Fishwaldo opened a new pull request, #19862:
URL: https://github.com/apache/nuttx/pull/19862
## Summary
The first twenty-eight commits belong to #19745, #19860 and #19861 and are
not
part of this change. This PR adds the last five.
The xHCI driver refused `CONFIG_USBHOST_HUB` at compile time. A controller
reaches a device behind a hub by the path to it, and for a low or full speed
device through the hub that translates for it, and none of that was
described.
- **A device is keyed by the port it occupies, not by the root port.** The
slot, the default control endpoint and the device context lived in the root
port structure, which holds only while every device is plugged straight
into
the controller. Two keys replace it: an endpoint records the slot it was
opened on, and a hub port belongs to one device wherever it sits. No
functional change for a directly attached device, whose port slot and
endpoint slot are the same one.
- **A hub reports what it is on the port it occupies.** Some controllers must
be told about the hubs in a topology, not only about the device at the end
of
it: a hub's slot context carries a hub flag, its downstream port count and
its transaction translator think time. The hub class driver already reads
both values from the hub descriptor; this publishes them on the hub's own
hub
port, beside the speed and function address already there. Nothing is
required to read them, so a controller that does not need them is
unaffected.
- **A device behind a hub is described to the controller.** The route string
is
the path to it, a nibble per tier with the tier nearest the root lowest,
stopping after five as the field and USB both require. Slot context dword 2
names the transaction translator, reported by slot rather than by USB
address
as EHCI does, and naming the nearest high speed ancestor rather than the
immediate parent. `xhci_epalloc()` carried a copy of `sam_ehci.c`'s
split-transaction block, writing fields this driver never read; both are
removed.
- **Hub support is implemented.** A device is created wherever it sits; the
hub
asks for a port's control endpoint before it reports the connection, so an
endpoint may exist before its slot does; a hub must be described as a hub
before anything behind it can be reached, which is only known once its
class
driver has read the descriptor, so `xhci_hub_update()` corrects the slot
context the first time something appears behind it; and a hub reports each
changed port without waiting for the last, so the connect method queues
them
rather than holding one pointer.
- **The host stack is told which controller a port belongs to.**
`struct usbhost_roothubport_s` carries that number and nothing set it.
One commit touches shared code outside the xHCI driver: reporting the hub's
port count and think time adds two fields to `struct usbhost_hubport_s`, both
inside `CONFIG_USBHOST_HUB`, as the parent pointer already is.
Multi-TT is not implemented. It comes from the hub's interface protocol
rather
than its descriptor, and driving a multi-TT hub as single-TT costs bandwidth
behind that hub but is correct.
## Impact
Enables `USBHOST_HUB` with `USBHOST_XHCI`, previously rejected at compile
time.
No change when hub support is disabled; the two new `usbhost_hubport_s`
fields
are compiled out without it.
## Testing
EIC7700 EVB, Synopsys DWC3 in host mode, with a real Fresco Logic 6-port USB
2.0 hub:
```
usb 1-1: keyboard, driver attached
usb 0-1: hub, driver attached (USB2.0 Hub, Fresco Logic)
usb 0-1.1: mass storage, driver attached -> /dev/sda
usb 0-1.2: misc, driver attached (SIPEED UARTx4 HS) -> /dev/ttyACM0-3
```
The device behind the hub mounts as a 59 GB vfat volume, its directory lists,
and `dd if=/dev/sda bs=512 count=256` reads at 703 KB/s.
QEMU with `-device qemu-xhci` plus a `usb-hub` carrying both a `usb-storage`
and a `usb-kbd`: both enumerate behind the hub, giving `/dev/sda` and
`/dev/kbda`, the file reads back `qemu-xhci-regression-ok`, and no assertions
fire.
Note for anyone reproducing this: `qemu-intel64:jumbo` has xHCI, mass storage
and HID keyboard but does not set `CONFIG_USBHOST_HUB`, so hub coverage needs
it enabled.
Depends-On: https://github.com/apache/nuttx/pull/19861
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
