Re: [Barry-devel] The many ways to enable/disable CONFIG_USB_SUSPEND's autosuspend mode
On Sat, Sep 29, 2007 at 07:44:18PM -0700, troy engel wrote: Chris, are you aware of any problems - other than more battery usage on a laptop e.g. - with disabling USB suspend? A quick Google I know of no such problems... actually the only problems I've had are with leaving it on. :-) Obviously if you're trying for optimum battery life, then suspend is handy, but in relation to the Blackberry, you're better off just unplugging it when you don't need it connected. Can't beat 0 watts :-) indicates we're not the only ones in this category (xsane/sane seems to have lots of issues with it as well) and folks are disabling it, as well I found a thread on a Fedora list where the dev's discuss all the devices with quirks but they feel it should be enabled globally. I took a peek at Ubuntu 7.10 Gutsy's latest kernel, and they patch driver/usb/core/quirks.c with a whole list of devices, the Blackberry being one of them, so it looks like this problem will go away for those folks upgrading early. The upcoming Barry 0.9 will hopefully fix this on Ubuntu 7.04 and Fedora 7 as well. I plan to do more testing, but so far CVS's bcharge is performing admirably on my system. (Debian stable) In your tests and setup, is it bcharge doing the disabling at runtime or are you passing the parameter to the kernel at boot/initrd and disabling it globally? Bcharge uses a hint from udev to find the right device under /sys, and then searches for the files level, autosuspend, and state in the following order, writing on, -1 or 0, and 0 respectively. This only disables autosuspend for the Blackberry that udev detected was just plugged in. Bcharge can't disable this globally. That requires a reboot. - Chris - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Barry-devel mailing list Barry-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/barry-devel
Re: [Barry-devel] The many ways to enable/disable CONFIG_USB_SUSPEND's autosuspend mode
On 9/29/07, Chris Frey [EMAIL PROTECTED] wrote: It works great in my tests here. I'm letting it sit for a while to make sure the kernel stays in on mode. So far so good. Chris, are you aware of any problems - other than more battery usage on a laptop e.g. - with disabling USB suspend? A quick Google indicates we're not the only ones in this category (xsane/sane seems to have lots of issues with it as well) and folks are disabling it, as well I found a thread on a Fedora list where the dev's discuss all the devices with quirks but they feel it should be enabled globally. In your tests and setup, is it bcharge doing the disabling at runtime or are you passing the parameter to the kernel at boot/initrd and disabling it globally? thx, -te -- I live in my own little world, but it's ok; they know me here. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Barry-devel mailing list Barry-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/barry-devel
Re: [Barry-devel] The many ways to enable/disable CONFIG_USB_SUSPEND's autosuspend mode
On Sat, Sep 29, 2007 at 01:35:53AM -0400, Chris Frey wrote: I'll update bcharge soon to try to handle this automatically, but I'm not sure it will work in every case. Ok, CVS has been updated with a fix to bcharge. If your system has the file level or autosuspend under /sys/class/usb_device/*/device/power/ then you should be able to charge your Blackberry without recompiling the kernel. It works great in my tests here. I'm letting it sit for a while to make sure the kernel stays in on mode. So far so good. - Chris - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Barry-devel mailing list Barry-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/barry-devel
Re: [Barry-devel] The many ways to enable/disable CONFIG_USB_SUSPEND's autosuspend mode
On 9/29/07, Chris Frey [EMAIL PROTECTED] wrote: I know of no such problems... actually the only problems I've had are with leaving it on. :-) Heh. I was more thinking like with a USB mouse and so forth, but your below comments clarified that we're only disabling it at runtime for the single device so this is a non-issue. Bcharge can't disable this globally. That requires a reboot. Right, I gotcha - sorry I meant disabling it via the dev/sys method you spelled out, or passing the global parameter on the kernel boot line (or modules.conf/initrd) for the entire machine. Since your previous email outlined how we could do that, I didn't know which method your test setup was using. All is clear now. :) -te -- I live in my own little world, but it's ok; they know me here. - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Barry-devel mailing list Barry-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/barry-devel
[Barry-devel] The many ways to enable/disable CONFIG_USB_SUSPEND's autosuspend mode
Hi folks, Well, I spent a chunk of tonight researching the practical ways of controlling CONFIG_USB_SUSPEND, or as we'll call it below: autosuspend. Here are the results: Kernels: Version 2.6.21 behaves with autosuspend=0 meaning off, while 2.6.22 and higher needs autosuspend=-1 to turn it off. In 2.6.22, a value of 0 means immediate instead of never. Version 2.6.22 adds variables internal to the system called autosuspend_disabled and autoresume_disabled. These are controlled by the /sys/class/usb_device/*/device/power/level file. (See below) Here's a summary of files under device/power. These may or may not exist on your system depending on your kernel version and configuration. autosuspend -1 or 0 means off, depending on kernel, otherwise it is the number of seconds to autosuspend level with the settings: on - suspend is disabled, device is fully powered auto- suspend is controlled by the kernel (default) suspend - suspend is enabled permanently You can write these strings to the file to control behaviour on a per-device basis. echo on /sys/usb_device/.../device/power/level state current state of device 0 - fully powered 2 - suspended You can write these numbers to control behaviour, but any change you make here might change automatically if autosuspend is on. echo -n 0 /sys/usb_device/.../device/power/state wakeup unknown Based on the feedback from Chuck Ebbert, it is possible to disable autosuspend on a system wide basis as well. This is either done on the kernel boot command line if usbcore is compiled into the kernel, or on the module command line, if not. Here are some practical notes for various distros: Debian / Ubuntu systems: Comes with usbcore (CONFIG_USB) compiled as a module and CONFIG_USB_SUSPEND enabled (at least on Ubuntu). Therefore, to disable autosuspend you either: - recompile kernel without CONFIG_USB_SUSPEND - configure /etc/modprobe.d/ with a file containing options usbcore autosuspend=-1 or set to 0 depending on your kernel If your system needs the modprobe configuration file above, and if your system uses initrd (probably does) then you will need to rebuild the initrd for your kernel for this to take effect. For example: dpkg-reconfigure linux-image-2.6.22.1 Fedora 6 and 7 systems: === Comes with usbcore (CONFIG_USB) compiled into the the kernel, and CONFIG_USB_SUSPEND enabled on Fedora 7. Therefore to disable autosuspend you either: - recompile kernel without CONFIG_USB_SUSPEND - boot with one of the following, depending on your kernel version (2.6.21 or 2.6.22 respectively): usbcore.autosuspend=0 usbcore.autosuspend=-1 Summary: It seems that bcharge could be coded to search for device/power/level, device/power/autosuspend, and device/power/state and use some heuristics to decide what best action to take. Fortunately, the kernel gives an error if you write -1 to autosuspend and the kernel doesn't support it, in which case writing 0 seems to be the next step. :-) Unfortunately, the files under device/power do not always exist, even while autosuspend is enabled, from what I can tell. I'll update bcharge soon to try to handle this automatically, but I'm not sure it will work in every case. Anyway, I'm getting verbose, and it's late, so I thought I'd post this in the meantime. At least the info is out there and people can test. :-) Good luck, and please send feedback if you have it. - Chris - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Barry-devel mailing list Barry-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/barry-devel