh.c
index 88026fc6a981..965b478d50fc 100644
--- a/fs/nfsd/nfsfh.c
+++ b/fs/nfsd/nfsfh.c
@@ -86,7 +86,7 @@ static __be32 nfsd_setuser_and_check_port(struct svc_rqst *rqstp,
86 int flags = nfsexp_flags(rqstp, exp); 86 int flags = nfsexp_flags(rqstp, exp);
87 87
88 /* Check if the request originated from a secure port. */ 88 /* Check if the request originated from a secure port. */
89 if (!rqstp->rq_secure && !(flags & NFSEXP_INSECURE_PORT)) { 89 if (!test_bit(RQ_SECURE, &rqstp->rq_flags) && !(flags & NFSEXP_INSECURE_PORT)) {
90 RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]); 90 RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
91 dprintk("nfsd: request from insecure port %s!\n", 91 dprintk("nfsd: request from insecure port %s!\n",
92 svc_print_addr(rqstp, buf, sizeof(buf))); 92 svc_print_addr(rqstp, buf, sizeof(buf)));
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index 752d56bbe0ba..314f5c8f8f1a 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -692,7 +692,7 @@ nfsd_dispatch(struct svc_rqst *rqstp, __be32 *statp)
692 /* Now call the procedure handler, and encode NFS status. */ 692 /* Now call the procedure handler, and encode NFS status. */
693 nfserr = proc->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp); 693 nfserr = proc->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
694 nfserr = map_new_errors(rqstp->rq_vers, nfserr); 694 nfserr = map_new_errors(rqstp->rq_vers, nfserr);
695 if (nfserr == nfserr_dropit || rqstp->rq_dropme) { 695 if (nfserr == nfserr_dropit || test_bit(RQ_DROPME, &rqstp->rq_flags)) {
696 dprintk("nfsd: Dropping request; may be revisited later\n"); 696 dprintk("nfsd: Dropping request; may be revisited later\n");
697 nfsd_cache_update(rqstp, RC_NOCACHE, NULL); 697 nfsd_cache_update(rqstp, RC_NOCACHE, NULL);
698 return 0; 698 return 0;
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index 2712042a66b1..9d3be371240a 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -463,17 +463,24 @@ static inline struct nfs4_lockowner * lockowner(struct nfs4_stateowner *so)
463/* 463/*
464 * nfs4_file: a file opened by some number of (open) nfs4_stateowners. 464 * nfs4_file: a file opened by some number of (open) nfs4_stateowners.
465 * 465 *
466 * These objects are global. nfsd only keeps one instance of a nfs4_file per 466 * These objects are global. nfsd keeps one instance of a nfs4_file per
467 * inode (though it may keep multiple file descriptors open per inode). These 467 * filehandle (though it may keep multiple file descriptors for each). Each
468 * are tracked in the file_hashtbl which is protected by the state_lock 468 * inode can have multiple filehandles associated with it, so there is
469 * spinlock. 469 * (potentially) a many to one relationship between this struct and struct
470 * inode.
471 *
472 * These are hashed by filehandle in the file_hashtbl, which is protected by
473 * the global state_lock spinlock.
470 */ 474 */
471struct nfs4_file { 475struct nfs4_file {
472 atomic_t fi_ref; 476 atomic_t fi_ref;
473 spinlock_t fi_lock; 477 spinlock_t fi_lock;
474 struct hlist_node fi_hash; /* hash by "struct inode *" */ 478 struct hlist_node fi_hash; /* hash on fi_fhandle */
475 struct list_head fi_stateids; 479 struct list_head fi_stateids;
476 struct list_head fi_delegations; 480 union {
481 struct list_head fi_delegations;
482 struct rcu_head fi_rcu;
483 };
477 /* One each for O_RDONLY, O_WRONLY, O_RDWR: */ 484 /* One each for O_RDONLY, O_WRONLY, O_RDWR: */
478 struct file * fi_fds[3]; 485 struct file * fi_fds[3];
479 /* 486 /*
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 0a82e3c033ee..5685c679dd93 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -16,6 +16,7 @@
16#include <linux/fs.h> 16#include <linux/fs.h>
17#include <linux/file.h> 17#include <linux/file.h>
18#include <linux/splice.h> 18#include <linux/splice.h>
19#include <linux/falloc.h>
19#include <linux/fcntl.h> 20#include <linux/fcntl.h>
20#include <linux/namei.h> 21#include <linux/namei.h>
21#include <linux/delay.h> 22#include <linux/delay.h>
@@ -533,6 +534,26 @@ __be32 nfsd4_set_nfs4_label(struct svc_rqst *rqstp, struct svc_fh *fhp,
533} 534}
534#endif 535#endif
535 536
537__be32 nfsd4_vfs_fallocate(struct svc_rqst *rqstp, struct svc_fh *fhp,
538 struct file *file, loff_t offset, loff_t len,
539 int flags)
540{
541 __be32 err;
542 int error;
543
544 if (!S_ISREG(file_inode(file)->i_mode))
545 return nfserr_inval;
546
547 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, NFSD_MAY_WRITE);
548 if (err)
549 return err;
550
551 error = vfs_fallocate(file, flags, offset, len);
552 if (!error)
553 error = commit_metadata(fhp);
554
555 return nfserrno(error);
556}
536#endif /* defined(CONFIG_NFSD_V4) */ 557#endif /* defined(CONFIG_NFSD_V4) */
537 558
538#ifdef CONFIG_NFSD_V3 559#ifdef CONFIG_NFSD_V3
@@ -881,7 +902,7 @@ static __be32
881nfsd_vfs_read(struct svc_rqst *rqstp, struct file *file, 902nfsd_vfs_read(struct svc_rqst *rqstp, struct file *file,
882 loff_t offset, struct kvec *vec, int vlen, unsigned long *count) 903 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
883{ 904{
884 if (file->f_op->splice_read && rqstp->rq_splice_ok) 905 if (file->f_op->splice_read && test_bit(RQ_SPLICE_OK, &rqstp->rq_flags))
885 return nfsd_splice_read(rqstp, file, offset, count); 906 return nfsd_splice_read(rqstp, file, offset, count);
886 else 907 else
887 return nfsd_readv(file, offset, vec, vlen, count); 908 return nfsd_readv(file, offset, vec, vlen, count);
@@ -937,9 +958,10 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
937 int stable = *stablep; 958 int stable = *stablep;
938 int use_wgather; 959 int use_wgather;
939 loff_t pos = offset; 960 loff_t pos = offset;
961 loff_t end = LLONG_MAX;
940 unsigned int pflags = current->flags; 962 unsigned int pflags = current->flags;
941 963
942 if (rqstp->rq_local) 964 if (test_bit(RQ_LOCAL, &rqstp->rq_flags))
943 /* 965 /*
944 * We want less throttling in balance_dirty_pages() 966 * We want less throttling in balance_dirty_pages()
945 * and shrink_inactive_list() so that nfs to 967 * and shrink_inactive_list() so that nfs to
@@ -967,10 +989,13 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
967 fsnotify_modify(file); 989 fsnotify_modify(file);
968 990
969 if (stable) { 991 if (stable) {
