Re: USB patch for better bus_dma and detachment support

From: Barry Bouwsma <freebsd-misuser_at_remove-NOSPAM-to-reply.NOSPAM.dyndns.dk>
Date: Tue, 8 Jun 2004 11:28:51 +0200 (CEST)
[drop the above ipv6-only address that only sometimes works from the
 recipients and I'll catch up later from the archives, thanks]


> 	http://people.freebsd.org/~iedowse/usb.diff
> that attempts to begin addressing a number of problems with our
> current USB code:
> 
>  o All data buffers were allocated using bus_dma as physically
>    contiguous memory. All transfers involve a memory copy to/from
>    these contiguous buffers, and the contiguous memory is never
>    freed. Use of these buffers is also inefficient, so low-memory
>    systems tend to quickly run out of usable physical memory.

Apologies for the lateness of this reply, but if you're interested,
I've had great success with this patch, apparently helping a few
annoying problems I've had.

I've massaged this patch to work against my -stable (with select
source files derived from -current), and so far I no longer see
the problems that had occurred regularly:
:-)  With my uhci controller, if I waited to mount and use a umass
     device until other programs had snarfed all the system memory,
     I'd get BBB timeouts when trying to access the device from
     which I could never recover.
:-(  My ehci controller, with certain devices, gave a `need data
     toggle' kernel message during access which either slowed
     access to a crawl, or inhibited any disk access, or could
     result in a panic.  This problem still does occur within
     hours of use if it occurs, but superficially it seems to
     be better than before.
The first problem has failed to recur in limited testing after
applying your patch, so it's a significant improvement for me.
So far, I appear to have stable operation of my USB storage
devices with my hardware for a change, though time will tell
if I just haven't yet triggered the above problems.


thanks, and sorry for the delay in this feedback.  if this
patch has not yet been written to apply against 4.x (only
a few trivial changes appear needed) during the time I've
been offline, I could be strong-armed into posting needed
changes to make it work with selected source.

barry bouwsma
Received on Tue Jun 08 2004 - 07:29:03 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:56 UTC