Improve periodic USB transfers (device intr, isoc) used for input
devices, audio, and video. It's still not perfect, and will need
further improvements.

High level, the diff contains following chan

Improve periodic USB transfers (device intr, isoc) used for input
devices, audio, and video. It's still not perfect, and will need
further improvements.

High level, the diff contains following changes:

* Sync up with the Linux code base, which did re-work the periodic
scheduling code path.
* Run the driver in IPL_VM instead of IPL_USB to prioritize us before
lower/equal interrupts (same what NetBSD does).
* Add two new flags to our USB stack required by the updated driver
code:
- 'multi' flag in the usbd_hub structure to keep track whether
a hub has one Transaction Translator for all ports (single TT)
or one Transaction Translator per port (multi TT).
- 'hcpriv' pointer in the usbd_tt structure for the HC driver to
allocate memory for the scheduling depending on single or multi TT.

"go for it" kettenis@

show more ...

constify USB host controller method tables and root hub descriptors

ok mglocker@

Instead of passing the flags as part of a struct that's supposed to be
filled by the callee, change the usb_allocmem() API to take another
argument for the flags.

ok mpi@

Switch USB to use non-coherent buffers for data transfers. Since
the import in '99 all buffers allocated using usb_allocmem() have
been mapped COHERENT. On some ARM SoCs, where the USB controller
i

Switch USB to use non-coherent buffers for data transfers. Since
the import in '99 all buffers allocated using usb_allocmem() have
been mapped COHERENT. On some ARM SoCs, where the USB controller
is not coherent with the caches, this means the buffers were mapped
uncached. This drastically reduces the performance, especially on
reads.

We already added cache syncs before and after USB transfers, but so
far those have essentially been no-ops. Since other drivers make
use of the same allocation code, and those haven't been proven to
have correct syncs, we can't just remove the COHERENT flag.

This splits the allocation into coherent and non-coherent blocks.
All drivers who call usb_allocmem() themselves now pass a flag to
show they require coherent blocks. Onced verified that they also
work fine without coherent, or once they have been refactored, we
can remove this again.

On a ure(4) connected to an i.MX8MQ, the receive performance is
10x as fast as before. The Raspberry Pi's ethernet receive speed
doubled.

Debugged using dt(4)
Original diff from Marius Strobl
Feedback from kettenis@
Tested on an i.MX8MQ (arm64) by patrick@
Tested on a Raspberry Pi (arm64) by tobhe@
Tested on an ERL (octeon) by Janne Johansson
ok mpi@ gerhard@ tobhe@

show more ...

Free sizes for the subdevs array.

ok anton@, visa@

Revert free sizes for `subdevs' array, it is incorrect.

Reported by anton@ and Natasha Kerensikova

free(9) size for USB subdevs array.

ok ratchov@, visa@

free(9) sizes for endpoints array.

ok okan@, tedu@, visa@

Simplify/refactor the way vendor/product/serial informations are
gathered from usb devices.

Cache them early in usbd_new_device() instead of querying the device
several times (the content wont chang

Simplify/refactor the way vendor/product/serial informations are
gathered from usb devices.

Cache them early in usbd_new_device() instead of querying the device
several times (the content wont change anyway) and uselessly generating
usb traffic (which could conflict with other access..)

The heuristic stays the same, first look in the device descriptor, then
try to match against usb_known_vendors/usb_known_products, then fallback
to use the idVendor/idProduct codes.

Remove the now useless 'usedev' parameter from usbd_fill_deviceinfo().

Tested from bsd & bsd.rd.

suggested by and ok mpi@

show more ...

Fill the 'card' member of v4l2_capability struct with the usb product
name coming from usbd_devinfo_vp() instead of the dummy "Generic USB
video class device" string, makes it easier to differentiate

Fill the 'card' member of v4l2_capability struct with the usb product
name coming from usbd_devinfo_vp() instead of the dummy "Generic USB
video class device" string, makes it easier to differentiate multiple
webcams in firefox's webrtc permission dialog.

ok/help mpi@

show more ...

Add support for dumping USB transfers via bpf(4) using USBPcap headers.

ok deraadt@, dlg@

A pile of sizes to free(9). In test for a few days in snapshots.
Errors will result in nice clean panic messages so we know what's wrong.
Reviewed by dhill visa natano jsg.

Get rid of usbd_get_device_strings() because we use it only once.

Return a char * rather using a void function for usbd_get_device_string()

ok patrick@

Cache vendor, product and serial info for each usb device. This allows
ioctl(USB_DEVICEINFO) not to issue any further requests to gather
information. Thus reducing stress on connected usb devices.

Cache vendor, product and serial info for each usb device. This allows
ioctl(USB_DEVICEINFO) not to issue any further requests to gather
information. Thus reducing stress on connected usb devices.
This fixes an issue where usbdevs called in a loop causes a USB mass
storage device to halt operation.

Adapted from a similar commit in NetBSD.

ok mpi@

show more ...

Always clear a port reset feature after setting it, or at least try to.
Otherwise we might leave a port in an unwanted state.

Found while investigating timeout issues on xhci(4).

When a bus is explored, do not probe the ports which status hasn't
changed. This saves a lot of I/O when attaching/detaching devices
and might help with some timing related problems.

Has been in sn

When a bus is explored, do not probe the ports which status hasn't
changed. This saves a lot of I/O when attaching/detaching devices
and might help with some timing related problems.

Has been in snap for ten days, committing now so that people testing
xhci(4) test the same thing w/ snapshots and their own kernel.

show more ...

Add some fields needed for TT support.

Remove USB locators. They are currently unused and this wont change due
to the way USB buses are discovered.

Remove the port status argument from usbd_reset_port(). We don't do
anything with it and it simplifies this mess in order to implement
warm reset.

Use understandable messages when the per-xfer poison value is incorrect.
Prodded by a comment from stsp@.

XFER_FREE is not used, ciao.

Do not store the whole USB hub descriptor in the "struct usbd_hub"
to help integrating super speed hubs that use a different descriptor.

autoconf(9) is your friend and it knows you more than you think. It
even knows your children! So let him handle the seperation correctly.

Do not rewrite your own custom config_detach_children(9) a

autoconf(9) is your friend and it knows you more than you think. It
even knows your children! So let him handle the seperation correctly.

Do not rewrite your own custom config_detach_children(9) and as a bonus
start removing the usbd_port tentacles from the stack.

ok pirofti@, yuo@, miod@, deraadt@

show more ...

Adds an optional bus function to set the address of a new device and
explicitly set it when required.

Right now xhci(4) does not need such function because it assigns
addresses when the first pipe o

Adds an optional bus function to set the address of a new device and
explicitly set it when required.

Right now xhci(4) does not need such function because it assigns
addresses when the first pipe of a device is opened.

ok yuo@, pirofti@

show more ...

Apart from the early exploration done to find a console keyboard during
the boot process, USB devices must be attached or detached from the usb
task thread in order to avoid races with periodical exp

Apart from the early exploration done to find a console keyboard during
the boot process, USB devices must be attached or detached from the usb
task thread in order to avoid races with periodical explorations issued
by uhub(4) interrupts.

Respect this rule when detaching root hubs during a suspend/resume cycle
and avoid some hangs due to the aforementioned race.

Tested by Mattieu Baptiste, thanks!

show more ...