970 if (use_wgather) 992 if (use_wgather) {
971 host_err = wait_for_concurrent_writes(file); 993 host_err = wait_for_concurrent_writes(file);
972 else 994 } else {
973 host_err = vfs_fsync_range(file, offset, offset+*cnt, 0); 995 if (*cnt)
996 end = offset + *cnt - 1;
997 host_err = vfs_fsync_range(file, offset, end, 0);
998 }
974 } 999 }
975 1000
976out_nfserr: 1001out_nfserr:
@@ -979,7 +1004,7 @@ out_nfserr:
979 err = 0; 1004 err = 0;
980 else 1005 else
981 err = nfserrno(host_err); 1006 err = nfserrno(host_err);
982 if (rqstp->rq_local) 1007 if (test_bit(RQ_LOCAL, &rqstp->rq_flags))
983 tsk_restore_flags(current, pflags, PF_LESS_THROTTLE); 1008 tsk_restore_flags(current, pflags, PF_LESS_THROTTLE);
984 return err; 1009 return err;
985} 1010}
diff --git a/fs/nfsd/vfs.h b/fs/nfsd/vfs.h
index b1796d6ee538..2050cb016998 100644
--- a/fs/nfsd/vfs.h
+++ b/fs/nfsd/vfs.h
@@ -54,6 +54,8 @@ int nfsd_mountpoint(struct dentry *, struct svc_export *);
54#ifdef CONFIG_NFSD_V4 54#ifdef CONFIG_NFSD_V4
55__be32 nfsd4_set_nfs4_label(struct svc_rqst *, struct svc_fh *, 55__be32 nfsd4_set_nfs4_label(struct svc_rqst *, struct svc_fh *,
56 struct xdr_netobj *); 56 struct xdr_netobj *);
57__be32 nfsd4_vfs_fallocate(struct svc_rqst *, struct svc_fh *,
58 struct file *, loff_t, loff_t, int);
57#endif /* CONFIG_NFSD_V4 */ 59#endif /* CONFIG_NFSD_V4 */
58__be32 nfsd_create(struct svc_rqst *, struct svc_fh *, 60__be32 nfsd_create(struct svc_rqst *, struct svc_fh *,
59 char *name, int len, struct iattr *attrs, 61 char *name, int len, struct iattr *attrs,
diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h
index 5720e9457f33..90a5925bd6ab 100644
--- a/fs/nfsd/xdr4.h
+++ b/fs/nfsd/xdr4.h
@@ -428,6 +428,13 @@ struct nfsd4_reclaim_complete {
428 u32 rca_one_fs; 428 u32 rca_one_fs;
429}; 429};
430 430
431struct nfsd4_fallocate {
432 /* request */
433 stateid_t falloc_stateid;
434 loff_t falloc_offset;
435 u64 falloc_length;
436};
437
431struct nfsd4_seek { 438struct nfsd4_seek {
432 /* request */ 439 /* request */
433 stateid_t seek_stateid; 440 stateid_t seek_stateid;
@@ -486,6 +493,8 @@ struct nfsd4_op {
486 struct nfsd4_free_stateid free_stateid; 493 struct nfsd4_free_stateid free_stateid;
487 494
488 /* NFSv4.2 */ 495 /* NFSv4.2 */
496 struct nfsd4_fallocate allocate;
497 struct nfsd4_fallocate deallocate;
489 struct nfsd4_seek seek; 498 struct nfsd4_seek seek;
490 } u; 499 } u;
491 struct nfs4_replay * replay; 500 struct nfs4_replay * replay;
diff --git a/fs/open.c b/fs/open.c
index d45bd905d418..813be037b412 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -222,7 +222,7 @@ SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
222#endif /* BITS_PER_LONG == 32 */ 222#endif /* BITS_PER_LONG == 32 */
223 223
224 224
225int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) 225int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
226{ 226{
227 struct inode *inode = file_inode(file); 227 struct inode *inode = file_inode(file);
228 long ret; 228 long ret;
@@ -309,6 +309,7 @@ int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
309 sb_end_write(inode->i_sb); 309 sb_end_write(inode->i_sb);
310 return ret; 310 return ret;
311} 311}
312EXPORT_SYMBOL_GPL(vfs_fallocate);
312 313
313SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) 314SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
314{ 315{
@@ -316,7 +317,7 @@ SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
316 int error = -EBADF; 317 int error = -EBADF;
317 318
318 if (f.file) { 319 if (f.file) {
319 error = do_fallocate(f.file, mode, offset, len); 320 error = vfs_fallocate(f.file, mode, offset, len);
320 fdput(f); 321 fdput(f);
321 } 322 }
322 return error; 323 return error;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 88157253b9e6..eeaccd37184f 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2086,7 +2086,7 @@ struct filename {
2086extern long vfs_truncate(struct path *, loff_t); 2086extern long vfs_truncate(struct path *, loff_t);
2087extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs, 2087extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs,
2088 struct file *filp); 2088 struct file *filp);
2089extern int do_fallocate(struct file *file, int mode, loff_t offset, 2089extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2090 loff_t len); 2090 loff_t len);
2091extern long do_sys_open(int dfd, const char __user *filename, int flags, 2091extern long do_sys_open(int dfd, const char __user *filename, int flags,
2092 umode_t mode); 2092 umode_t mode);
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 21678464883a..6f22cfeef5e3 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -26,10 +26,10 @@ typedef int (*svc_thread_fn)(void *);
26 26
27/* statistics for svc_pool structures */ 27/* statistics for svc_pool structures */
28struct svc_pool_stats { 28struct svc_pool_stats {
29 unsigned long packets; 29 atomic_long_t packets;
30 unsigned long sockets_queued; 30 unsigned long sockets_queued;
31 unsigned long threads_woken; 31 atomic_long_t threads_woken;
32 unsigned long threads_timedout; 32 atomic_long_t threads_timedout;
33}; 33};
34 34
35/* 35/*
@@ -45,12 +45,13 @@ struct svc_pool_stats {
45struct svc_pool { 45struct svc_pool {
46 unsigned int sp_id; /* pool id; also node id on NUMA */ 46 unsigned int sp_id; /* pool id; also node id on NUMA */
47 spinlock_t sp_lock; /* protects all fields */ 47 spinlock_t sp_lock; /* protects all fields */
48 struct list_head sp_threads; /* idle server threads */
49 struct list_head sp_sockets; /* pending sockets */ 48 struct list_head sp_sockets; /* pending sockets */
50 unsigned int sp_nrthreads; /* # of threads in pool */ 49 unsigned int sp_nrthreads; /* # of threads in pool */
51 struct list_head sp_all_threads; /* all server threads */ 50 struct list_head sp_all_threads; /* all server threads */
52 struct svc_pool_stats sp_stats; /* statistics on pool operation */ 51 struct svc_pool_stats sp_stats; /* statistics on pool operation */
53 int sp_task_pending;/* has pending task */ 52#define SP_TASK_PENDING (0) /* still work to do even if no
53 * xprt is queued. */
54 unsigned long sp_flags;
54} ____cacheline_aligned_in_smp; 55} ____cacheline_aligned_in_smp;
55 56
56/* 57/*
@@ -219,8 +220,8 @@ static inline void svc_putu32(struct kvec *iov, __be32 val)
219 * processed. 220 * processed.
