[PATCH 5/5] xhci: support omitting shared hcd if either of the root hubs has no ports
Mathias Nyman
mathias.nyman at linux.intel.com
Tue Mar 8 06:55:28 PST 2022
On 4.3.2022 20.37, Heiner Kallweit wrote:
> If either of the root hubs has no ports, then we can get rid of
> overhead like the shared hcd. A major internal change is that now
> the main hcd can be USB2 or USB3.
>
> Signed-off-by: Heiner Kallweit <hkallweit1 at gmail.com>
> ---
> drivers/usb/host/xhci-mem.c | 11 +++++------
> drivers/usb/host/xhci.c | 9 ++++++---
> 2 files changed, 11 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
> index a1a17713a..ced139583 100644
> --- a/drivers/usb/host/xhci-mem.c
> +++ b/drivers/usb/host/xhci-mem.c
> @@ -2362,12 +2362,11 @@ static int xhci_setup_port_arrays(struct xhci_hcd *xhci, gfp_t flags)
> xhci->usb2_rhub.num_ports = USB_MAXCHILDREN;
> }
>
> - xhci->needs_shared_hcd = 1;
> -
> - /*
> - * Note we could have all USB 3.0 ports, or all USB 2.0 ports.
> - * Not sure how the USB core will handle a hub with no ports...
> - */
> + if (xhci->usb2_rhub.num_ports && xhci->usb3_rhub.num_ports)
> + xhci->needs_shared_hcd = 1;
> + else
> + xhci_info(xhci, "USB%u root hub has no ports\n",
> + xhci->usb2_rhub.num_ports ? 3 : 2);
This now works for xhci controllers using xhci-plat.c, but in all other cases
the the secondary hcd will still be added.
Would it make sense to instead of setting xhci->needs_shared_hcd, we
set a xhci->allow_single_roothub flag in the .reset override function?
In the xhci-plat.c case this would be in xhci_plat_setup()
We would only add the flag if the respective probe supports one roothub.
Add a helper function to check if we if really should set up just one
hcd in probe, and should call xhci_run_finished() already the he first time
xhci_run() is called (like you do in patch 3/5).
Something like:
bool xhci_has_one_roothub(struct xhci_hcd *xhci)
{
return xhci->allow_single_roothub && (!xhci->usb2_rhub.num_ports != !xhci->usb3_rhub.num_ports);
}
Thanks
-Mathias
More information about the linux-amlogic
mailing list