On Thu, Feb 13 2014, Sitsofe Wheeler wrote:
> On Thu, Feb 13, 2014 at 09:17:33AM -0700, Jens Axboe wrote:
> > 
> > Initially I didn't see the issue, but then I realized that ->io_us is a
> > pointer to the io_u pointer. So it is an issue. The fix isn't super
> > pretty, but it gets rid of the bug, so I'll apply it. It might be nicer
> > to split it into a top and bottom define.
> 
> Can you explain this top and bottom define more - would this let me turn
> it into a while loop?
> 
> The only reason for the current abuse was because I couldn't think of a
> another way to fix it while preserving the macro...

Basically you turn it into a do-while with two macros. Ala the below,
completely untested...

diff --git a/backend.c b/backend.c
index 32bc2652bd0b..e355447d1e3a 100644
--- a/backend.c
+++ b/backend.c
@@ -255,13 +255,13 @@ static void cleanup_pending_aio(struct thread_data *td)
                struct io_u *io_u;
                int i;
 
-               io_u_qiter(&td->io_u_all, io_u, i) {
+               io_u_do_qiter(&td->io_u_all, io_u, i) {
                        if (io_u->flags & IO_U_F_FLIGHT) {
                                r = td->io_ops->cancel(td, io_u);
                                if (!r)
                                        put_io_u(td, io_u);
                        }
-               }
+               } while_each_io_u(&td->io_u_all, io_u, i);
        }
 
        if (td->cur_depth)
diff --git a/engines/posixaio.c b/engines/posixaio.c
index 2df26af3848e..3b27fcdfcc34 100644
--- a/engines/posixaio.c
+++ b/engines/posixaio.c
@@ -111,7 +111,7 @@ static int fio_posixaio_getevents(struct thread_data *td, 
unsigned int min,
 restart:
        memset(suspend_list, 0, sizeof(*suspend_list));
        suspend_entries = 0;
-       io_u_qiter(&td->io_u_all, io_u, i) {
+       io_u_do_qiter(&td->io_u_all, io_u, i) {
                int err;
 
                if (io_u->seen || !(io_u->flags & IO_U_F_FLIGHT))
@@ -138,7 +138,7 @@ restart:
                        io_u->resid = io_u->xfer_buflen - retval;
                } else
                        io_u->error = err;
-       }
+       } while_each_io_u(&td->io_u_all, io_u, i);
 
        if (r >= min)
                return r;
diff --git a/engines/windowsaio.c b/engines/windowsaio.c
index 16df74035f18..c6ff27a4408f 100644
--- a/engines/windowsaio.c
+++ b/engines/windowsaio.c
@@ -275,7 +275,7 @@ static int fio_windowsaio_getevents(struct thread_data *td, 
unsigned int min,
        }
 
        do {
-               io_u_qiter(&td->io_u_all, io_u, i) {
+               io_u_do_qiter(&td->io_u_all, io_u, i) {
                        if (!(io_u->flags & IO_U_F_FLIGHT))
                                continue;
 
@@ -290,7 +290,7 @@ static int fio_windowsaio_getevents(struct thread_data *td, 
unsigned int min,
 
                        if (dequeued >= min)
                                break;
-               }
+               } while_each_io_u(&td->io_u_all, io_u, i);
 
                if (dequeued < min) {
                        status = WaitForSingleObject(wd->iocomplete_event, 
mswait);
diff --git a/io_u_queue.h b/io_u_queue.h
index 5b6cad0ef173..649dcb5ca67c 100644
--- a/io_u_queue.h
+++ b/io_u_queue.h
@@ -28,8 +28,14 @@ static inline int io_u_qempty(struct io_u_queue *q)
        return !q->nr;
 }
 
-#define io_u_qiter(q, io_u, i) \
-       for (i = 0; i < (q)->nr && (io_u = (q)->io_us[i]); i++)
+#define io_u_do_qiter(q, io_u, i)                              \
+       (i) = 0;                                                \
+       do {                                                    \
+               (io_u) = (q)->io_us[i];                         \
+               (i)++;                                          \
+
+#define while_each_io_u(q, io_u, i)                            \
+       } while ((i) < (q)->nr);                                \
 
 int io_u_qinit(struct io_u_queue *q, unsigned int nr);
 void io_u_qexit(struct io_u_queue *q);

-- 
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to