Rockbox

  • Status Closed
  • Percent Complete
    100%
  • Task Type Patches
  • Category Operating System/Drivers
  • Assigned To No-one
  • Operating System All players
  • Severity Low
  • Priority Very Low
  • Reported Version Release 3.10
  • Due in Version Undecided
  • Due Date Undecided
  • Votes
  • Private
Attached to Project: Rockbox
Opened by undo - 2011-12-30
Last edited by fg - 2012-01-04

FS#12497 - Reorganize USB initialization to make it work with FreeBSD

When USB_DETECT_BY_CORE is defined (as is the case for almost all targets), the Rockbox USB stack distinguishes between connections to a USB charger and a real USB host. Initially, a charger is assumed and all USB drivers are disabled. If a real USB host is present, the first control request it sends triggers the usb_drv_usb_detect_event() callback which in turn enables the USB HID and/or mass storage drivers. Unfortunately, the call to allocate_interfaces_and_endpoints() which should allocate the endpoints required by these drivers has already occurred at that point and cannot simply be repeated due to its side effects. Instead, the USB stack relies on the host issuing a reset that will cause allocate_interfaces_and_endpoints() to be run again. This reset is issued by Linux, Windows and OS X but not by other operating systems such as FreeBSD.

The attached patch implements a fix as discussed on IRC. A new function usb_set_host_present() ensures that USB HID and/or mass storage drivers are enabled *before* the handling of the first control request begins and allocate_interfaces_and_endpoints() is called. This new function is located in usb.c, cleaning up the USB code by removing the need for a callback from usb_core.c. The define that controls this functionality is therefore renamed from USB_DETECT_BY_CORE to USB_DETECT_BY_REQUEST.

This patch has been tested with FreeBSD and Linux USB hosts.

   patch_usb.diff (11.4 KiB)
Closed by  fg
2012-01-04 22:21
Reason for closing:  Accepted
Additional comments about closing:   Warning: Undefined array key "typography" in /home/rockbox/flyspray/plugins/dokuwiki/inc/parserutils.php on line 371 Warning: Undefined array key "camelcase" in /home/rockbox/flyspray/plugins/dokuwiki/inc/parserutils.php on line 407

Committed as r31582

Admin
fg commented on 2011-12-30 20:40

I've tested this on fuzev2, e200v1 and gigabeat f with both linux (3.1) and windows (7). As far as I'm concerned this can go in as soon as we've had testing on a non-USB_DETECT_BY_CORE/USB_DETECT_BY_REQUEST software usb device (such as onda, nano2g, or classic)

MikeS commented on 2011-12-31 02:09

I posted a more extemporaneous version of this on IRC. Noone said anything. It also got rid of USB_HOSTED entirely.

MikeS commented on 2011-12-31 03:03

Looks good, works good. Tested on a stack device and bridge chip device. :)

Here's a minor tweak to have a consistent manner of setting usb_state (inside main thread loop) and keeping the code that changes usb_num_acks_to_expect nearby for easier viewing.

@frank: I don't see any change worth noting for "promiscuous" detection. It basically the same stuff as a bridge IC.

Also worth noting since I happened to notice when testing: the USB stack still isn't big enough. Screendump on gigabeat F went into an infinite loop, creating lots of .bmp images. Beast usage is 96% after screendump.

   patch_usb_02.diff (12.3 KiB)
MikeS commented on 2011-12-31 09:05

I did test the clip v1 bootloader USB mode which does not define USB_DETECT_BY_REQUEST and it was ok.

pamaury commented on 2011-12-31 10:50

Looks fine to me, I'll test it on the fuze+ but I'm confident it will work since the usb driver is the same as the e200v1.

Loading...

Available keyboard shortcuts

Tasklist

Task Details

Task Editing