On Sat, Jul 10, 2010 at 02:10:19PM -0400, Alan Stern wrote: > On Sat, 10 Jul 2010, Sarah Sharp wrote: > > > (I recently found this patch languishing in my git tree. I've updated it > > to 2.6.35-rc4, but I haven't retested it.) > > > > A while back, Oliver Neukum made a patch to suspend mass storage devices > > between SCSI commands. This patch uses some of the work from Oliver's > > patch. > > > > Unfortunately, Oliver's patch made SD readers lose their minds. After a > > suspend, all previous internal state is lost and they can't tell > > whether a card had been recently inserted or removed. After a suspend, > > the first TEST_UNIT_READY command sent to an SD card reader will report > > "Media may have changed", and the next TEST_UNIT_READY command will report > > actual media presence or removal. If the SD readers are suspended between > > TEST_UNIT_READY commands, they will continually report "Media may have > > changed", and no media change will ever be detected. > > > > Disable autosuspend for a mass storage device until we know medium is not > > present, by peeking into the results of a TEST_UNIT_READY command. I > > matched against the key, asc, and ascq fields, but I believe I only really > > need to match the asc and asq fields. > > > > This patch is only useful if the autosuspend timeout is set to 0 seconds > > (an immediate suspend). You can generate the udev rule to enable > > autosuspend and set the timeout with my usb-pm tool script > > (vid-pid-to-udev-rule.sh) at http://git.moblin.org/cgit.cgi/usb-pm-tool/ > > > > This patch would be even more useful if we could get userspace to stop > > polling for new media so often. > > > > This patch works fine with empty SD card readers. Some thumb drives > > will also autosuspend when they are unmounted or ejected from the Gnome > > filesystem GUI. Other thumb drives always report "Medium may have > > changed" if they need to flash stupid LEDs, like the original Linux > > Plumbers Conference thumb drive. > > I would prefer not to have this accepted. There already are patches > waiting in James Bottomley's queue to add runtime PM support to the > SCSI subsystem; once they are merged, adding autosuspend to usb-storage > will be a 4-line change -- already written and tested. Ok. I knew that the runtime-PM patches for SCSI were in the works. However, they've been in the works for a long time now, and I wasn't sure if anyone was interested in an interim solution. I hacked this together last February, and it's saved about 0.25 watts on average for netbooks with built-in SD cards running Moblin/MeeGo. (It actually makes the power consumption of Linux better than Windows XP.) I know this is a layering violation and not *the right* solution, but I wanted to make sure the patch didn't languish out-of-tree if it could get merged. That said, I don't feel strongly about whether this particular patch gets merged. I do feel strongly that there needs to be a solution to the USB storage PM issue. > (Unfortunately James is very bad about letting people know the status > of pending patches. I have no idea when these will be merged or even > if he is willing to accept them.) I think he's giving a talk at OSCON next week, so I'll poke him if I see him. Sarah Sharp -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
