Re: [PATCH v2 0/4] usb: ehci-vf: take the register bases from the device tree

[Mehmet Fide <mehmet.fide@gmail.com>]

Hello Marek,

Please forgive my eagerness. Your reviews made valid points, and my way
of answering them was to fold the answers into the cover letter and
post fixed patches right away. From now on I will proceed the way you
describe: reply in the thread first and let the review settle before a
new version goes out, with per-patch changelogs below the --- line
rather than in the cover letter.

To put the answers where they belong:

- SPL without DM_USB (v1 1/3): no build of this driver can take that
  path. drivers/Makefile builds usb/host/ for SPL only when
  SPL_USB_HOST is set, and none of the five boards selecting
  ARCH_VF610 sets it, or CONFIG_SPL at all. The removed branch also
  does not compile, which is how it went unnoticed.

- The two Fixes tags (v1 2/3): both added in v2 2/4, thank you.

- fdtdec (v1 3/3): v2 4/4 uses the livetree calls throughout and the
  driver no longer includes fdtdec.h. The alias of a phandle-referenced
  node had no livetree accessor, so v2 3/4 moves the body of
  dev_read_alias_seq() down to the ofnode level; if you would rather
  see a different shape for that helper, I am happy to adjust.

- -ENOMEM from the regulator lookup (v2 2/2): v3 2/2 tolerates only
  -ENOENT (no vbus-supply) and -ENOSYS (no DM_REGULATOR, which is the
  case for four of the five vf610 boards) and fails the probe on
  anything else.

On applying the feedback throughout: I checked every return value the
series handles. The five DT parse failures all fail the probe, and the
enable path fails it on anything but -ENOSYS. The one place a return
value is ignored is the regulator disable in ehci_usb_remove(), left
best-effort on teardown as in ehci-mx6 - say the word if you would
rather have it logged.

Best regards,
Mehmet