The branch main has been updated by rmacklem:
URL:
https://cgit.FreeBSD.org/src/commit/?id=96b40b896772bbce5f93d62eaa640e1eb51b4a30
commit 96b40b896772bbce5f93d62eaa640e1eb51b4a30
Author: Rick Macklem <rmack...@freebsd.org>
AuthorDate: 2021-05-28 02:08:36 +0000
Commit: Rick Macklem <rmack...@freebsd.org>
CommitDate: 2021-05-28 02:08:36 +0000
nfscl: Use hash lists to improve expected search performance for opens
A problem was reported via email, where a large (130000+) accumulation
of NFSv4 opens on an NFSv4 mount caused significant lock contention
on the mutex used to protect the client mount's open/lock state.
Although the root cause for the accumulation of opens was not
resolved, it is obvious that the NFSv4 client is not designed to
handle 100000+ opens efficiently. When searching for an open,
usually for a match by file handle, a linear search of all opens
is done.
Commit 3f7e14ad9345 added a hash table of lists hashed on file handle
for the opens. This patch uses the hash lists for searching for
a matching open based of file handle instead of an exhaustive
linear search of all opens.
This change appears to be performance neutral for a small number
of opens, but should improve expected performance for a large
number of opens.
This commit should not affect the high level semantics of open
handling.
MFC after: 2 weeks
---
sys/fs/nfsclient/nfs_clstate.c | 165 +++++++++++++++++++----------------------
1 file changed, 75 insertions(+), 90 deletions(-)
diff --git a/sys/fs/nfsclient/nfs_clstate.c b/sys/fs/nfsclient/nfs_clstate.c
index 63c70ebcfdf5..5e4ac2ae9d4d 100644
--- a/sys/fs/nfsclient/nfs_clstate.c
+++ b/sys/fs/nfsclient/nfs_clstate.c
@@ -1229,7 +1229,6 @@ nfscl_relbytelock(vnode_t vp, u_int64_t off, u_int64_t
len,
struct nfscllockowner **lpp, int *dorpcp)
{
struct nfscllockowner *lp;
- struct nfsclowner *owp;
struct nfsclopen *op;
struct nfscllock *nlop, *other_lop = NULL;
struct nfscldeleg *dp;
@@ -1291,24 +1290,21 @@ nfscl_relbytelock(vnode_t vp, u_int64_t off, u_int64_t
len,
*/
lp = NULL;
fnd = 0;
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, np->n_fhp->nfh_fh,
+ np->n_fhp->nfh_len), nfso_hash) {
if (op->nfso_fhlen == np->n_fhp->nfh_len &&
!NFSBCMP(op->nfso_fh, np->n_fhp->nfh_fh, op->nfso_fhlen)) {
- LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
- if (lp->nfsl_inprog == NULL &&
- !NFSBCMP(lp->nfsl_owner, own,
- NFSV4CL_LOCKNAMELEN)) {
- fnd = 1;
- break;
+ LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
+ if (lp->nfsl_inprog == NULL &&
+ !NFSBCMP(lp->nfsl_owner, own,
+ NFSV4CL_LOCKNAMELEN)) {
+ fnd = 1;
+ break;
+ }
}
- }
- if (fnd)
- break;
}
- }
- if (fnd)
- break;
+ if (fnd)
+ break;
}
if (lp != NULL) {
@@ -1338,7 +1334,6 @@ void
nfscl_releasealllocks(struct nfsclclient *clp, vnode_t vp, NFSPROC_T *p,
void *id, int flags)
{
- struct nfsclowner *owp;
struct nfsclopen *op;
struct nfscllockowner *lp;
struct nfsnode *np;
@@ -1347,20 +1342,19 @@ nfscl_releasealllocks(struct nfsclclient *clp, vnode_t
vp, NFSPROC_T *p,
np = VTONFS(vp);
nfscl_filllockowner(id, own, flags);
NFSLOCKCLSTATE();
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, np->n_fhp->nfh_fh,
+ np->n_fhp->nfh_len), nfso_hash) {
if (op->nfso_fhlen == np->n_fhp->nfh_len &&
!NFSBCMP(op->nfso_fh, np->n_fhp->nfh_fh, op->nfso_fhlen)) {
- LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
- if (lp->nfsl_inprog == p &&
- !NFSBCMP(lp->nfsl_owner, own,
- NFSV4CL_LOCKNAMELEN)) {
- lp->nfsl_inprog = NULL;
- nfscl_lockunlock(&lp->nfsl_rwlock);
+ LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
+ if (lp->nfsl_inprog == p &&
+ !NFSBCMP(lp->nfsl_owner, own,
+ NFSV4CL_LOCKNAMELEN)) {
+ lp->nfsl_inprog = NULL;
+ nfscl_lockunlock(&lp->nfsl_rwlock);
+ }
}
- }
}
- }
}
nfscl_clrelease(clp);
NFSUNLOCKCLSTATE();
@@ -1376,7 +1370,6 @@ int
nfscl_checkwritelocked(vnode_t vp, struct flock *fl,
struct ucred *cred, NFSPROC_T *p, void *id, int flags)
{
- struct nfsclowner *owp;
struct nfscllockowner *lp;
struct nfsclopen *op;
struct nfsclclient *clp;
@@ -1445,30 +1438,29 @@ nfscl_checkwritelocked(vnode_t vp, struct flock *fl,
/*
* Now, check state against the server.