220 */ 221 */
221struct svc_rqst { 222struct svc_rqst {
222 struct list_head rq_list; /* idle list */
223 struct list_head rq_all; /* all threads list */ 223 struct list_head rq_all; /* all threads list */
224 struct rcu_head rq_rcu_head; /* for RCU deferred kfree */
224 struct svc_xprt * rq_xprt; /* transport ptr */ 225 struct svc_xprt * rq_xprt; /* transport ptr */
225 226
226 struct sockaddr_storage rq_addr; /* peer address */ 227 struct sockaddr_storage rq_addr; /* peer address */
@@ -236,7 +237,6 @@ struct svc_rqst {
236 struct svc_cred rq_cred; /* auth info */ 237 struct svc_cred rq_cred; /* auth info */
237 void * rq_xprt_ctxt; /* transport specific context ptr */ 238 void * rq_xprt_ctxt; /* transport specific context ptr */
238 struct svc_deferred_req*rq_deferred; /* deferred request we are replaying */ 239 struct svc_deferred_req*rq_deferred; /* deferred request we are replaying */
239 bool rq_usedeferral; /* use deferral */
240 240
241 size_t rq_xprt_hlen; /* xprt header len */ 241 size_t rq_xprt_hlen; /* xprt header len */
242 struct xdr_buf rq_arg; 242 struct xdr_buf rq_arg;
@@ -253,9 +253,17 @@ struct svc_rqst {
253 u32 rq_vers; /* program version */ 253 u32 rq_vers; /* program version */
254 u32 rq_proc; /* procedure number */ 254 u32 rq_proc; /* procedure number */
255 u32 rq_prot; /* IP protocol */ 255 u32 rq_prot; /* IP protocol */
256 unsigned short 256 int rq_cachetype; /* catering to nfsd */
257 rq_secure : 1; /* secure port */ 257#define RQ_SECURE (0) /* secure port */
258 unsigned short rq_local : 1; /* local request */ 258#define RQ_LOCAL (1) /* local request */
259#define RQ_USEDEFERRAL (2) /* use deferral */
260#define RQ_DROPME (3) /* drop current reply */
261#define RQ_SPLICE_OK (4) /* turned off in gss privacy
262 * to prevent encrypting page
263 * cache pages */
264#define RQ_VICTIM (5) /* about to be shut down */
265#define RQ_BUSY (6) /* request is busy */
266 unsigned long rq_flags; /* flags field */
259 267
260 void * rq_argp; /* decoded arguments */ 268 void * rq_argp; /* decoded arguments */
261 void * rq_resp; /* xdr'd results */ 269 void * rq_resp; /* xdr'd results */
@@ -271,16 +279,12 @@ struct svc_rqst {
271 struct cache_req rq_chandle; /* handle passed to caches for 279 struct cache_req rq_chandle; /* handle passed to caches for
272 * request delaying 280 * request delaying
273 */ 281 */
274 bool rq_dropme;
275 /* Catering to nfsd */ 282 /* Catering to nfsd */
276 struct auth_domain * rq_client; /* RPC peer info */ 283 struct auth_domain * rq_client; /* RPC peer info */
277 struct auth_domain * rq_gssclient; /* "gss/"-style peer info */ 284 struct auth_domain * rq_gssclient; /* "gss/"-style peer info */
278 int rq_cachetype;
279 struct svc_cacherep * rq_cacherep; /* cache info */ 285 struct svc_cacherep * rq_cacherep; /* cache info */
280 bool rq_splice_ok; /* turned off in gss privacy
281 * to prevent encrypting page
282 * cache pages */
283 struct task_struct *rq_task; /* service thread */ 286 struct task_struct *rq_task; /* service thread */
287 spinlock_t rq_lock; /* per-request lock */
284}; 288};
285 289
286#define SVC_NET(svc_rqst) (svc_rqst->rq_xprt->xpt_net) 290#define SVC_NET(svc_rqst) (svc_rqst->rq_xprt->xpt_net)
diff --git a/include/linux/sunrpc/svc_xprt.h b/include/linux/sunrpc/svc_xprt.h
index ce6e4182a5b2..79f6f8f3dc0a 100644
--- a/include/linux/sunrpc/svc_xprt.h
+++ b/include/linux/sunrpc/svc_xprt.h
@@ -63,10 +63,9 @@ struct svc_xprt {
63#define XPT_CHNGBUF 7 /* need to change snd/rcv buf sizes */ 63#define XPT_CHNGBUF 7 /* need to change snd/rcv buf sizes */
64#define XPT_DEFERRED 8 /* deferred request pending */ 64#define XPT_DEFERRED 8 /* deferred request pending */
65#define XPT_OLD 9 /* used for xprt aging mark+sweep */ 65#define XPT_OLD 9 /* used for xprt aging mark+sweep */
66#define XPT_DETACHED 10 /* detached from tempsocks list */ 66#define XPT_LISTENER 10 /* listening endpoint */
67#define XPT_LISTENER 11 /* listening endpoint */ 67#define XPT_CACHE_AUTH 11 /* cache auth info */
68#define XPT_CACHE_AUTH 12 /* cache auth info */ 68#define XPT_LOCAL 12 /* connection from loopback interface */
69#define XPT_LOCAL 13 /* connection from loopback interface */
70 69
71 struct svc_serv *xpt_server; /* service for transport */ 70 struct svc_serv *xpt_server; /* service for transport */
72 atomic_t xpt_reserved; /* space on outq that is rsvd */ 71 atomic_t xpt_reserved; /* space on outq that is rsvd */
diff --git a/include/trace/events/sunrpc.h b/include/trace/events/sunrpc.h
index 171ca4ff6d99..b9c1dc6c825a 100644
--- a/include/trace/events/sunrpc.h
+++ b/include/trace/events/sunrpc.h
@@ -8,6 +8,7 @@
8#include <linux/sunrpc/clnt.h> 8#include <linux/sunrpc/clnt.h>
9#include <linux/sunrpc/svc.h> 9#include <linux/sunrpc/svc.h>
10#include <linux/sunrpc/xprtsock.h> 10#include <linux/sunrpc/xprtsock.h>
11#include <linux/sunrpc/svc_xprt.h>
11#include <net/tcp_states.h> 12#include <net/tcp_states.h>
12#include <linux/net.h> 13#include <linux/net.h>
13#include <linux/tracepoint.h> 14#include <linux/tracepoint.h>
@@ -412,6 +413,16 @@ TRACE_EVENT(xs_tcp_data_recv,
412 __entry->copied, __entry->reclen, __entry->offset) 413 __entry->copied, __entry->reclen, __entry->offset)
413); 414);
414 415
416#define show_rqstp_flags(flags) \
417 __print_flags(flags, "|", \
418 { (1UL << RQ_SECURE), "RQ_SECURE"}, \
419 { (1UL << RQ_LOCAL), "RQ_LOCAL"}, \
420 { (1UL << RQ_USEDEFERRAL), "RQ_USEDEFERRAL"}, \
421 { (1UL << RQ_DROPME), "RQ_DROPME"}, \
422 { (1UL << RQ_SPLICE_OK), "RQ_SPLICE_OK"}, \
423 { (1UL << RQ_VICTIM), "RQ_VICTIM"}, \
424 { (1UL << RQ_BUSY), "RQ_BUSY"})
425
415TRACE_EVENT(svc_recv, 426TRACE_EVENT(svc_recv,
416 TP_PROTO(struct svc_rqst *rqst, int status), 427 TP_PROTO(struct svc_rqst *rqst, int status),
417 428
@@ -421,16 +432,19 @@ TRACE_EVENT(svc_recv,
421 __field(struct sockaddr *, addr) 432 __field(struct sockaddr *, addr)
422 __field(__be32, xid) 433 __field(__be32, xid)
423 __field(int, status) 434 __field(int, status)
435 __field(unsigned long, flags)
424 ), 436 ),
425 437
426 TP_fast_assign( 438 TP_fast_assign(
427 __entry->addr = (struct sockaddr *)&rqst->rq_addr; 439 __entry->addr = (struct sockaddr *)&rqst->rq_addr;
428 __entry->xid = status > 0 ? rqst->rq_xid : 0; 440 __entry->xid = status > 0 ? rqst->rq_xid : 0;
429 __entry->status = status; 441 __entry->status = status;
442 __entry->flags = rqst->rq_flags;
430 ), 443 ),
431 444
432 TP_printk("addr=%pIScp xid=0x%x status=%d", __entry->addr, 445 TP_printk("addr=%pIScp xid=0x%x status=%d flags=%s", __entry->addr,
