${ux_ehci_ed_microframe_ssplit_count} is not calculated correctly

[xuzihan351]

I am using USBX host to access the mouse device. I found that some mouse insertion and removal operations caused the program to enter a hard fault. I have found that as long as the device has multiple interrupt endpoints, this situation will occur. So I tracked the correlation function of the interrupt endpoint. I found that in the ux_hcd_ehci_interrupt_endpoint_destroy function, the value of this variable changed from 0 to 255. When the device is plugged in again, _ux_host_stack_new_device_create will return UX_NO_BANDWIDTH_AVAILABLE because of the ux_ehci_ed_microframe_ssplit_count == 255. So I add a if ux_ehci_ed_microframe_ssplit_count > 0 before the ux_ehci_ed_microframe_ssplit_count--, the bug has disapperared. Please check this error and let me know the correct repair method.

[xiaocq2001]

Hi, please try change interval in following lines to 8:

https://github.com/azure-rtos/usbx/blob/ef2fa7717be8ab751fdfb5d3f91e24fbbaea8dd6/common/usbx_host_controllers/src/ux_hcd_ehci_interrupt_endpoint_create.c#L269

https://github.com/azure-rtos/usbx/blob/ef2fa7717be8ab751fdfb5d3f91e24fbbaea8dd6/common/usbx_host_controllers/src/ux_hcd_ehci_interrupt_endpoint_create.c#L296

and check if that helps to remove the split count check.

[xuzihan351]

It also result in an exception.

[xiaocq2001]

Thanks for the feedback. I think we may need more time to find out what's the root cause. In normal case, ux_ehci_ed_microframe_ssplit_count is increased on endpoint creating and decreased on endpoint destroy.

[xuzihan351]

I tested several mouse and tracked the process of creating interrupt endpoints. For mouses with interrupt reporting cycles longer than 1 millisecond, ux_ehci_ed_microframe_ssplit_count will not increase during endpoiint creating. But it will decrease during destory.

[xiaocq2001]

https://github.com/azure-rtos/usbx/blob/7c928b43db68b72970b3effd5a2582eb5a6869c7/common/usbx_host_controllers/src/ux_hcd_ehci_interrupt_endpoint_create.c#L243 interval calculation is not correct, if endpoint interval is larger than 1ms, final checked interval in code should be 8. Please try following code:

    /* Keep interval <= 1ms for micro-frame calculation.  */
    /* Make it index steps to move.  */
    if (interval >= 4)
    {
        interval = 3; /* Uses 3 for 1ms calculation.  */
    }
    else if (interval > 0)
    {
        interval --;
        interval &= 0x3;
    }
    interval = (1u << interval); /* 1 (1/8ms), 2, 4, 8 (1ms)  */

With this change, the previous interval change in L269 and L296 is not necessary.

[xuzihan351]

It works!

[xiaocq2001]

Glad to hear that.

[xiaocq2001]

Fix is in latest code. Can close.