glusterfs: discard (hole punch) support

Add support for the DISCARD file operation. Discard punches a hole
in a file in the provided range. Block de-allocation is implemented
via fallocate() (as requested via fuse and passed on to the brick
fs) but a separate fop is created within gluster to emphasize the
fact that discard changes file data (the discarded region is
replaced with zeroes) and must invalidate caches where appropriate.

BUG: 963678
Change-Id: I34633a0bfff2187afeab4292a15f3cc9adf261af
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-on: http://review.gluster.org/5090
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

3 files changed, 258 insertions, 0 deletions

diff --git a/xlators/protocol/client/src/client-rpc-fops.c b/xlators/protocol/client/src/client-rpc-fops.c
index 53cf3968a..c05542834 100644
--- a/xlators/protocol/client/src/client-rpc-fops.c
+++ b/xlators/protocol/client/src/client-rpc-fops.c
@@ -1993,6 +1993,61 @@ out:
 }
 
 int
+client3_3_discard_cbk(struct rpc_req *req, struct iovec *iov, int count,
+		      void *myframe)
+{
+        call_frame_t    *frame      = NULL;
+        gfs3_discard_rsp rsp      = {0,};
+        struct iatt      prestat    = {0,};
+        struct iatt      poststat   = {0,};
+        int              ret        = 0;
+        xlator_t *this       = NULL;
+        dict_t  *xdata       = NULL;
+
+        this = THIS;
+
+        frame = myframe;
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = ENOTCONN;
+                goto out;
+        }
+        ret = xdr_to_generic(*iov, &rsp, (xdrproc_t) xdr_gfs3_discard_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "XDR decoding failed");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        if (-1 != rsp.op_ret) {
+                gf_stat_to_iatt (&rsp.statpre, &prestat);
+                gf_stat_to_iatt (&rsp.statpost, &poststat);
+        }
+
+        GF_PROTOCOL_DICT_UNSERIALIZE (this, xdata, (rsp.xdata.xdata_val),
+                                      (rsp.xdata.xdata_len), ret,
+                                      rsp.op_errno, out);
+
+out:
+        if (rsp.op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING, "remote operation failed: %s",
+                        strerror (gf_error_to_errno (rsp.op_errno)));
+        }
+        CLIENT_STACK_UNWIND (discard, frame, rsp.op_ret,
+                             gf_error_to_errno (rsp.op_errno), &prestat,
+                             &poststat, xdata);
+
+        free (rsp.xdata.xdata_val);
+
+        if (xdata)
+                dict_unref (xdata);
+
+        return 0;
+}
+
+int
 client3_3_setattr_cbk (struct rpc_req *req, struct iovec *iov, int count,
                        void *myframe)
 {
@@ -5888,6 +5943,50 @@ unwind:
         return 0;
 }
 
+int32_t
+client3_3_discard(call_frame_t *frame, xlator_t *this, void *data)
+{
+        clnt_args_t       *args     = NULL;
+        int64_t            remote_fd = -1;
+        clnt_conf_t       *conf     = NULL;
+	gfs3_discard_req   req	    = {{0},};
+        int                op_errno = ESTALE;
+        int                ret        = 0;
+
+        if (!frame || !this || !data)
+                goto unwind;
+
+        args = data;
+        conf = this->private;
+
+        CLIENT_GET_REMOTE_FD (this, args->fd, DEFAULT_REMOTE_FD,
+                              remote_fd, op_errno, unwind);
+
+        req.fd = remote_fd;
+	req.offset = args->offset;
+	req.size = args->size;
+	memcpy(req.gfid, args->fd->inode->gfid, 16);
+
+        GF_PROTOCOL_DICT_SERIALIZE (this, args->xdata, (&req.xdata.xdata_val),
+                                    req.xdata.xdata_len, op_errno, unwind);
+
+        ret = client_submit_request(this, &req, frame, conf->fops,
+                                    GFS3_OP_DISCARD, client3_3_discard_cbk,
+				    NULL, NULL, 0, NULL, 0, NULL,
+                                    (xdrproc_t) xdr_gfs3_discard_req);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "failed to send the fop");
+
+        GF_FREE (req.xdata.xdata_val);
+
+        return 0;
+unwind:
+        CLIENT_STACK_UNWIND(discard, frame, -1, op_errno, NULL, NULL, NULL);
+        GF_FREE (req.xdata.xdata_val);
+
+        return 0;
+}
+
 /* Table Specific to FOPS */
 
