On Thu, 03 May 2007 14:40:01 +0200 Jiri Slaby <[EMAIL PROTECTED]> wrote:
> >> +static int phantom_release(struct inode *inode, struct file *file) > >> +{ > >> + struct phantom_device *dev = file->private_data; > >> + > >> + if (mutex_lock_interruptible(&dev->open_lock)) > >> + return -ERESTARTSYS; > >> + > >> + dev->opened = 0; > >> + phantom_status(dev, dev->status & ~PHB_RUNNING); > >> + > >> + mutex_unlock(&dev->open_lock); > >> + > >> + return 0; > >> +} > > > > The mutex_lock_interruptible() is wrong, I think. This function is called > > when we do the final close(). If the signal _is_ taken then it's game > > over: the device can no longer be opened. > > Yes, starving at sys_close, I think so. > > <fast searching in drivers/*> > > Wouldn't this be a problem here too: > drivers/media/dvb/cinergyT2/cinergyT2.c > drivers/usb/misc/auerswald.c > at least that the device won't be stopped or something like that? Yes, those drivers appear to have the same problem. I'll fix 'em. > >> +/* > >> + * Init and deinit driver > >> + */ > >> + > >> +static unsigned int __devinit phantom_get_free(void) > >> +{ > >> + unsigned int i; > >> + > >> + for (i = 0; i < PHANTOM_MAX_MINORS; i++) > >> + if (phantom_devices[i] == 0) > >> + break; > >> + > >> + return i; > >> +} > > > > This function assumes that the phantom_devices[] array will never have any > > holes in it. But if phantom_remove() is called out-of-order, it _will_ > > have holes. Perhaps - I didn't look too hard. > > I'm afraid I didn't get this. It goes through all phantom_devices and > returns first free index. If a hole is present it will return "i" which > corresponds to the hole, because phantom_devices[i] is 0. Note that it is > unsigned char array. > I guess I must have misread the code. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/