433 be32_to_cpu(__entry->xid), __entry->status) 446 be32_to_cpu(__entry->xid), __entry->status,
447 show_rqstp_flags(__entry->flags))
434); 448);
435 449
436DECLARE_EVENT_CLASS(svc_rqst_status, 450DECLARE_EVENT_CLASS(svc_rqst_status,
@@ -444,18 +458,19 @@ DECLARE_EVENT_CLASS(svc_rqst_status,
444 __field(__be32, xid) 458 __field(__be32, xid)
445 __field(int, dropme) 459 __field(int, dropme)
446 __field(int, status) 460 __field(int, status)
461 __field(unsigned long, flags)
447 ), 462 ),
448 463
449 TP_fast_assign( 464 TP_fast_assign(
450 __entry->addr = (struct sockaddr *)&rqst->rq_addr; 465 __entry->addr = (struct sockaddr *)&rqst->rq_addr;
451 __entry->xid = rqst->rq_xid; 466 __entry->xid = rqst->rq_xid;
452 __entry->dropme = (int)rqst->rq_dropme;
453 __entry->status = status; 467 __entry->status = status;
468 __entry->flags = rqst->rq_flags;
454 ), 469 ),
455 470
456 TP_printk("addr=%pIScp rq_xid=0x%x dropme=%d status=%d", 471 TP_printk("addr=%pIScp rq_xid=0x%x status=%d flags=%s",
457 __entry->addr, be32_to_cpu(__entry->xid), __entry->dropme, 472 __entry->addr, be32_to_cpu(__entry->xid),
458 __entry->status) 473 __entry->status, show_rqstp_flags(__entry->flags))
459); 474);
460 475
461DEFINE_EVENT(svc_rqst_status, svc_process, 476DEFINE_EVENT(svc_rqst_status, svc_process,
@@ -466,6 +481,99 @@ DEFINE_EVENT(svc_rqst_status, svc_send,
466 TP_PROTO(struct svc_rqst *rqst, int status), 481 TP_PROTO(struct svc_rqst *rqst, int status),
467 TP_ARGS(rqst, status)); 482 TP_ARGS(rqst, status));
468 483
484#define show_svc_xprt_flags(flags) \
485 __print_flags(flags, "|", \
486 { (1UL << XPT_BUSY), "XPT_BUSY"}, \
487 { (1UL << XPT_CONN), "XPT_CONN"}, \
488 { (1UL << XPT_CLOSE), "XPT_CLOSE"}, \
489 { (1UL << XPT_DATA), "XPT_DATA"}, \
490 { (1UL << XPT_TEMP), "XPT_TEMP"}, \
491 { (1UL << XPT_DEAD), "XPT_DEAD"}, \
492 { (1UL << XPT_CHNGBUF), "XPT_CHNGBUF"}, \
493 { (1UL << XPT_DEFERRED), "XPT_DEFERRED"}, \
494 { (1UL << XPT_OLD), "XPT_OLD"}, \
495 { (1UL << XPT_LISTENER), "XPT_LISTENER"}, \
496 { (1UL << XPT_CACHE_AUTH), "XPT_CACHE_AUTH"}, \
497 { (1UL << XPT_LOCAL), "XPT_LOCAL"})
498
499TRACE_EVENT(svc_xprt_do_enqueue,
500 TP_PROTO(struct svc_xprt *xprt, struct svc_rqst *rqst),
501
502 TP_ARGS(xprt, rqst),
503
504 TP_STRUCT__entry(
505 __field(struct svc_xprt *, xprt)
506 __field(struct svc_rqst *, rqst)
507 ),
508
509 TP_fast_assign(
510 __entry->xprt = xprt;
511 __entry->rqst = rqst;
512 ),
513
514 TP_printk("xprt=0x%p addr=%pIScp pid=%d flags=%s", __entry->xprt,
515 (struct sockaddr *)&__entry->xprt->xpt_remote,
516 __entry->rqst ? __entry->rqst->rq_task->pid : 0,
517 show_svc_xprt_flags(__entry->xprt->xpt_flags))
518);
519
520TRACE_EVENT(svc_xprt_dequeue,
521 TP_PROTO(struct svc_xprt *xprt),
522
523 TP_ARGS(xprt),
524
525 TP_STRUCT__entry(
526 __field(struct svc_xprt *, xprt)
527 __field_struct(struct sockaddr_storage, ss)
528 __field(unsigned long, flags)
529 ),
530
531 TP_fast_assign(
532 __entry->xprt = xprt,
533 xprt ? memcpy(&__entry->ss, &xprt->xpt_remote, sizeof(__entry->ss)) : memset(&__entry->ss, 0, sizeof(__entry->ss));
534 __entry->flags = xprt ? xprt->xpt_flags : 0;
535 ),
536
537 TP_printk("xprt=0x%p addr=%pIScp flags=%s", __entry->xprt,
538 (struct sockaddr *)&__entry->ss,
539 show_svc_xprt_flags(__entry->flags))
540);
541
542TRACE_EVENT(svc_wake_up,
543 TP_PROTO(int pid),
544
545 TP_ARGS(pid),
546
547 TP_STRUCT__entry(
548 __field(int, pid)
549 ),
550
551 TP_fast_assign(
552 __entry->pid = pid;
553 ),
554
555 TP_printk("pid=%d", __entry->pid)
556);
557
558TRACE_EVENT(svc_handle_xprt,
559 TP_PROTO(struct svc_xprt *xprt, int len),
560
561 TP_ARGS(xprt, len),
562
563 TP_STRUCT__entry(
564 __field(struct svc_xprt *, xprt)
565 __field(int, len)
566 ),
567
568 TP_fast_assign(
569 __entry->xprt = xprt;
570 __entry->len = len;
571 ),
572
573 TP_printk("xprt=0x%p addr=%pIScp len=%d flags=%s", __entry->xprt,
574 (struct sockaddr *)&__entry->xprt->xpt_remote, __entry->len,
575 show_svc_xprt_flags(__entry->xprt->xpt_flags))
576);
469#endif /* _TRACE_SUNRPC_H */ 577#endif /* _TRACE_SUNRPC_H */
470 578
471#include <trace/define_trace.h> 579#include <trace/define_trace.h>
diff --git a/mm/madvise.c b/mm/madvise.c
index 0938b30da4ab..a271adc93289 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -326,7 +326,7 @@ static long madvise_remove(struct vm_area_struct *vma,
326 */ 326 */
327 get_file(f); 327 get_file(f);
328 up_read(&current->mm->mmap_sem); 328 up_read(&current->mm->mmap_sem);
329 error = do_fallocate(f, 329 error = vfs_fallocate(f,
330 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, 330 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
331 offset, end - start); 331 offset, end - start);
332 fput(f); 332 fput(f);
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index de856ddf5fed..224a82f24d3c 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -886,7 +886,7 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
886 u32 priv_len, maj_stat; 886 u32 priv_len, maj_stat;
887 int pad, saved_len, remaining_len, offset; 887 int pad, saved_len, remaining_len, offset;
888 888
889 rqstp->rq_splice_ok = false; 889 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
890 890
891 priv_len = svc_getnl(&buf->head[0]); 891 priv_len = svc_getnl(&buf->head[0]);
892 if (rqstp->rq_deferred) { 892 if (rqstp->rq_deferred) {
diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c
index 066362141133..33fb105d4352 100644
--- a/net/sunrpc/cache.c
+++ b/net/sunrpc/cache.c
@@ -20,6 +20,7 @@
20#include <linux/list.h> 20#include <linux/list.h>
21#include <linux/module.h> 21#include <linux/module.h>
22#include <linux/ctype.h> 22#include <linux/ctype.h>
23#include <linux/string_helpers.h>
23#include <asm/uaccess.h> 24#include <asm/uaccess.h>
24#include <linux/poll.h> 25#include <linux/poll.h>
25#include <linux/seq_file.h> 26#include <linux/seq_file.h>
@@ -1067,30 +1068,15 @@ void qword_add(char **bpp, int *lp, char *str)
1067{ 1068{
1068 char *bp = *bpp; 1069 char *bp = *bpp;
1069 int len = *lp; 1070 int len = *lp;
1070 char c; 1071 int ret;
1071 1072
1072 if (len < 0) return; 1073 if (len < 0) return;
1073 1074
1074 while ((c=*str++) && len) 1075 ret = string_escape_str(str, &bp, len, ESCAPE_OCTAL, "\\ \n\t");
1075 switch(c) { 1076 if (ret < 0 || ret == len)
1076 case ' ': 1077 len = -1;
1077 case '\t':
1078 case '\n':
1079 case '\\':
1080 if (len >= 4) {
1081 *bp++ = '\\';
1082 *bp++ = '0' + ((c & 0300)>>6);
1083 *bp++ = '0' + ((c & 0070)>>3);
1084 *bp++ = '0' + ((c & 0007)>>0);
