summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJeff Darcy <jdarcy@redhat.com>2013-02-19 21:14:23 -0800
committerAnand Avati <avati@redhat.com>2013-03-06 23:30:14 -0800
commitc37546cf11555678be6fefdbfec0007272aeb336 (patch)
treed8f67450ebf7e6270fb6a9bf7ee0af38723f5db3
parent6a9cbcb5b54d32db36ac26d006eae8d08071fff6 (diff)
performance/write-behind: guarantee non-overlapping concurrent writesv3.4.0alpha2
Maintain a list of writes (either written behind or SYNC) which are currently "in progress" (i.e, STACK_WIND'ed towards server) and hold off any new STACK_WIND of write (either written behind or SYNC) which overlaps with any of the "in progress" writes. This is a guarantee which AFR's eager-lock depends upon (though not strictly a write-behind requirement) Change-Id: Icedd0b51b440366a906dc9223d62b7fd6ef2ca03 BUG: 857673 Original-author: Anand Avati <avati@redhat.com> Signed-off-by: Anand Avati <avati@redhat.com> Signed-off-by: Jeff Darcy <jdarcy@redhat.com> Reviewed-on: http://review.gluster.org/4642 Tested-by: Gluster Build System <jenkins@build.gluster.com>
-rw-r--r--xlators/performance/write-behind/src/write-behind.c66
1 files changed, 65 insertions, 1 deletions
diff --git a/xlators/performance/write-behind/src/write-behind.c b/xlators/performance/write-behind/src/write-behind.c
index b0d984ea2ef..b6757aebda1 100644
--- a/xlators/performance/write-behind/src/write-behind.c
+++ b/xlators/performance/write-behind/src/write-behind.c
@@ -82,6 +82,12 @@ typedef struct wb_inode {
write-behind from this list, and therefore
get "upgraded" to the "liability" list.
*/
+ list_head_t wip; /* List of write calls in progress, SYNC or non-SYNC
+ which are currently STACK_WIND'ed towards the server.
+ This is for guaranteeing that no two overlapping
+ writes are in progress at the same time. Modules
+ like eager-lock in AFR depend on this behavior.
+ */
uint64_t gen; /* Liability generation number. Represents
the current 'state' of liability. Every
new addition to the liability list bumps
@@ -113,6 +119,7 @@ typedef struct wb_request {
list_head_t lie; /* either in @liability or @temptation */
list_head_t winds;
list_head_t unwinds;
+ list_head_t wip;
call_stub_t *stub;
@@ -319,6 +326,30 @@ wb_liability_has_conflict (wb_inode_t *wb_inode, wb_request_t *req)
}
+gf_boolean_t
+wb_wip_has_conflict (wb_inode_t *wb_inode, wb_request_t *req)
+{
+ wb_request_t *each = NULL;
+
+ if (req->stub->fop != GF_FOP_WRITE)
+ /* non-writes fundamentally never conflict with WIP requests */
+ return _gf_false;
+
+ list_for_each_entry (each, &wb_inode->wip, wip) {
+ if (each == req)
+ /* request never conflicts with itself,
+ though this condition should never occur.
+ */
+ continue;
+
+ if (wb_requests_overlap (each, req))
+ return _gf_true;
+ }
+
+ return _gf_false;
+}
+
+
static int
__wb_request_unref (wb_request_t *req)
{
@@ -337,6 +368,7 @@ __wb_request_unref (wb_request_t *req)
if (req->refcount == 0) {
list_del_init (&req->todo);
list_del_init (&req->lie);
+ list_del_init (&req->wip);
list_del_init (&req->all);
if (list_empty (&wb_inode->all)) {
@@ -442,6 +474,7 @@ wb_enqueue_common (wb_inode_t *wb_inode, call_stub_t *stub, int tempted)
INIT_LIST_HEAD (&req->lie);
INIT_LIST_HEAD (&req->winds);
INIT_LIST_HEAD (&req->unwinds);
+ INIT_LIST_HEAD (&req->wip);
req->stub = stub;
req->wb_inode = wb_inode;
@@ -558,6 +591,7 @@ __wb_inode_create (xlator_t *this, inode_t *inode)
INIT_LIST_HEAD (&wb_inode->todo);
INIT_LIST_HEAD (&wb_inode->liability);
INIT_LIST_HEAD (&wb_inode->temptation);
+ INIT_LIST_HEAD (&wb_inode->wip);
wb_inode->this = this;
@@ -1084,6 +1118,18 @@ __wb_pick_winds (wb_inode_t *wb_inode, list_head_t *tasks,
/* wait some more */
continue;
+ if (req->stub->fop == GF_FOP_WRITE) {
+ if (wb_wip_has_conflict (wb_inode, req))
+ continue;
+
+ list_add_tail (&req->wip, &wb_inode->wip);
+
+ if (!req->ordering.tempted)
+ /* unrefed in wb_writev_cbk */
+ req->stub->frame->local =
+ __wb_request_ref (req);
+ }
+
list_del_init (&req->todo);
if (req->ordering.tempted)
@@ -1143,11 +1189,29 @@ wb_process_queue (wb_inode_t *wb_inode)
int
+wb_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+ int32_t op_ret, int32_t op_errno,
+ struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+ wb_request_t *req = NULL;
+
+ req = frame->local;
+ frame->local = NULL;
+
+ wb_request_unref (req);
+
+ STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+ xdata);
+ return 0;
+}
+
+
+int
wb_writev_helper (call_frame_t *frame, xlator_t *this, fd_t *fd,
struct iovec *vector, int32_t count, off_t offset,
uint32_t flags, struct iobref *iobref, dict_t *xdata)
{
- STACK_WIND (frame, default_writev_cbk,
+ STACK_WIND (frame, wb_writev_cbk,
FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
fd, vector, count, offset, flags, iobref, xdata);
return 0;