From: James Simmons <jsimmons@infradead.org>
To: Andreas Dilger <adilger@whamcloud.com>,
Oleg Drokin <green@whamcloud.com>, NeilBrown <neilb@suse.de>
Cc: Lustre Development List <lustre-devel@lists.lustre.org>
Subject: [lustre-devel] [PATCH 08/24] lnet: socklnd: factor out key calculation for ksnd_peers
Date: Tue, 21 Sep 2021 22:19:45 -0400 [thread overview]
Message-ID: <1632277201-6920-9-git-send-email-jsimmons@infradead.org> (raw)
In-Reply-To: <1632277201-6920-1-git-send-email-jsimmons@infradead.org>
From: Mr NeilBrown <neilb@suse.de>
The hash_table library requires a "long" to be used as a key. We
currently provide the nid, which at 64bits is a suitable long on 64bit
hosts, but isn't really correct on 32bit hosts.
When we change to an extend nid (which is 160bits) it will be even
less appropriate.
So create a separate function to compute a 'long' key, and implement
by simply xoring 'long'-sized parts of the nid together. On a 64bit
machine, this is currently optimized away for lnet_nid_t, but that
will change when we convert to struct lnet_nid.
This new function is placed in lnet-types.h as it will be more
generally useful later.
The hash_table library calls hash_long() on the key, so we don't need
to do anything more interesting than xoring.
WC-bug-id: https://jira.whamcloud.com/browse/LU-10391
Lustre-commit: 96a0c378c2e0a0c8f ("LU-10391 socklnd: factor out key calculation for ksnd_peers")
Signed-off-by: Mr NeilBrown <neilb@suse.de>
Reviewed-on: https://review.whamcloud.com/42103
Reviewed-by: James Simmons <jsimmons@infradead.org>
Reviewed-by: Chris Horn <chris.horn@hpe.com>
Reviewed-by: Serguei Smirnov <ssmirnov@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Signed-off-by: James Simmons <jsimmons@infradead.org>
---
include/uapi/linux/lnet/lnet-types.h | 10 ++++++++++
net/lnet/klnds/socklnd/socklnd.c | 17 +++++++++++------
2 files changed, 21 insertions(+), 6 deletions(-)
diff --git a/include/uapi/linux/lnet/lnet-types.h b/include/uapi/linux/lnet/lnet-types.h
index ba8a079..5538d4e 100644
--- a/include/uapi/linux/lnet/lnet-types.h
+++ b/include/uapi/linux/lnet/lnet-types.h
@@ -163,6 +163,16 @@ static inline int nid_same(const struct lnet_nid *n1,
n1->nid_addr[3] == n2->nid_addr[3];
}
+/* This can be used when we need to hash a nid */
+static inline unsigned long nidhash(lnet_nid_t nid)
+{
+ unsigned long hash = 0;
+
+ hash ^= LNET_NIDNET(nid);
+ hash ^= LNET_NIDADDR(nid);
+ return hash;
+}
+
struct lnet_counters_health {
__u32 lch_rst_alloc;
__u32 lch_resend_count;
diff --git a/net/lnet/klnds/socklnd/socklnd.c b/net/lnet/klnds/socklnd/socklnd.c
index 21569fb..08d1cf4 100644
--- a/net/lnet/klnds/socklnd/socklnd.c
+++ b/net/lnet/klnds/socklnd/socklnd.c
@@ -221,9 +221,10 @@ struct ksock_peer_ni *
ksocknal_find_peer_locked(struct lnet_ni *ni, struct lnet_process_id id)
{
struct ksock_peer_ni *peer_ni;
+ unsigned long hash = nidhash(id.nid);
hash_for_each_possible(ksocknal_data.ksnd_peers, peer_ni,
- ksnp_list, id.nid) {
+ ksnp_list, hash) {
LASSERT(!peer_ni->ksnp_closing);
if (peer_ni->ksnp_ni != ni)
@@ -602,7 +603,8 @@ struct ksock_peer_ni *
peer_ni = peer2;
} else {
/* peer_ni table takes my ref on peer_ni */
- hash_add(ksocknal_data.ksnd_peers, &peer_ni->ksnp_list, id.nid);
+ hash_add(ksocknal_data.ksnd_peers, &peer_ni->ksnp_list,
+ nidhash(id.nid));
}
ksocknal_add_conn_cb_locked(peer_ni, conn_cb);
@@ -656,7 +658,8 @@ struct ksock_peer_ni *
write_lock_bh(&ksocknal_data.ksnd_global_lock);
if (id.nid != LNET_NID_ANY) {
- lo = hash_min(id.nid, HASH_BITS(ksocknal_data.ksnd_peers));
+ lo = hash_min(nidhash(id.nid),
+ HASH_BITS(ksocknal_data.ksnd_peers));
hi = lo;
} else {
lo = 0;
@@ -935,7 +938,7 @@ struct ksock_peer_ni *
* table (which takes my ref)
*/
hash_add(ksocknal_data.ksnd_peers,
- &peer_ni->ksnp_list, peerid.nid);
+ &peer_ni->ksnp_list, nidhash(peerid.nid));
} else {
ksocknal_peer_decref(peer_ni);