1085 }
1086 len -= 4;
1087 break;
1088 default:
1089 *bp++ = c;
1090 len--;
1091 }
1092 if (c || len <1) len = -1;
1093 else { 1078 else {
1079 len -= ret;
1094 *bp++ = ' '; 1080 *bp++ = ' ';
1095 len--; 1081 len--;
1096 } 1082 }
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 2783fd80c229..91eaef1844c8 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -191,7 +191,7 @@ svc_pool_map_init_percpu(struct svc_pool_map *m)
191 return err; 191 return err;
192 192
193 for_each_online_cpu(cpu) { 193 for_each_online_cpu(cpu) {
194 BUG_ON(pidx > maxpools); 194 BUG_ON(pidx >= maxpools);
195 m->to_pool[cpu] = pidx; 195 m->to_pool[cpu] = pidx;
196 m->pool_to[pidx] = cpu; 196 m->pool_to[pidx] = cpu;
197 pidx++; 197 pidx++;
@@ -476,15 +476,11 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
476 i, serv->sv_name); 476 i, serv->sv_name);
477 477
478 pool->sp_id = i; 478 pool->sp_id = i;
479 INIT_LIST_HEAD(&pool->sp_threads);
480 INIT_LIST_HEAD(&pool->sp_sockets); 479 INIT_LIST_HEAD(&pool->sp_sockets);
481 INIT_LIST_HEAD(&pool->sp_all_threads); 480 INIT_LIST_HEAD(&pool->sp_all_threads);
482 spin_lock_init(&pool->sp_lock); 481 spin_lock_init(&pool->sp_lock);
483 } 482 }
484 483
485 if (svc_uses_rpcbind(serv) && (!serv->sv_shutdown))
486 serv->sv_shutdown = svc_rpcb_cleanup;
487
488 return serv; 484 return serv;
489} 485}
490 486
@@ -505,13 +501,15 @@ svc_create_pooled(struct svc_program *prog, unsigned int bufsize,
505 unsigned int npools = svc_pool_map_get(); 501 unsigned int npools = svc_pool_map_get();
506 502
507 serv = __svc_create(prog, bufsize, npools, shutdown); 503 serv = __svc_create(prog, bufsize, npools, shutdown);
504 if (!serv)
505 goto out_err;
508 506
509 if (serv != NULL) { 507 serv->sv_function = func;
510 serv->sv_function = func; 508 serv->sv_module = mod;
511 serv->sv_module = mod;
512 }
513
514 return serv; 509 return serv;
510out_err:
511 svc_pool_map_put();
512 return NULL;
515} 513}
516EXPORT_SYMBOL_GPL(svc_create_pooled); 514EXPORT_SYMBOL_GPL(svc_create_pooled);
517 515
@@ -615,12 +613,14 @@ svc_prepare_thread(struct svc_serv *serv, struct svc_pool *pool, int node)
615 goto out_enomem; 613 goto out_enomem;
616 614
617 serv->sv_nrthreads++; 615 serv->sv_nrthreads++;
616 __set_bit(RQ_BUSY, &rqstp->rq_flags);
617 spin_lock_init(&rqstp->rq_lock);
618 rqstp->rq_server = serv;
619 rqstp->rq_pool = pool;
618 spin_lock_bh(&pool->sp_lock); 620 spin_lock_bh(&pool->sp_lock);
619 pool->sp_nrthreads++; 621 pool->sp_nrthreads++;
620 list_add(&rqstp->rq_all, &pool->sp_all_threads); 622 list_add_rcu(&rqstp->rq_all, &pool->sp_all_threads);
621 spin_unlock_bh(&pool->sp_lock); 623 spin_unlock_bh(&pool->sp_lock);
622 rqstp->rq_server = serv;
623 rqstp->rq_pool = pool;
624 624
625 rqstp->rq_argp = kmalloc_node(serv->sv_xdrsize, GFP_KERNEL, node); 625 rqstp->rq_argp = kmalloc_node(serv->sv_xdrsize, GFP_KERNEL, node);
626 if (!rqstp->rq_argp) 626 if (!rqstp->rq_argp)
@@ -685,7 +685,8 @@ found_pool:
685 * so we don't try to kill it again. 685 * so we don't try to kill it again.
686 */ 686 */
687 rqstp = list_entry(pool->sp_all_threads.next, struct svc_rqst, rq_all); 687 rqstp = list_entry(pool->sp_all_threads.next, struct svc_rqst, rq_all);
688 list_del_init(&rqstp->rq_all); 688 set_bit(RQ_VICTIM, &rqstp->rq_flags);
689 list_del_rcu(&rqstp->rq_all);
689 task = rqstp->rq_task; 690 task = rqstp->rq_task;
690 } 691 }
691 spin_unlock_bh(&pool->sp_lock); 692 spin_unlock_bh(&pool->sp_lock);
@@ -783,10 +784,11 @@ svc_exit_thread(struct svc_rqst *rqstp)
783 784
784 spin_lock_bh(&pool->sp_lock); 785 spin_lock_bh(&pool->sp_lock);
785 pool->sp_nrthreads--; 786 pool->sp_nrthreads--;
786 list_del(&rqstp->rq_all); 787 if (!test_and_set_bit(RQ_VICTIM, &rqstp->rq_flags))
788 list_del_rcu(&rqstp->rq_all);
787 spin_unlock_bh(&pool->sp_lock); 789 spin_unlock_bh(&pool->sp_lock);
788 790
789 kfree(rqstp); 791 kfree_rcu(rqstp, rq_rcu_head);
790 792
791 /* Release the server */ 793 /* Release the server */
792 if (serv) 794 if (serv)
@@ -1086,10 +1088,10 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
1086 goto err_short_len; 1088 goto err_short_len;
1087 1089
1088 /* Will be turned off only in gss privacy case: */ 1090 /* Will be turned off only in gss privacy case: */
1089 rqstp->rq_splice_ok = true; 1091 set_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
1090 /* Will be turned off only when NFSv4 Sessions are used */ 1092 /* Will be turned off only when NFSv4 Sessions are used */
1091 rqstp->rq_usedeferral = true; 1093 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1092 rqstp->rq_dropme = false; 1094 clear_bit(RQ_DROPME, &rqstp->rq_flags);
1093 1095
1094 /* Setup reply header */ 1096 /* Setup reply header */
1095 rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp); 1097 rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp);
@@ -1189,7 +1191,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
1189 *statp = procp->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp); 1191 *statp = procp->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
1190 1192
1191 /* Encode reply */ 1193 /* Encode reply */
1192 if (rqstp->rq_dropme) { 1194 if (test_bit(RQ_DROPME, &rqstp->rq_flags)) {
1193 if (procp->pc_release) 1195 if (procp->pc_release)
1194 procp->pc_release(rqstp, NULL, rqstp->rq_resp); 1196 procp->pc_release(rqstp, NULL, rqstp->rq_resp);
1195 goto dropit; 1197 goto dropit;
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index bbb3b044b877..c69358b3cf7f 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -220,9 +220,11 @@ static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
220 */ 220 */
221static void svc_xprt_received(struct svc_xprt *xprt) 221static void svc_xprt_received(struct svc_xprt *xprt)
222{ 222{
223 WARN_ON_ONCE(!test_bit(XPT_BUSY, &xprt->xpt_flags)); 223 if (!test_bit(XPT_BUSY, &xprt->xpt_flags)) {
224 if (!test_bit(XPT_BUSY, &xprt->xpt_flags)) 224 WARN_ONCE(1, "xprt=0x%p already busy!", xprt);
225 return; 225 return;
226 }
227
226 /* As soon as we clear busy, the xprt could be closed and 228 /* As soon as we clear busy, the xprt could be closed and
227 * 'put', so we need a reference to call svc_xprt_do_enqueue with: 229 * 'put', so we need a reference to call svc_xprt_do_enqueue with:
228 */ 230 */
@@ -310,25 +312,6 @@ char *svc_print_addr(struct svc_rqst *rqstp, char *buf, size_t len)
310} 312}
311EXPORT_SYMBOL_GPL(svc_print_addr); 313EXPORT_SYMBOL_GPL(svc_print_addr);
312 314
313/*
314 * Queue up an idle server thread. Must have pool->sp_lock held.
