glusterfs/SOURCES/0525-open-behind-implement-create-fop.patch

From 41aae052b5e3afe64d3e0668643726bab0e77265 Mon Sep 17 00:00:00 2001
From: Xavi Hernandez <xhernandez@redhat.com>
Date: Fri, 4 Sep 2020 14:49:50 +0200
Subject: [PATCH 525/526] open-behind: implement create fop

Open behind didn't implement create fop. This caused that files created
were not accounted for the number of open fd's. This could cause future
opens to be delayed when they shouldn't.

This patch implements the create fop. It also fixes a problem when
destroying the stack: when frame->local was not NULL, STACK_DESTROY()
tried to mem_put() it, which is not correct.

Upstream patch:
> Upstream-patch-link: https://review.gluster.org/#/c/glusterfs/+/24953
> Fixes: #1440
> Change-Id: Ic982bad07d4af30b915d7eb1fbcef7a847a45869
> Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>

BUG: 1830713
Change-Id: Ic982bad07d4af30b915d7eb1fbcef7a847a45869
Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/224489
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
 xlators/performance/open-behind/src/open-behind.c | 52 +++++++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/xlators/performance/open-behind/src/open-behind.c b/xlators/performance/open-behind/src/open-behind.c
index 1ab635e..600c3b6 100644
--- a/xlators/performance/open-behind/src/open-behind.c
+++ b/xlators/performance/open-behind/src/open-behind.c
@@ -336,6 +336,7 @@ ob_stub_dispatch(xlator_t *xl, ob_inode_t *ob_inode, fd_t *fd,
 static void
 ob_open_destroy(call_stub_t *stub, fd_t *fd)
 {
+    stub->frame->local = NULL;
     STACK_DESTROY(stub->frame->root);
     call_stub_destroy(stub);
     fd_unref(fd);
@@ -516,6 +517,56 @@ ob_open(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, fd_t *fd,
 }
 
 static int32_t
+ob_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+          mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+    ob_inode_t *ob_inode;
+    call_stub_t *stub;
+    fd_t *first_fd;
+    ob_state_t state;
+
+    /* Create requests are never delayed. We always send them synchronously. */
+    state = ob_open_and_resume_fd(this, fd, 1, true, true, &ob_inode,
+                                  &first_fd);
+    if (state == OB_STATE_READY) {
+        /* There's no pending open, but there are other file descriptors opened
+         * so we simply forward the request synchronously. */
+        return default_create(frame, this, loc, flags, mode, umask, fd, xdata);
+    }
+
+    if (state == OB_STATE_OPEN_TRIGGERED) {
+        /* The first open is in progress (either because it was already issued
+         * or because this request triggered it). We try to create a new stub
+         * to retry the operation once the initial open completes. */
+        stub = fop_create_stub(frame, ob_create, loc, flags, mode, umask, fd,
+                               xdata);
+        if (stub != NULL) {
+            return ob_stub_dispatch(this, ob_inode, first_fd, stub);
+        }
+
+        state = -ENOMEM;
+    }
+
+    /* Since we forced a synchronous request, OB_STATE_FIRST_OPEN will never
+     * be returned by ob_open_and_resume_fd(). If we are here it can only be
+     * because there has been a problem. */
+
+    /* In case of failure we need to decrement the number of open files because
+     * ob_fdclose() won't be called. */
+
+    LOCK(&fd->inode->lock);
+    {
+        ob_inode->open_count--;
+    }
+    UNLOCK(&fd->inode->lock);
+
+    gf_smsg(this->name, GF_LOG_ERROR, -state, OPEN_BEHIND_MSG_FAILED, "fop=%s",
+            "create", "path=%s", loc->path, NULL);
+
+    return default_create_failure_cbk(frame, -state);
+}
+
+static int32_t
 ob_readv(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
          off_t offset, uint32_t flags, dict_t *xdata)
 {
@@ -946,6 +997,7 @@ fini(xlator_t *this)
 
 struct xlator_fops fops = {
     .open = ob_open,
+    .create = ob_create,
     .readv = ob_readv,
     .writev = ob_writev,
     .flush = ob_flush,
-- 
1.8.3.1