From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Brownell Date: Sat, 18 Aug 2001 16:49:52 +0000 Subject: Re: xircom cbem56g-100 support Message-Id: List-Id: References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-hotplug@vger.kernel.org > > Workaround: comment it out of that blacklist. > > it now works. As should be ... :) > another quick question: > > this is a multifuntion card and hotplug loads the network driver, but is > hotplug or cardmgr responsible for loading the serial driver (esp. since > i'm testing the in-kernel pcmcia drivers and that means just loading the > regular serial.o module)? If this is done with multiple PCI functions, then I'd expect normal PCI hotplugging to trigger loading the right module. If the serial function isn't exposed using a PCI function (that is, if "lspci" doesn't show it) then hotplugging could load "serial.o" by adding a /etc/modules/pci.handmap functionality ... assuming that cardmgr is really trying to be ignorant of this card. That's quite possibly a gap in the current cardbus hotplug support. (Want to implement such support?) > and finally, there seems to be a disadvantage of hotplug at this point: > > with pcmcia-cs, if you tried to do a cardctl eject when the interface was > still UP it would return a "ioctl(): Device or resource busy." with > hotplug, since cardmgr doesn't do the configuration, you can eject a card > that still has a live interface. i know people should know better. does > hotplug have functionality similar to "unconfigure and remove?" I don't know of a good solution for such problems. As I recall, "cardmgr" and "cardctl" interact through some sort of state that records driver/hardware bindings. Using just "modutils", that sort of functionality doesn't exist. The lack shows up in other places; it's why hotplugging doesn't have an automated way to "rmmod". What's needed is IMO a generic mechanism that can work for USB, PCI/Cardbus, and other kinds of hotpluggable drivers. It should happen automatically when the hardware is removed. Presumably there should be a software-driven version like "cardctl eject", to support cleaner device shutdown protocols. And when power management suspends the host, the clean shutdown should kick in (devices get removed then). I know that USB needs some new kernel driver hooks to handle that, we know roughly what they need to be, but I suspect those new APIs won't get added till the 2.5 work starts. (Backport should be easy, but every HC driver needs updating.) - Dave _______________________________________________ Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net Linux-hotplug-devel@lists.sourceforge.net http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel