Re: deadlocks in using autopm interfaces
Alan Stern <[email protected]>
| Newsgroups | gmane.linux.usb.devel |
|---|---|
| Message-ID | <[email protected]> |
On Mon, 12 Feb 2007, Oliver Neukum wrote: > Hi Alan, list, > > I am having trouble using the usb_autopm_get/put_interface. > To wake a device on demand I need to use usb_autopm_get_interface(). > However the driver needs to hold a lock while calling > usb_autopm_put_interface(). The driver's suspend method in turn > needs to take a lock to wait for ongoing IO. As the actual autosuspend > takes places in a workqueue, there's a classic AB-BA deadlock involving > the driver's lock and pm_mutex. Are you talking about the proposed usb-storage autosuspend changes? Where and why does the driver need to hold a lock while calling usb_autopm_put_interface()? > I think the cleanest solution is to provide lockless versions and let > the caller take pm_mutex. That doesn't feel right... Incidentally, you might consider the problems involved in adding autosuspend support to usb-skeleton. Nothing special is needed, but getting the locking just right requires a good deal of care; you have to worry about the BKL in addition to the driver's own io_mutex. Alan Stern ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier. Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ [email protected] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel