> On 8 Jun 2015, at 20:25, Guenter Roeck <[email protected]> wrote: > >> On Tue, Mar 24, 2015 at 12:59:22PM +0000, Brian Russell wrote: >> >> >>> On 23/03/15 20:41, Greg Kroah-Hartman wrote: >>>> On Fri, Mar 20, 2015 at 02:54:44PM +0000, Brian Russell wrote: >>>> Protect uio driver from its owner being unplugged while there are open fds. >>>> Embed struct device in struct uio_device, use refcounting on device, free >>>> uio_device on release. >>>> info struct passed in uio_register_device can be freed on unregister, so >>>> null >>>> out the field in uio_unregister_device and check accesses. >>> >>> That's really not protecting anything except heavy-handed problems... >>> >>> Look at the code: >>> >>>> @@ -493,7 +499,7 @@ static unsigned int uio_poll(struct file *filep, >>>> poll_table *wait) >>>> struct uio_listener *listener = filep->private_data; >>>> struct uio_device *idev = listener->dev; >>>> >>>> - if (!idev->info->irq) >>>> + if (!idev->info || !idev->info->irq) >>>> return -EIO; >>> >>> Great, you checked the irq value, but what if it changes the very next >>> line: >>> >>>> poll_wait(filep, &idev->wait, wait); >>> >>> Or any other line within this function? Or any other function that you >>> try to check the value for in the beginning... >>> >>> This really isn't protecting anything "properly", sorry. Either we >>> don't care about it (hint, I don't think we really do), or we need to >>> properly lock things and check, and protect, things that way. >> >> The checks for irq value are already there. I added the checks for the >> idev->info ptr and deliberately nulled it in uio_unregister_device as >> the caller module may free uio_info after unregistering (dpdk's igb_uio >> does anyway) and then release will be called later when fds are closed. >> >> So I think I definitely need the check in uio_release. I didn't think >> it hurt to return early from poll/read/write if we know the device >> has been unregistered? > > What is the final verdict on this patch ? We are seeing the crash in our > system, and I would like to apply a 'final' patch if possible to get it > fixed. > It needs a bit more work. uio_info needs to live as long as the corresponding uio_device. Since they seem to always be 1:1, uio_info could embedded within uio_device (but then all the users of uio need changed) or uio_info could be a refcounted object.
Brian > Thanks, > Guenter -- 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/