315 * Note: this is really a stack rather than a queue, so that we only
316 * use as many different threads as we need, and the rest don't pollute
317 * the cache.
318 */
319static void svc_thread_enqueue(struct svc_pool *pool, struct svc_rqst *rqstp)
320{
321 list_add(&rqstp->rq_list, &pool->sp_threads);
322}
323
324/*
325 * Dequeue an nfsd thread. Must have pool->sp_lock held.
326 */
327static void svc_thread_dequeue(struct svc_pool *pool, struct svc_rqst *rqstp)
328{
329 list_del(&rqstp->rq_list);
330}
331
332static bool svc_xprt_has_something_to_do(struct svc_xprt *xprt) 315static bool svc_xprt_has_something_to_do(struct svc_xprt *xprt)
333{ 316{
334 if (xprt->xpt_flags & ((1<<XPT_CONN)|(1<<XPT_CLOSE))) 317 if (xprt->xpt_flags & ((1<<XPT_CONN)|(1<<XPT_CLOSE)))
@@ -341,11 +324,12 @@ static bool svc_xprt_has_something_to_do(struct svc_xprt *xprt)
341static void svc_xprt_do_enqueue(struct svc_xprt *xprt) 324static void svc_xprt_do_enqueue(struct svc_xprt *xprt)
342{ 325{
343 struct svc_pool *pool; 326 struct svc_pool *pool;
344 struct svc_rqst *rqstp; 327 struct svc_rqst *rqstp = NULL;
345 int cpu; 328 int cpu;
329 bool queued = false;
346 330
347 if (!svc_xprt_has_something_to_do(xprt)) 331 if (!svc_xprt_has_something_to_do(xprt))
348 return; 332 goto out;
349 333
350 /* Mark transport as busy. It will remain in this state until 334 /* Mark transport as busy. It will remain in this state until
351 * the provider calls svc_xprt_received. We update XPT_BUSY 335 * the provider calls svc_xprt_received. We update XPT_BUSY
@@ -355,43 +339,69 @@ static void svc_xprt_do_enqueue(struct svc_xprt *xprt)
355 if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags)) { 339 if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags)) {
356 /* Don't enqueue transport while already enqueued */ 340 /* Don't enqueue transport while already enqueued */
357 dprintk("svc: transport %p busy, not enqueued\n", xprt); 341 dprintk("svc: transport %p busy, not enqueued\n", xprt);
358 return; 342 goto out;
359 } 343 }
360 344
361 cpu = get_cpu(); 345 cpu = get_cpu();
362 pool = svc_pool_for_cpu(xprt->xpt_server, cpu); 346 pool = svc_pool_for_cpu(xprt->xpt_server, cpu);
363 spin_lock_bh(&pool->sp_lock);
364 347
365 pool->sp_stats.packets++; 348 atomic_long_inc(&pool->sp_stats.packets);
366 349
367 if (!list_empty(&pool->sp_threads)) { 350redo_search:
368 rqstp = list_entry(pool->sp_threads.next, 351 /* find a thread for this xprt */
369 struct svc_rqst, 352 rcu_read_lock();
370 rq_list); 353 list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
371 dprintk("svc: transport %p served by daemon %p\n", 354 /* Do a lockless check first */
372 xprt, rqstp); 355 if (test_bit(RQ_BUSY, &rqstp->rq_flags))
373 svc_thread_dequeue(pool, rqstp); 356 continue;
374 if (rqstp->rq_xprt) 357
375 printk(KERN_ERR 358 /*
376 "svc_xprt_enqueue: server %p, rq_xprt=%p!\n", 359 * Once the xprt has been queued, it can only be dequeued by
377 rqstp, rqstp->rq_xprt); 360 * the task that intends to service it. All we can do at that
378 /* Note the order of the following 3 lines: 361 * point is to try to wake this thread back up so that it can
379 * We want to assign xprt to rqstp->rq_xprt only _after_ 362 * do so.
380 * we've woken up the process, so that we don't race with
381 * the lockless check in svc_get_next_xprt().
382 */ 363 */
383 svc_xprt_get(xprt); 364 if (!queued) {
365 spin_lock_bh(&rqstp->rq_lock);
366 if (test_and_set_bit(RQ_BUSY, &rqstp->rq_flags)) {
367 /* already busy, move on... */
368 spin_unlock_bh(&rqstp->rq_lock);
369 continue;
370 }
371
372 /* this one will do */
373 rqstp->rq_xprt = xprt;
374 svc_xprt_get(xprt);
375 spin_unlock_bh(&rqstp->rq_lock);
376 }
377 rcu_read_unlock();
378
379 atomic_long_inc(&pool->sp_stats.threads_woken);
384 wake_up_process(rqstp->rq_task); 380 wake_up_process(rqstp->rq_task);
385 rqstp->rq_xprt = xprt; 381 put_cpu();
386 pool->sp_stats.threads_woken++; 382 goto out;
387 } else { 383 }
384 rcu_read_unlock();
385
386 /*
387 * We didn't find an idle thread to use, so we need to queue the xprt.
388 * Do so and then search again. If we find one, we can't hook this one
389 * up to it directly but we can wake the thread up in the hopes that it
390 * will pick it up once it searches for a xprt to service.
391 */
392 if (!queued) {
393 queued = true;
388 dprintk("svc: transport %p put into queue\n", xprt); 394 dprintk("svc: transport %p put into queue\n", xprt);
395 spin_lock_bh(&pool->sp_lock);
389 list_add_tail(&xprt->xpt_ready, &pool->sp_sockets); 396 list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
390 pool->sp_stats.sockets_queued++; 397 pool->sp_stats.sockets_queued++;
398 spin_unlock_bh(&pool->sp_lock);
399 goto redo_search;
391 } 400 }
392 401 rqstp = NULL;
393 spin_unlock_bh(&pool->sp_lock);
394 put_cpu(); 402 put_cpu();
403out:
404 trace_svc_xprt_do_enqueue(xprt, rqstp);
395} 405}
396 406
397/* 407/*
@@ -408,22 +418,28 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
408EXPORT_SYMBOL_GPL(svc_xprt_enqueue); 418EXPORT_SYMBOL_GPL(svc_xprt_enqueue);
409 419
410/* 420/*
411 * Dequeue the first transport. Must be called with the pool->sp_lock held. 421 * Dequeue the first transport, if there is one.
412 */ 422 */
413static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool) 423static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
414{ 424{
415 struct svc_xprt *xprt; 425 struct svc_xprt *xprt = NULL;
416 426
417 if (list_empty(&pool->sp_sockets)) 427 if (list_empty(&pool->sp_sockets))
418 return NULL; 428 goto out;
419
420 xprt = list_entry(pool->sp_sockets.next,
421 struct svc_xprt, xpt_ready);
422 list_del_init(&xprt->xpt_ready);
423 429
424 dprintk("svc: transport %p dequeued, inuse=%d\n", 430 spin_lock_bh(&pool->sp_lock);
425 xprt, atomic_read(&xprt->xpt_ref.refcount)); 431 if (likely(!list_empty(&pool->sp_sockets))) {
432 xprt = list_first_entry(&pool->sp_sockets,
433 struct svc_xprt, xpt_ready);
434 list_del_init(&xprt->xpt_ready);
435 svc_xprt_get(xprt);
426 436
437 dprintk("svc: transport %p dequeued, inuse=%d\n",
438 xprt, atomic_read(&xprt->xpt_ref.refcount));
439 }
440 spin_unlock_bh(&pool->sp_lock);
441out:
442 trace_svc_xprt_dequeue(xprt);
427 return xprt; 443 return xprt;
428} 444}
429 445
@@ -484,34 +500,36 @@ static void svc_xprt_release(struct svc_rqst *rqstp)
484} 500}
485 501
486/* 502/*
487 * External function to wake up a server waiting for data 503 * Some svc_serv's will have occasional work to do, even when a xprt is not
488 * This really only makes sense for services like lockd 504 * waiting to be serviced. This function is there to "kick" a task in one of
489 * which have exactly one thread anyway. 505 * those services so that it can wake up and do that work. Note that we only
506 * bother with pool 0 as we don't need to wake up more than one thread for
507 * this purpose.