peer_ni = peer2;
@@ -1567,7 +1570,8 @@ struct ksock_peer_ni *
write_lock_bh(&ksocknal_data.ksnd_global_lock);
if (id.nid != LNET_NID_ANY) {
- lo = hash_min(id.nid, HASH_BITS(ksocknal_data.ksnd_peers));
+ lo = hash_min(nidhash(id.nid),
+ HASH_BITS(ksocknal_data.ksnd_peers));
hi = lo;
} else {
lo = 0;
@@ -1662,7 +1666,8 @@ static int ksocknal_push(struct lnet_ni *ni, struct lnet_process_id id)
int rc = -ENOENT;
if (id.nid != LNET_NID_ANY) {
- lo = hash_min(id.nid, HASH_BITS(ksocknal_data.ksnd_peers));
+ lo = hash_min(nidhash(id.nid),
+ HASH_BITS(ksocknal_data.ksnd_peers));
hi = lo;
} else {
lo = 0;
--
1.8.3.1
_______________________________________________
lustre-devel mailing list
lustre-devel@lists.lustre.org
http://lists.lustre.org/listinfo.cgi/lustre-devel-lustre.org
next prev parent reply other threads:[~2021-09-22 2:20 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-22 2:19 [lustre-devel] [PATCH 00/24] lustre: Update to OpenSFS Sept 21, 2021 James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 01/24] lnet: Lock primary NID logic James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 02/24] lustre: quota: enforce block quota for chgrp James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 03/24] lnet: introduce struct lnet_nid James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 04/24] lnet: add string formating/parsing for IPv6 nids James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 05/24] lnet: change lpni_nid in lnet_peer_ni to lnet_nid James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 06/24] lnet: change lp_primary_nid to struct lnet_nid James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 07/24] lnet: change lp_disc_*_nid " James Simmons
2021-09-22 2:19 ` James Simmons [this message]
2021-09-22 2:19 ` [lustre-devel] [PATCH 09/24] lnet: introduce lnet_processid for ksock_peer_ni James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 10/24] lnet: enhance connect/accept to support large addr James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 11/24] lnet: change lr_nid to struct lnet_nid James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 12/24] lnet: extend rspt_next_hop_nid in lnet_rsp_tracker James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 13/24] lustre: ptlrpc: two replay lock threads James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 14/24] lustre: llite: Always do lookup on ENOENT in open James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 15/24] lustre: llite: Remove inode locking in ll_fsync James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 16/24] lnet: socklnd: fix link state detection James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 17/24] lustre: llite: check read only mount for setquota James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 18/24] lustre: llite: don't touch vma after filemap_fault James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 19/24] lnet: Check for -ESHUTDOWN in lnet_parse James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 20/24] lustre: obdclass: EAGAIN after rhashtable_walk_next() James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 21/24] lustre: sec: filename encryption James Simmons
2021-09-22 2:19 ` [lustre-devel] [PATCH 22/24] lustre: uapi: fixup UAPI headers for native Linux client James Simmons
2021-09-22 2:20 ` [lustre-devel] [PATCH 23/24] lustre: ptlrpc: separate out server code for wiretest James Simmons
2021-09-22 2:20 ` [lustre-devel] [PATCH 24/24] lustre: pcc: VM_WRITE should not trigger layout write James Simmons
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1632277201-6920-9-git-send-email-jsimmons@infradead.org \
--to=jsimmons@infradead.org \
--cc=adilger@whamcloud.com \
--cc=green@whamcloud.com \
--cc=lustre-devel@lists.lustre.org \
--cc=neilb@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).