On Tue, Feb 12, 2019 at 1:18 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > > On Tue, Jan 22, 2019 at 08:13:06PM +0000, Venkatesh Srinivas wrote: > > From: Venkatesh Srinivas <venkateshs@xxxxxxxxxx> > > > > Userspace code using uio_pci_generic may enable bus-mastering by > > directly manipulating a PCI device's command register. If a userspace > > program enables bus-mastering but exits/crashes uncleanly, bus- > > mastering will still be enabled and stale DMA addresses may be > > programmed and live in the device. > > > > Disable bus-mastering unconditionally on last close of a UIO PCI fd > > to avoid this. If the device did not have bus-mastering enabled, > > pci_clear_master() is a no-op. > > > > Signed-off-by: Venkatesh Srinivas <venkateshs@xxxxxxxxxx> > > Reviewed-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx> > > Reviewed-by: Catherine Sullivan <csully@xxxxxxxxxx> > > Acked-by: Michael S. Tsirkin <mst@xxxxxxxxxx> > > --- > > drivers/uio/uio_pci_generic.c | 17 +++++++++++++++++ > > 1 file changed, 17 insertions(+) > > Why is this a "resend" and what changed from the other versions? That > should go below the --- line. > > thanks, > > greg k-h I originally sent V2 of this patch in November (https://www.spinics.net/lists/kvm/msg177574.html), it was acked by mst@xxxxxxxxxx. I didn't realize I had to cc Greg, the resend was so Greg and Patchwork would see it. Sorry about that, -- vs;