490 */ 508 */
491void svc_wake_up(struct svc_serv *serv) 509void svc_wake_up(struct svc_serv *serv)
492{ 510{
493 struct svc_rqst *rqstp; 511 struct svc_rqst *rqstp;
494 unsigned int i;
495 struct svc_pool *pool; 512 struct svc_pool *pool;
496 513
497 for (i = 0; i < serv->sv_nrpools; i++) { 514 pool = &serv->sv_pools[0];
498 pool = &serv->sv_pools[i];
499 515
500 spin_lock_bh(&pool->sp_lock); 516 rcu_read_lock();
501 if (!list_empty(&pool->sp_threads)) { 517 list_for_each_entry_rcu(rqstp, &pool->sp_all_threads, rq_all) {
502 rqstp = list_entry(pool->sp_threads.next, 518 /* skip any that aren't queued */
503 struct svc_rqst, 519 if (test_bit(RQ_BUSY, &rqstp->rq_flags))
504 rq_list); 520 continue;
505 dprintk("svc: daemon %p woken up.\n", rqstp); 521 rcu_read_unlock();
506 /* 522 dprintk("svc: daemon %p woken up.\n", rqstp);
507 svc_thread_dequeue(pool, rqstp); 523 wake_up_process(rqstp->rq_task);
508 rqstp->rq_xprt = NULL; 524 trace_svc_wake_up(rqstp->rq_task->pid);
509 */ 525 return;
510 wake_up_process(rqstp->rq_task);
511 } else
512 pool->sp_task_pending = 1;
513 spin_unlock_bh(&pool->sp_lock);
514 } 526 }
527 rcu_read_unlock();
528
529 /* No free entries available */
530 set_bit(SP_TASK_PENDING, &pool->sp_flags);
531 smp_wmb();
532 trace_svc_wake_up(0);
515} 533}
516EXPORT_SYMBOL_GPL(svc_wake_up); 534EXPORT_SYMBOL_GPL(svc_wake_up);
517 535
@@ -622,75 +640,86 @@ static int svc_alloc_arg(struct svc_rqst *rqstp)
622 return 0; 640 return 0;
623} 641}
624 642
643static bool
644rqst_should_sleep(struct svc_rqst *rqstp)
645{
646 struct svc_pool *pool = rqstp->rq_pool;
647
648 /* did someone call svc_wake_up? */
649 if (test_and_clear_bit(SP_TASK_PENDING, &pool->sp_flags))
650 return false;
651
652 /* was a socket queued? */
653 if (!list_empty(&pool->sp_sockets))
654 return false;
655
656 /* are we shutting down? */
657 if (signalled() || kthread_should_stop())
658 return false;
659
660 /* are we freezing? */
661 if (freezing(current))
662 return false;
663
664 return true;
665}
666
625static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp, long timeout) 667static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp, long timeout)
626{ 668{
627 struct svc_xprt *xprt; 669 struct svc_xprt *xprt;
628 struct svc_pool *pool = rqstp->rq_pool; 670 struct svc_pool *pool = rqstp->rq_pool;
629 long time_left = 0; 671 long time_left = 0;
630 672
673 /* rq_xprt should be clear on entry */
674 WARN_ON_ONCE(rqstp->rq_xprt);
675
631 /* Normally we will wait up to 5 seconds for any required 676 /* Normally we will wait up to 5 seconds for any required
632 * cache information to be provided. 677 * cache information to be provided.
633 */ 678 */
634 rqstp->rq_chandle.thread_wait = 5*HZ; 679 rqstp->rq_chandle.thread_wait = 5*HZ;
635 680
636 spin_lock_bh(&pool->sp_lock);
637 xprt = svc_xprt_dequeue(pool); 681 xprt = svc_xprt_dequeue(pool);
638 if (xprt) { 682 if (xprt) {
639 rqstp->rq_xprt = xprt; 683 rqstp->rq_xprt = xprt;
640 svc_xprt_get(xprt);
641 684
642 /* As there is a shortage of threads and this request 685 /* As there is a shortage of threads and this request
643 * had to be queued, don't allow the thread to wait so 686 * had to be queued, don't allow the thread to wait so
644 * long for cache updates. 687 * long for cache updates.
645 */ 688 */
646 rqstp->rq_chandle.thread_wait = 1*HZ; 689 rqstp->rq_chandle.thread_wait = 1*HZ;
647 pool->sp_task_pending = 0; 690 clear_bit(SP_TASK_PENDING, &pool->sp_flags);
648 } else { 691 return xprt;
649 if (pool->sp_task_pending) { 692 }
650 pool->sp_task_pending = 0;
651 xprt = ERR_PTR(-EAGAIN);
652 goto out;
653 }
654 /*
655 * We have to be able to interrupt this wait
656 * to bring down the daemons ...
657 */
658 set_current_state(TASK_INTERRUPTIBLE);
659 693
660 /* No data pending. Go to sleep */ 694 /*
661 svc_thread_enqueue(pool, rqstp); 695 * We have to be able to interrupt this wait
662 spin_unlock_bh(&pool->sp_lock); 696 * to bring down the daemons ...
697 */
698 set_current_state(TASK_INTERRUPTIBLE);
699 clear_bit(RQ_BUSY, &rqstp->rq_flags);
700 smp_mb();
663 701
664 if (!(signalled() || kthread_should_stop())) { 702 if (likely(rqst_should_sleep(rqstp)))
665 time_left = schedule_timeout(timeout); 703 time_left = schedule_timeout(timeout);
666 __set_current_state(TASK_RUNNING); 704 else
705 __set_current_state(TASK_RUNNING);
667 706
668 try_to_freeze(); 707 try_to_freeze();
669 708
670 xprt = rqstp->rq_xprt; 709 spin_lock_bh(&rqstp->rq_lock);
671 if (xprt != NULL) 710 set_bit(RQ_BUSY, &rqstp->rq_flags);
672 return xprt; 711 spin_unlock_bh(&rqstp->rq_lock);
673 } else
674 __set_current_state(TASK_RUNNING);
675 712
676 spin_lock_bh(&pool->sp_lock); 713 xprt = rqstp->rq_xprt;
677 if (!time_left) 714 if (xprt != NULL)
678 pool->sp_stats.threads_timedout++; 715 return xprt;
679 716
680 xprt = rqstp->rq_xprt; 717 if (!time_left)
681 if (!xprt) { 718 atomic_long_inc(&pool->sp_stats.threads_timedout);
682 svc_thread_dequeue(pool, rqstp); 719
683 spin_unlock_bh(&pool->sp_lock); 720 if (signalled() || kthread_should_stop())
684 dprintk("svc: server %p, no data yet\n", rqstp); 721 return ERR_PTR(-EINTR);
685 if (signalled() || kthread_should_stop()) 722 return ERR_PTR(-EAGAIN);
686 return ERR_PTR(-EINTR);
687 else
688 return ERR_PTR(-EAGAIN);
689 }
690 }
691out:
692 spin_unlock_bh(&pool->sp_lock);
693 return xprt;
694} 723}
695 724
696static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt) 725static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt)
@@ -719,7 +748,7 @@ static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
719 dprintk("svc_recv: found XPT_CLOSE\n"); 748 dprintk("svc_recv: found XPT_CLOSE\n");
720 svc_delete_xprt(xprt); 749 svc_delete_xprt(xprt);
721 /* Leave XPT_BUSY set on the dead xprt: */ 750 /* Leave XPT_BUSY set on the dead xprt: */
722 return 0; 751 goto out;
723 } 752 }
724 if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) { 753 if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
725 struct svc_xprt *newxpt; 754 struct svc_xprt *newxpt;
