On Mon, 9 May 2011, Jan Andersson wrote:
> This patch (as1462) updates the special accessor functions defined in
> uhci-hcd.h. Rather than using a full compiler barrier, all we really
> need is the ACCESS_ONCE() mechanism, because the idea is to prevent
> the compiler from caching a value that can change at any time.
Reading through this again, I think the last line of the description
could be improved. Here's the same patch with a better description,
for your V3.
Alan Stern
This patch (as1462) updates the special accessor functions defined in
uhci-hcd.h. Rather than using a full compiler barrier, all we really
need is the ACCESS_ONCE() mechanism, because the idea is to force the
compiler to store a fixed copy of a possibly changing value.
Signed-off-by: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
---
drivers/usb/host/uhci-hcd.h | 14 ++------------
1 file changed, 2 insertions(+), 12 deletions(-)
Index: usb-2.6/drivers/usb/host/uhci-hcd.h
===================================================================
--- usb-2.6.orig/drivers/usb/host/uhci-hcd.h
+++ usb-2.6/drivers/usb/host/uhci-hcd.h
@@ -168,12 +168,7 @@ struct uhci_qh {
* We need a special accessor for the element pointer because it is
* subject to asynchronous updates by the controller.
*/
-static inline __le32 qh_element(struct uhci_qh *qh) {
- __le32 element = qh->element;
-
- barrier();
- return element;
-}
+#define qh_element(qh) ACCESS_ONCE((qh)->element)
#define LINK_TO_QH(qh) (UHCI_PTR_QH | cpu_to_le32((qh)->dma_handle))
@@ -263,12 +258,7 @@ struct uhci_td {
* We need a special accessor for the control/status word because it is
* subject to asynchronous updates by the controller.
*/
-static inline u32 td_status(struct uhci_td *td) {
- __le32 status = td->status;
-
- barrier();
- return le32_to_cpu(status);
-}
+#define td_status(td) le32_to_cpu(ACCESS_ONCE((td)->status))
#define LINK_TO_TD(td) (cpu_to_le32((td)->dma_handle))
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
