* [PATCH 1/3] SUNRPC: Replace division by multiplication in calculation of queue length
@ 2019-07-16 20:04 Trond Myklebust
2019-07-16 20:04 ` [PATCH 2/3] SUNRPC: Skip zero-refcount transports Trond Myklebust
0 siblings, 1 reply; 3+ messages in thread
From: Trond Myklebust @ 2019-07-16 20:04 UTC (permalink / raw)
To: linux-nfs
When checking whether or not a particular xprt queue length is shorter
than the average queue length for all xprts, prefer to use multiplication
rather than division for performance reasons.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
net/sunrpc/xprtmultipath.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c
index 9d66ce53355d..5df4e7adedf0 100644
--- a/net/sunrpc/xprtmultipath.c
+++ b/net/sunrpc/xprtmultipath.c
@@ -322,7 +322,6 @@ struct rpc_xprt *xprt_iter_next_entry_roundrobin(struct rpc_xprt_iter *xpi)
struct rpc_xprt *xprt;
unsigned long xprt_queuelen;
unsigned long xps_queuelen;
- unsigned long xps_avglen;
do {
xprt = xprt_iter_next_entry_multiple(xpi,
@@ -333,8 +332,8 @@ struct rpc_xprt *xprt_iter_next_entry_roundrobin(struct rpc_xprt_iter *xpi)
if (xprt_queuelen <= 2)
break;
xps_queuelen = atomic_long_read(&xps->xps_queuelen);
- xps_avglen = DIV_ROUND_UP(xps_queuelen, xps->xps_nactive);
- } while (xprt_queuelen > xps_avglen);
+ /* Exit loop if xprt_queuelen <= average queue length */
+ } while (xprt_queuelen * READ_ONCE(xps->xps_nactive) > xps_queuelen);
return xprt;
}
--
2.21.0
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH 2/3] SUNRPC: Skip zero-refcount transports
2019-07-16 20:04 [PATCH 1/3] SUNRPC: Replace division by multiplication in calculation of queue length Trond Myklebust
@ 2019-07-16 20:04 ` Trond Myklebust
2019-07-16 20:04 ` [PATCH 3/3] SUNRPC: Fix a hang in xprt_switch_set_next_cursor() Trond Myklebust
0 siblings, 1 reply; 3+ messages in thread
From: Trond Myklebust @ 2019-07-16 20:04 UTC (permalink / raw)
To: linux-nfs
When looking for the next transport to use for an RPC call, skip those
that are in the process of being destroyed and that have a zero refcount.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
net/sunrpc/xprtmultipath.c | 20 +++++++++++++++++++-
1 file changed, 19 insertions(+), 1 deletion(-)
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c
index 5df4e7adedf0..c12778e1235e 100644
--- a/net/sunrpc/xprtmultipath.c
+++ b/net/sunrpc/xprtmultipath.c
@@ -193,10 +193,22 @@ void xprt_iter_default_rewind(struct rpc_xprt_iter *xpi)
WRITE_ONCE(xpi->xpi_cursor, NULL);
}
+static
+bool xprt_is_active(const struct rpc_xprt *xprt)
+{
+ return kref_read(&xprt->kref) != 0;
+}
+
static
struct rpc_xprt *xprt_switch_find_first_entry(struct list_head *head)
{
- return list_first_or_null_rcu(head, struct rpc_xprt, xprt_switch);
+ struct rpc_xprt *pos;
+
+ list_for_each_entry_rcu(pos, head, xprt_switch) {
+ if (xprt_is_active(pos))
+ return pos;
+ }
+ return NULL;
}
static
@@ -214,9 +226,12 @@ struct rpc_xprt *xprt_switch_find_current_entry(struct list_head *head,
const struct rpc_xprt *cur)
{
struct rpc_xprt *pos;
+ bool found = false;
list_for_each_entry_rcu(pos, head, xprt_switch) {
if (cur == pos)
+ found = true;
+ if (found && xprt_is_active(pos))
return pos;
}
return NULL;
@@ -261,9 +276,12 @@ struct rpc_xprt *xprt_switch_find_next_entry(struct list_head *head,
const struct rpc_xprt *cur)
{
struct rpc_xprt *pos, *prev = NULL;
+ bool found = false;
list_for_each_entry_rcu(pos, head, xprt_switch) {
if (cur == prev)
+ found = true;
+ if (found && xprt_is_active(pos))
return pos;
prev = pos;
}
--
2.21.0
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH 3/3] SUNRPC: Fix a hang in xprt_switch_set_next_cursor()
2019-07-16 20:04 ` [PATCH 2/3] SUNRPC: Skip zero-refcount transports Trond Myklebust
@ 2019-07-16 20:04 ` Trond Myklebust
0 siblings, 0 replies; 3+ messages in thread
From: Trond Myklebust @ 2019-07-16 20:04 UTC (permalink / raw)
To: linux-nfs
Apparently we do need the memory barrier.
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
net/sunrpc/xprtmultipath.c | 12 ++++--------
1 file changed, 4 insertions(+), 8 deletions(-)
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c
index c12778e1235e..852da66cdaec 100644
--- a/net/sunrpc/xprtmultipath.c
+++ b/net/sunrpc/xprtmultipath.c
@@ -293,18 +293,14 @@ struct rpc_xprt *xprt_switch_set_next_cursor(struct list_head *head,
struct rpc_xprt **cursor,
xprt_switch_find_xprt_t find_next)
{
- struct rpc_xprt *cur, *pos, *old;
+ struct rpc_xprt *pos, *old;
- cur = READ_ONCE(*cursor);
- for (;;) {
- old = cur;
+ old = READ_ONCE(*cursor);
+ do {
pos = find_next(head, old);
if (pos == NULL)
break;
- cur = cmpxchg_relaxed(cursor, old, pos);
- if (cur == old)
- break;
- }
+ } while (!try_cmpxchg(cursor, &old, pos));
return pos;
}
--
2.21.0
^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-07-16 20:06 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-16 20:04 [PATCH 1/3] SUNRPC: Replace division by multiplication in calculation of queue length Trond Myklebust
2019-07-16 20:04 ` [PATCH 2/3] SUNRPC: Skip zero-refcount transports Trond Myklebust
2019-07-16 20:04 ` [PATCH 3/3] SUNRPC: Fix a hang in xprt_switch_set_next_cursor() Trond Myklebust
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).