Re: kern/40607: uvideo(4) queues too many xfers to uhci / ohci

[Rafal Boni <rafal%pobox.com@localhost>]

The following reply was made to PR kern/40607; it has been noted by GNATS.

From: Rafal Boni <rafal%pobox.com@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/40607: uvideo(4) queues too many xfers to uhci / ohci
Date: Wed, 11 Feb 2009 09:37:32 -0500

 As a follow-up, taking Jeremy's suggestion to up UHCI_VFRAMELIST_COUNT
 in src/sys/dev/usb/uhcivar.h from 128 -> 1024 seems to make my machine
 not crash -- the following happens instead:
        * running 'mplayer /dev/video0' or 'mplayer tv://' complains about
          a bunch of ioctls (EINVALs IIRC) and then exits.
        * running 'mplayer -tv fps=25 tv://' produces a good video stream
          which is no longer broken up; I think mplayer still complains
          about a bunch of ioctls being rejected with EINVALs (have to
          double-check this).
        * ekiga is now able to open the device and render video, which
          it didn't do before.  The video stream is good without breakup
          or other artifacts.
        * cheese still doesn't recognize a camera for some reason.
 
        (Note that along with this change I updated my source tree to
         5.0_RC2, so it isn't exactly an apples-to-apples comparison,
         but I don't think in this area much has changed).
 
 Anyway, it seems like a net win.  However, ohci(4) probably needs some
 similar love, and I suspect both uhci(4) and ohci(4) need some bounds-
 checking on queued xfers.
 
 (I presume the errors from mplayer are artifacts of our incomplete v4l
  implementation / emulation and unrelated to any of the rest of this,
  but if those are of interest I'll collect them the next time I'm at
  the machine and append them here).
 
 --rafal