 
@@ -5934,6 +6033,7 @@ rpc_clnt_procedure_t clnt3_3_fop_actors[GF_FOP_MAXVALUE] = {
         [GF_FOP_FSETATTR]    = { "FSETATTR",    client3_3_fsetattr },
         [GF_FOP_READDIRP]    = { "READDIRP",    client3_3_readdirp },
 	[GF_FOP_FALLOCATE]   = { "FALLOCATE",	client3_3_fallocate },
+	[GF_FOP_DISCARD]     = { "DISCARD",	client3_3_discard },
         [GF_FOP_RELEASE]     = { "RELEASE",     client3_3_release },
         [GF_FOP_RELEASEDIR]  = { "RELEASEDIR",  client3_3_releasedir },
         [GF_FOP_GETSPEC]     = { "GETSPEC",     client3_getspec },
@@ -5984,6 +6084,7 @@ char *clnt3_3_fop_names[GFS3_OP_MAXVALUE] = {
         [GFS3_OP_FSETATTR]    = "FSETATTR",
         [GFS3_OP_READDIRP]    = "READDIRP",
 	[GFS3_OP_FALLOCATE]   = "FALLOCATE",
+	[GFS3_OP_DISCARD]     = "DISCARD",
         [GFS3_OP_RELEASE]     = "RELEASE",
         [GFS3_OP_RELEASEDIR]  = "RELEASEDIR",
         [GFS3_OP_FREMOVEXATTR] = "FREMOVEXATTR",
diff --git a/xlators/protocol/client/src/client.c b/xlators/protocol/client/src/client.c
index 377263af6..aa2b8eef8 100644
--- a/xlators/protocol/client/src/client.c
+++ b/xlators/protocol/client/src/client.c
@@ -1997,6 +1997,40 @@ out:
 }
 
 int32_t
+client_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       size_t len, dict_t *xdata)
+{
+        int          ret  = -1;
+        clnt_conf_t *conf = NULL;
+        rpc_clnt_procedure_t *proc = NULL;
+        clnt_args_t  args = {0,};
+
+        conf = this->private;
+        if (!conf || !conf->fops)
+                goto out;
+
+        args.fd = fd;
+	args.offset = offset;
+	args.size = len;
+        args.xdata = xdata;
+
+        proc = &conf->fops->proctable[GF_FOP_DISCARD];
+        if (!proc) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "rpc procedure not found for %s",
+                        gf_fop_list[GF_FOP_DISCARD]);
+                goto out;
+        }
+        if (proc->fn)
+                ret = proc->fn (frame, this, &args);
+out:
+        if (ret)
+                STACK_UNWIND_STRICT(discard, frame, -1, ENOTCONN, NULL, NULL, NULL);
+
+	return 0;
+}
+
+int32_t
 client_getspec (call_frame_t *frame, xlator_t *this, const char *key,
                 int32_t flags)
 {
@@ -2714,6 +2748,7 @@ struct xlator_fops fops = {
         .setattr     = client_setattr,
         .fsetattr    = client_fsetattr,
 	.fallocate   = client_fallocate,
+	.discard     = client_discard,
         .getspec     = client_getspec,
 };
 
diff --git a/xlators/protocol/server/src/server-rpc-fops.c b/xlators/protocol/server/src/server-rpc-fops.c
index 4b847eabd..55e30fdd7 100644
--- a/xlators/protocol/server/src/server-rpc-fops.c
+++ b/xlators/protocol/server/src/server-rpc-fops.c
@@ -2008,6 +2008,45 @@ out:
         return 0;
 }
 