*/
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, np->n_fhp->nfh_fh,
+ np->n_fhp->nfh_len), nfso_hash) {
if (op->nfso_fhlen == np->n_fhp->nfh_len &&
!NFSBCMP(op->nfso_fh, np->n_fhp->nfh_fh, op->nfso_fhlen)) {
- LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
- if (!NFSBCMP(lp->nfsl_owner, own,
- NFSV4CL_LOCKNAMELEN))
- break;
- }
- if (lp != NULL) {
- LIST_FOREACH(lop, &lp->nfsl_lock, nfslo_list) {
- if (lop->nfslo_first >= end)
- break;
- if (lop->nfslo_end <= off)
- continue;
- if (lop->nfslo_type == F_WRLCK) {
- nfscl_clrelease(clp);
- NFSUNLOCKCLSTATE();
- return (1);
- }
+ LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
+ if (!NFSBCMP(lp->nfsl_owner, own,
+ NFSV4CL_LOCKNAMELEN))
+ break;
+ }
+ if (lp != NULL) {
+ LIST_FOREACH(lop, &lp->nfsl_lock, nfslo_list) {
+ if (lop->nfslo_first >= end)
+ break;
+ if (lop->nfslo_end <= off)
+ continue;
+ if (lop->nfslo_type == F_WRLCK) {
+ nfscl_clrelease(clp);
+ NFSUNLOCKCLSTATE();
+ return (1);
+ }
+ }
}
- }
}
- }
}
nfscl_clrelease(clp);
NFSUNLOCKCLSTATE();
@@ -3243,23 +3235,22 @@ nfscl_getclose(vnode_t vp, struct nfsclclient **clpp)
}
/* Now process the opens against the server. */
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
- if (op->nfso_fhlen == nfhp->nfh_len &&
- !NFSBCMP(op->nfso_fh, nfhp->nfh_fh,
- nfhp->nfh_len)) {
- /* Found an open, decrement cnt if possible */
- if (notdecr && op->nfso_opencnt > 0) {
- notdecr = 0;
- op->nfso_opencnt--;
- }
- /*
- * There are more opens, so just return.
- */
- if (op->nfso_opencnt > 0) {
- NFSUNLOCKCLSTATE();
- return (0);
- }
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, nfhp->nfh_fh, nfhp->nfh_len),
+ nfso_hash) {
+ if (op->nfso_fhlen == nfhp->nfh_len &&
+ !NFSBCMP(op->nfso_fh, nfhp->nfh_fh,
+ nfhp->nfh_len)) {
+ /* Found an open, decrement cnt if possible */
+ if (notdecr && op->nfso_opencnt > 0) {
+ notdecr = 0;
+ op->nfso_opencnt--;
+ }
+ /*
+ * There are more opens, so just return.
+ */
+ if (op->nfso_opencnt > 0) {
+ NFSUNLOCKCLSTATE();
+ return (0);
}
}
}
@@ -3310,24 +3301,21 @@ nfscl_doclose(vnode_t vp, struct nfsclclient **clpp,
NFSPROC_T *p)
/* Now process the opens against the server. */
lookformore:
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- op = LIST_FIRST(&owp->nfsow_open);
- while (op != NULL) {
- if (op->nfso_fhlen == nfhp->nfh_len &&
- !NFSBCMP(op->nfso_fh, nfhp->nfh_fh,
- nfhp->nfh_len)) {
- /* Found an open, close it. */
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, nfhp->nfh_fh, nfhp->nfh_len),
+ nfso_hash) {
+ if (op->nfso_fhlen == nfhp->nfh_len &&
+ !NFSBCMP(op->nfso_fh, nfhp->nfh_fh,
+ nfhp->nfh_len)) {
+ /* Found an open, close it. */
#ifdef DIAGNOSTIC
- KASSERT((op->nfso_opencnt == 0),
- ("nfscl: bad open cnt on server (%d)",
- op->nfso_opencnt));
+ KASSERT((op->nfso_opencnt == 0),
+ ("nfscl: bad open cnt on server (%d)",
+ op->nfso_opencnt));
#endif
- NFSUNLOCKCLSTATE();
- nfsrpc_doclose(VFSTONFS(vp->v_mount), op, p);
- NFSLOCKCLSTATE();
- goto lookformore;
- }
- op = LIST_NEXT(op, nfso_list);
+ NFSUNLOCKCLSTATE();
+ nfsrpc_doclose(VFSTONFS(vp->v_mount), op, p);
+ NFSLOCKCLSTATE();
+ goto lookformore;
}
}
NFSUNLOCKCLSTATE();
@@ -3956,7 +3944,6 @@ nfscl_localconflict(struct nfsclclient *clp, u_int8_t
*fhp, int fhlen,
struct nfscllock *nlop, u_int8_t *own, struct nfscldeleg *dp,
struct nfscllock **lopp)
{
- struct nfsclowner *owp;
struct nfsclopen *op;
int ret;
@@ -3965,15 +3952,13 @@ nfscl_localconflict(struct nfsclclient *clp, u_int8_t
*fhp, int fhlen,
if (ret)
return (ret);
}
- LIST_FOREACH(owp, &clp->nfsc_owner, nfsow_list) {
- LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
- if (op->nfso_fhlen == fhlen &&
- !NFSBCMP(op->nfso_fh, fhp, fhlen)) {
- ret = nfscl_checkconflict(&op->nfso_lock, nlop,
- own, lopp);
- if (ret)
- return (ret);
- }
+ LIST_FOREACH(op, NFSCLOPENHASH(clp, fhp, fhlen), nfso_hash) {
+ if (op->nfso_fhlen == fhlen &&
+ !NFSBCMP(op->nfso_fh, fhp, fhlen)) {
+ ret = nfscl_checkconflict(&op->nfso_lock, nlop,
+ own, lopp);
+ if (ret)
+ return (ret);
}
}
return (0);
_______________________________________________
dev-commits-src-all@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/dev-commits-src-all
To unsubscribe, send any mail to "dev-commits-src-all-unsubscr...@freebsd.org"
