On Thu, 23 Mar 2000, Alan Cox wrote:
> PCI posting also takes time and will delay the cmdStatus write until the next
> read from the pci or the posting timeout expiring.
>
> As an aside I see nothing that forbids the OHCI controller from issuing
> speculative reads either
I was going to suggest something like in the patch below but I really
don't know anything about the issues you described. What are their effects?
--- linux/drivers/usb/usb-ohci.c.bak Thu Mar 23 21:52:14 2000
+++ linux/drivers/usb/usb-ohci.c Thu Mar 23 21:54:58 2000
@@ -871,6 +872,7 @@
TD_CC | TD_DP_OUT : TD_CC | TD_R | TD_DP_IN ;
td_fill (info | (cnt? TD_T_TOGGLE:toggle), data, data_len,
urb, (cnt? 0: ST_ADDR) | ADD_LEN, cnt);
cnt++;
+ wmb();
writel (OHCI_BLF, &ohci->regs->cmdstatus); /* start bulk list
*/
break;
@@ -891,6 +893,7 @@
info = usb_pipeout (urb->pipe)?
TD_CC | TD_DP_IN | TD_T_DATA1: TD_CC | TD_DP_OUT |
TD_T_DATA1;
td_fill (info, NULL, 0, urb, 0, cnt++);
+ wmb();
writel (OHCI_CLF, &ohci->regs->cmdstatus); /* start Control
list */
break;
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