+int
+server_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno,
+                   struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
+{
+        gfs3_discard_rsp rsp	 = {0,};
+        server_state_t    *state = NULL;
+        rpcsvc_request_t  *req   = NULL;
+
+        req = frame->local;
+        state  = CALL_STATE (frame);
+
+        GF_PROTOCOL_DICT_SERIALIZE (this, xdata, (&rsp.xdata.xdata_val),
+                                    rsp.xdata.xdata_len, op_errno, out);
+
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%"PRId64": DISCARD %"PRId64" (%s) ==> (%s)",
+                        frame->root->unique, state->resolve.fd_no,
+                        uuid_utoa (state->resolve.gfid),
+                        strerror (op_errno));
+                goto out;
+        }
+
+        gf_stat_from_iatt (&rsp.statpre, statpre);
+        gf_stat_from_iatt (&rsp.statpost, statpost);
+
+out:
+        rsp.op_ret    = op_ret;
+        rsp.op_errno  = gf_errno_to_error (op_errno);
+
+        server_submit_reply(frame, req, &rsp, NULL, 0, NULL,
+                            (xdrproc_t) xdr_gfs3_discard_rsp);
+
+        GF_FREE (rsp.xdata.xdata_val);
+
+        return 0;
+}
+
 /* Resume function section */
 
 int
@@ -2975,6 +3014,26 @@ err:
         return 0;
 }
 
+int
+server_discard_resume (call_frame_t *frame, xlator_t *bound_xl)
+{
+        server_state_t *state = NULL;
+
+        state = CALL_STATE (frame);
+
+        if (state->resolve.op_ret != 0)
+                goto err;
+
+        STACK_WIND (frame, server_discard_cbk,
+                    bound_xl, bound_xl->fops->discard,
+                    state->fd, state->offset, state->size, state->xdata);
+        return 0;
+err:
+        server_discard_cbk(frame, NULL, frame->this, state->resolve.op_ret,
+			   state->resolve.op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
 
 
 /* Fop section */
@@ -3213,6 +3272,68 @@ out:
 
         return ret;
 }
+
+
+int
+server3_3_discard(rpcsvc_request_t *req)
+{
+        server_state_t    *state = NULL;
+        call_frame_t      *frame = NULL;
+        gfs3_discard_req args	 = {{0},};
+        int                ret   = -1;
+        int                op_errno = 0;
+
+        if (!req)
+                return ret;
+
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_gfs3_discard_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        frame = get_frame_from_request (req);
+        if (!frame) {
+                // something wrong, mostly insufficient memory
+                req->rpc_err = GARBAGE_ARGS; /* TODO */
+                goto out;
+        }
+        frame->root->op = GF_FOP_DISCARD;
+
+        state = CALL_STATE (frame);
+        if (!state->conn->bound_xl) {
+                /* auth failure, request on subvolume without setvolume */
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        state->resolve.type   = RESOLVE_MUST;
+        state->resolve.fd_no  = args.fd;
+
+	state->offset = args.offset;
+	state->size = args.size;
+	memcpy(state->resolve.gfid, args.gfid, 16);
+
+        GF_PROTOCOL_DICT_UNSERIALIZE (state->conn->bound_xl, state->xdata,
+                                      (args.xdata.xdata_val),
+                                      (args.xdata.xdata_len), ret,
+                                      op_errno, out);
+
+        ret = 0;
+        resolve_and_resume (frame, server_discard_resume);
+
+out:
+        free (args.xdata.xdata_val);
+
+        if (op_errno)
+                req->rpc_err = GARBAGE_ARGS;
+
+        return ret;
+}
+
+
 int
 server3_3_readlink (rpcsvc_request_t *req)
 {
@@ -5867,6 +5988,7 @@ rpcsvc_actor_t glusterfs3_3_fop_actors[] = {
         [GFS3_OP_FSETATTR]    = { "FSETATTR",   GFS3_OP_FSETATTR, server3_3_fsetattr, NULL, 0},
         [GFS3_OP_READDIRP]    = { "READDIRP",   GFS3_OP_READDIRP, server3_3_readdirp, NULL, 0},
 	[GFS3_OP_FALLOCATE]    = { "FALLOCATE",	GFS3_OP_FALLOCATE, server3_3_fallocate, NULL, 0},
+	[GFS3_OP_DISCARD]     = { "DISCARD",	GFS3_OP_DISCARD, server3_3_discard, NULL, 0},
         [GFS3_OP_RELEASE]     = { "RELEASE",    GFS3_OP_RELEASE, server3_3_release, NULL, 0},
         [GFS3_OP_RELEASEDIR]  = { "RELEASEDIR", GFS3_OP_RELEASEDIR, server3_3_releasedir, NULL, 0},
         [GFS3_OP_FREMOVEXATTR] = { "FREMOVEXATTR", GFS3_OP_FREMOVEXATTR, server3_3_fremovexattr, NULL, 0},