OSDN Git Service

xen: slightly simplify bufioreq handling
authorJan Beulich <JBeulich@suse.com>
Fri, 25 Nov 2016 10:06:33 +0000 (03:06 -0700)
committerStefano Stabellini <sstabellini@kernel.org>
Mon, 28 Nov 2016 19:26:04 +0000 (11:26 -0800)
There's no point setting fields always receiving the same value on each
iteration, as handle_ioreq() doesn't alter them anyway. Set state and
count once ahead of the loop, drop the redundant clearing of
data_is_ptr, and avoid the meaningless (because count is 1) setting of
df altogether.

Also avoid doing an unsigned long calculation of size when the field to
be initialized is only 32 bits wide (and the shift value in the range
0...3).

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
xen-hvm.c

index d74e233..124ae10 100644 (file)
--- a/xen-hvm.c
+++ b/xen-hvm.c
@@ -995,6 +995,8 @@ static int handle_buffered_iopage(XenIOState *state)
     }
 
     memset(&req, 0x00, sizeof(req));
+    req.state = STATE_IOREQ_READY;
+    req.count = 1;
 
     for (;;) {
         uint32_t rdptr = buf_page->read_pointer, wrptr;
@@ -1009,15 +1011,11 @@ static int handle_buffered_iopage(XenIOState *state)
             break;
         }
         buf_req = &buf_page->buf_ioreq[rdptr % IOREQ_BUFFER_SLOT_NUM];
-        req.size = 1UL << buf_req->size;
-        req.count = 1;
+        req.size = 1U << buf_req->size;
         req.addr = buf_req->addr;
         req.data = buf_req->data;
-        req.state = STATE_IOREQ_READY;
         req.dir = buf_req->dir;
-        req.df = 1;
         req.type = buf_req->type;
-        req.data_is_ptr = 0;
         xen_rmb();
         qw = (req.size == 8);
         if (qw) {
@@ -1032,6 +1030,13 @@ static int handle_buffered_iopage(XenIOState *state)
 
         handle_ioreq(state, &req);
 
+        /* Only req.data may get updated by handle_ioreq(), albeit even that
+         * should not happen as such data would never make it to the guest.
+         */
+        assert(req.state == STATE_IOREQ_READY);
+        assert(req.count == 1);
+        assert(!req.data_is_ptr);
+
         atomic_add(&buf_page->read_pointer, qw + 1);
     }