glusterfs/SOURCES/0199-upcall-Avoid-sending-notifications-for-invalid-inode.patch
2021-09-09 17:41:22 +00:00

81 lines
2.9 KiB
Diff

From e41b4a45f9f5c07ffa38582d0bb4517f6a66eaa3 Mon Sep 17 00:00:00 2001
From: Soumya Koduri <skoduri@redhat.com>
Date: Fri, 7 Jun 2019 19:33:07 +0530
Subject: [PATCH 199/221] upcall: Avoid sending notifications for invalid
inodes
For nameless LOOKUPs, server creates a new inode which shall
remain invalid until the fop is successfully processed post
which it is linked to the inode table.
But incase if there is an already linked inode for that entry,
it discards that newly created inode which results in upcall
notification. This may result in client being bombarded with
unnecessary upcalls affecting performance if the data set is huge.
This issue can be avoided by looking up and storing the upcall
context in the original linked inode (if exists), thus saving up on
those extra callbacks.
This is backport of below upstream fix -
mainline: https://review.gluster.org/22840
release-6: https://review.gluster.org/22873
Change-Id: I044a1737819bb40d1a049d2f53c0566e746d2a17
fixes: bz#1717784
Signed-off-by: Soumya Koduri <skoduri@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/173507
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Kaleb Keithley <kkeithle@redhat.com>
---
xlators/features/upcall/src/upcall-internal.c | 19 ++++++++++++++++++-
1 file changed, 18 insertions(+), 1 deletion(-)
diff --git a/xlators/features/upcall/src/upcall-internal.c b/xlators/features/upcall/src/upcall-internal.c
index 46cf6f8..7998dd2 100644
--- a/xlators/features/upcall/src/upcall-internal.c
+++ b/xlators/features/upcall/src/upcall-internal.c
@@ -520,6 +520,7 @@ upcall_cache_invalidate(call_frame_t *frame, xlator_t *this, client_t *client,
upcall_client_t *tmp = NULL;
upcall_inode_ctx_t *up_inode_ctx = NULL;
gf_boolean_t found = _gf_false;
+ inode_t *linked_inode = NULL;
if (!is_upcall_enabled(this))
return;
@@ -532,7 +533,20 @@ upcall_cache_invalidate(call_frame_t *frame, xlator_t *this, client_t *client,
return;
}
- if (inode)
+ /* For nameless LOOKUPs, inode created shall always be
+ * invalid. Hence check if there is any already linked inode.
+ * If yes, update the inode_ctx of that valid inode
+ */
+ if (inode && (inode->ia_type == IA_INVAL) && stbuf) {
+ linked_inode = inode_find(inode->table, stbuf->ia_gfid);
+ if (linked_inode) {
+ gf_log("upcall", GF_LOG_DEBUG,
+ "upcall_inode_ctx_get of linked inode (%p)", inode);
+ up_inode_ctx = upcall_inode_ctx_get(linked_inode, this);
+ }
+ }
+
+ if (inode && !up_inode_ctx)
up_inode_ctx = upcall_inode_ctx_get(inode, this);
if (!up_inode_ctx) {
@@ -600,6 +614,9 @@ upcall_cache_invalidate(call_frame_t *frame, xlator_t *this, client_t *client,
}
pthread_mutex_unlock(&up_inode_ctx->client_list_lock);
out:
+ /* release the ref from inode_find */
+ if (linked_inode)
+ inode_unref(linked_inode);
return;
}
--
1.8.3.1