Handle Raspberry Pi 4 xhci firmware loading.
ClosedPublic
Actions

Authored by crowston_protonmail.com on Jun 14 2020, 12:22 AM.

Details

Reviewers: None
Commits: rS363795: Handle Raspberry Pi 4 xhci firmware loading.

Summary

The newer hardware revisions of the Raspberry Pi 4 removed the ability
of the VIA VL805 xhci controller to load its own firmware. Instead the
firmware must be installed at the appropriate time by the VideoCore
coprocessor.

This differential builds on top of D25068.

Test Plan

Tested with new and old firmware, on several different Pi4s.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

crowston_protonmail.com created this revision.Jun 14 2020, 12:22 AM

Herald added subscribers: emaste, andrew. · View Herald TranscriptJun 14 2020, 12:22 AM

crowston_protonmail.com requested review of this revision.Jun 14 2020, 12:22 AM

Harbormaster completed remote builds in B31691: Diff 73087.Jun 14 2020, 12:22 AM

crowston_protonmail.com edited the summary of this revision. (Show Details)Jun 14 2020, 12:23 AM

crowston_protonmail.com edited the test plan for this revision. (Show Details)

crowston_protonmail.com edited the summary of this revision. (Show Details)

andrew added inline comments.Jun 15 2020, 6:06 PM

sys/arm/broadcom/bcm2835/bcm2838_xhci.c
86–87 ↗	(On Diff #73087)	Is this correct? There is a device tree node for the XHCI device?

crowston_protonmail.com added inline comments.Jun 15 2020, 7:22 PM

sys/arm/broadcom/bcm2835/bcm2838_xhci.c
86–87 ↗	(On Diff #73087)	There is a device tree node, but it's disabled and in any case we do not use it. The controller is discovered through ordinary enumeration of the PCI bus, both on Linux and in the driver I wrote. The intention of the device tree inspection here is so I can check if the computer is a Raspberry Pi 4. The VIA controller is a commodity controller, but this quirk where the Pi Foundation removed the firmware EEPROM is limited to the Pi SBC.

maciphone2_googlemail.com mentioned this in D25068: Add driver for bcm2838 PCI express controller.Jun 25 2020, 3:03 PM

Add additional comments to explain the probe.

Harbormaster completed remote builds in B32186: Diff 74170.Jul 7 2020, 8:13 PM

I tried this on my Raspberry Pi 4 & got the following error.

xhci0: <VL805 USB 3.0 controller (on the Raspberry Pi 4b)> irq 78 at device 0.0 on pci1
mbox0: mbox response error
xhci0: warning: xhci firmware install failed (error 5).
xhci0: 32 bytes context size, 64-bit DMA

sys/arm/broadcom/bcm2835/bcm2838_xhci.c
94 ↗	(On Diff #74170)	You should place the variable decorations at the start of the function.
98 ↗	(On Diff #74170)	I thunk you can use `ofw_bus_is_compatible` here.
86–87 ↗	(On Diff #73087)	This device has no node associated with it do the call to `ofw_bus_status_okay` will call `ofw_bus_get_status` that returns `NULL` so we return a status of 'ok'.
sys/dev/usb/controller/xhci_pci.c
88–89 ↗	(On Diff #74170)	Remove the above `driver_t` & replace it with the `DEFINE_CLASS_0` version. They are equivalent, other than the variable name.

crowston_protonmail.com added inline comments.Jul 19 2020, 9:30 PM

sys/arm/broadcom/bcm2835/bcm2838_xhci.c
98 ↗	(On Diff #74170)	What's the advantage of using `ofw_bus_is_compatible`? I would have to discover the right `device_t` for the "/" device.

andrew added inline comments.Jul 20 2020, 10:21 AM

sys/arm/broadcom/bcm2835/bcm2838_xhci.c
98 ↗	(On Diff #74170)	Sorry, that should be `ofw_bus_node_is_compatible`. `fdt_is_compatible_strict` should be removed as it's redundant given we have `ofw_` versions.