@@ -750,6 +779,8 @@ static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
750 } 779 }
751 /* clear XPT_BUSY: */ 780 /* clear XPT_BUSY: */
752 svc_xprt_received(xprt); 781 svc_xprt_received(xprt);
782out:
783 trace_svc_handle_xprt(xprt, len);
753 return len; 784 return len;
754} 785}
755 786
@@ -797,7 +828,10 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
797 828
798 clear_bit(XPT_OLD, &xprt->xpt_flags); 829 clear_bit(XPT_OLD, &xprt->xpt_flags);
799 830
800 rqstp->rq_secure = xprt->xpt_ops->xpo_secure_port(rqstp); 831 if (xprt->xpt_ops->xpo_secure_port(rqstp))
832 set_bit(RQ_SECURE, &rqstp->rq_flags);
833 else
834 clear_bit(RQ_SECURE, &rqstp->rq_flags);
801 rqstp->rq_chandle.defer = svc_defer; 835 rqstp->rq_chandle.defer = svc_defer;
802 rqstp->rq_xid = svc_getu32(&rqstp->rq_arg.head[0]); 836 rqstp->rq_xid = svc_getu32(&rqstp->rq_arg.head[0]);
803 837
@@ -895,7 +929,6 @@ static void svc_age_temp_xprts(unsigned long closure)
895 continue; 929 continue;
896 list_del_init(le); 930 list_del_init(le);
897 set_bit(XPT_CLOSE, &xprt->xpt_flags); 931 set_bit(XPT_CLOSE, &xprt->xpt_flags);
898 set_bit(XPT_DETACHED, &xprt->xpt_flags);
899 dprintk("queuing xprt %p for closing\n", xprt); 932 dprintk("queuing xprt %p for closing\n", xprt);
900 933
901 /* a thread will dequeue and close it soon */ 934 /* a thread will dequeue and close it soon */
@@ -935,8 +968,7 @@ static void svc_delete_xprt(struct svc_xprt *xprt)
935 xprt->xpt_ops->xpo_detach(xprt); 968 xprt->xpt_ops->xpo_detach(xprt);
936 969
937 spin_lock_bh(&serv->sv_lock); 970 spin_lock_bh(&serv->sv_lock);
938 if (!test_and_set_bit(XPT_DETACHED, &xprt->xpt_flags)) 971 list_del_init(&xprt->xpt_list);
939 list_del_init(&xprt->xpt_list);
940 WARN_ON_ONCE(!list_empty(&xprt->xpt_ready)); 972 WARN_ON_ONCE(!list_empty(&xprt->xpt_ready));
941 if (test_bit(XPT_TEMP, &xprt->xpt_flags)) 973 if (test_bit(XPT_TEMP, &xprt->xpt_flags))
942 serv->sv_tmpcnt--; 974 serv->sv_tmpcnt--;
@@ -1080,7 +1112,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req)
1080 struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle); 1112 struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle);
1081 struct svc_deferred_req *dr; 1113 struct svc_deferred_req *dr;
1082 1114
1083 if (rqstp->rq_arg.page_len || !rqstp->rq_usedeferral) 1115 if (rqstp->rq_arg.page_len || !test_bit(RQ_USEDEFERRAL, &rqstp->rq_flags))
1084 return NULL; /* if more than a page, give up FIXME */ 1116 return NULL; /* if more than a page, give up FIXME */
1085 if (rqstp->rq_deferred) { 1117 if (rqstp->rq_deferred) {
1086 dr = rqstp->rq_deferred; 1118 dr = rqstp->rq_deferred;
@@ -1109,7 +1141,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req)
1109 } 1141 }
1110 svc_xprt_get(rqstp->rq_xprt); 1142 svc_xprt_get(rqstp->rq_xprt);
1111 dr->xprt = rqstp->rq_xprt; 1143 dr->xprt = rqstp->rq_xprt;
1112 rqstp->rq_dropme = true; 1144 set_bit(RQ_DROPME, &rqstp->rq_flags);
1113 1145
1114 dr->handle.revisit = svc_revisit; 1146 dr->handle.revisit = svc_revisit;
1115 return &dr->handle; 1147 return &dr->handle;
@@ -1311,10 +1343,10 @@ static int svc_pool_stats_show(struct seq_file *m, void *p)
1311 1343
1312 seq_printf(m, "%u %lu %lu %lu %lu\n", 1344 seq_printf(m, "%u %lu %lu %lu %lu\n",
1313 pool->sp_id, 1345 pool->sp_id,
1314 pool->sp_stats.packets, 1346 (unsigned long)atomic_long_read(&pool->sp_stats.packets),
1315 pool->sp_stats.sockets_queued, 1347 pool->sp_stats.sockets_queued,
1316 pool->sp_stats.threads_woken, 1348 (unsigned long)atomic_long_read(&pool->sp_stats.threads_woken),
1317 pool->sp_stats.threads_timedout); 1349 (unsigned long)atomic_long_read(&pool->sp_stats.threads_timedout));
1318 1350
1319 return 0; 1351 return 0;
1320} 1352}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index f9c052d508f0..cc331b6cf573 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1145,7 +1145,10 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
1145 1145
1146 rqstp->rq_xprt_ctxt = NULL; 1146 rqstp->rq_xprt_ctxt = NULL;
1147 rqstp->rq_prot = IPPROTO_TCP; 1147 rqstp->rq_prot = IPPROTO_TCP;
1148 rqstp->rq_local = !!test_bit(XPT_LOCAL, &svsk->sk_xprt.xpt_flags); 1148 if (test_bit(XPT_LOCAL, &svsk->sk_xprt.xpt_flags))
1149 set_bit(RQ_LOCAL, &rqstp->rq_flags);
1150 else
1151 clear_bit(RQ_LOCAL, &rqstp->rq_flags);
1149 1152
1150 p = (__be32 *)rqstp->rq_arg.head[0].iov_base; 1153 p = (__be32 *)rqstp->rq_arg.head[0].iov_base;
1151 calldir = p[1]; 1154 calldir = p[1];
diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index 290af97bf6f9..1cb61242e55e 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -617,9 +617,10 @@ void xdr_truncate_encode(struct xdr_stream *xdr, size_t len)
617 fraglen = min_t(int, buf->len - len, tail->iov_len); 617 fraglen = min_t(int, buf->len - len, tail->iov_len);
618 tail->iov_len -= fraglen; 618 tail->iov_len -= fraglen;
619 buf->len -= fraglen; 619 buf->len -= fraglen;
620 if (tail->iov_len && buf->len == len) { 620 if (tail->iov_len) {
621 xdr->p = tail->iov_base + tail->iov_len; 621 xdr->p = tail->iov_base + tail->iov_len;
622 /* xdr->end, xdr->iov should be set already */ 622 WARN_ON_ONCE(!xdr->end);
623 WARN_ON_ONCE(!xdr->iov);
623 return; 624 return;
624 } 625 }
625 WARN_ON_ONCE(fraglen); 626 WARN_ON_ONCE(fraglen);
@@ -631,11 +632,11 @@ void xdr_truncate_encode(struct xdr_stream *xdr, size_t len)
631 old = new + fraglen; 632 old = new + fraglen;
632 xdr->page_ptr -= (old >> PAGE_SHIFT) - (new >> PAGE_SHIFT); 633 xdr->page_ptr -= (old >> PAGE_SHIFT) - (new >> PAGE_SHIFT);
633 634
634 if (buf->page_len && buf->len == len) { 635 if (buf->page_len) {
635 xdr->p = page_address(*xdr->page_ptr); 636 xdr->p = page_address(*xdr->page_ptr);
636 xdr->end = (void *)xdr->p + PAGE_SIZE; 637 xdr->end = (void *)xdr->p + PAGE_SIZE;
637 xdr->p = (void *)xdr->p + (new % PAGE_SIZE); 638 xdr->p = (void *)xdr->p + (new % PAGE_SIZE);
638 /* xdr->iov should already be NULL */ 639 WARN_ON_ONCE(xdr->iov);
639 return; 640 return;
640 } 641 }
641 if (fraglen) { 642 if (fraglen) {
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-29 07:09:35 -0500