Date: Sun, 27 Oct 2002 19:02:40 -0700 (MST) From: "M. Warner Losh" <imp@bsdimp.com> To: iedowse@maths.tcd.ie Cc: freebsd-mobile@FreeBSD.ORG Subject: Re: Patch to fix/shorten "wi" freezes Message-ID: <20021027.190240.59654773.imp@bsdimp.com> In-Reply-To: <200210280111.aa16990@salmon.maths.tcd.ie> References: <20021027.170646.118304063.imp@bsdimp.com> <200210280111.aa16990@salmon.maths.tcd.ie>
next in thread | previous in thread | raw e-mail | index | archive | help
In message: <200210280111.aa16990@salmon.maths.tcd.ie>
Ian Dowse <iedowse@maths.tcd.ie> writes:
: In message <20021027.170646.118304063.imp@bsdimp.com>, "M. Warner Losh" writes:
: >
: >For the removal, it would be better to use the bus_child_present() api
: >for the eject case.
:
: Does the pccard system implement bus_child_present()? I just did a
: quick grep for *_child_present there without finding anything, but
: maybe I'm looking for the wrong thing. Is the idea to use something
: like
:
: if (sc->wi_gone)
: return;
: <hardware operation>
: if (timeout) {
: if (!bus_child_present(sc->dev)) {
: sc->wi_gone = 1;
: return;
: }
: device_printf(sc->dev, "device timeout\n");
: }
:
: so that further slow timeouts can be avoided if the device has
: really been removed? I presume it is too much overhead to just call
: bus_child_present() everywhere instead of testing `gone'.
Well, I think I have them in my tree, but haven't committed them to
the tree yet.
The reason that I'd not call it all the time is because once a device
is gone, it should stay gone until it is reattached. At least that's
been my experience with debouncing the pccard stuff. Also, once it is
gone, it is a lot cheaper to check a flag, than to go to hardware,
which may take several microseconds.
Warner
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-mobile" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20021027.190240.59654773.imp>