[Spice-devel] [PATCH qxl-wddm-dod v5 7/7] Fixing possible BSOD
Sameeh Jubran
sameeh at daynix.com
Mon Sep 26 14:19:28 UTC 2016
On Mon, Sep 26, 2016 at 4:33 PM, Frediano Ziglio <fziglio at redhat.com> wrote:
> >
> > Interrupts seem to arrive to the driver before the initialization phase
> > is over (m_pHWDevice = NULL), in that case we can't handle interrupts
> yet.
>
> I would remove the "seem"
>
> > Even
> > when m_pHWDevice isn't NULL, other fields aren't necessarly fully
> intialized
> > till
> > the StartDevice function has finished initialization, thus the flag
> > DriverStarted should be checked upon interrupts.
> >
> > Note: There is no way provided by microsoft to disable interrupts in WDDM
> > drivers.
> >
> > Signed-off-by: Sameeh Jubran <sameeh at daynix.com>
> > ---
> > qxldod/QxlDod.cpp | 6 +++++-
> > 1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/qxldod/QxlDod.cpp b/qxldod/QxlDod.cpp
> > index 3fb30ce..7de64e9 100755
> > --- a/qxldod/QxlDod.cpp
> > +++ b/qxldod/QxlDod.cpp
> > @@ -62,6 +62,7 @@ QxlDod::QxlDod(_In_ DEVICE_OBJECT*
> pPhysicalDeviceObject) :
> > m_pPhysicalDevice(pP
> > PAGED_CODE();
> > DbgPrint(TRACE_LEVEL_INFORMATION, ("---> %s\n", __FUNCTION__));
> > *((UINT*)&m_Flags) = 0;
> > + m_Flags.DriverStarted = FALSE;
> > RtlZeroMemory(&m_DxgkInterface, sizeof(m_DxgkInterface));
> > RtlZeroMemory(&m_DeviceInfo, sizeof(m_DeviceInfo));
> > RtlZeroMemory(m_CurrentModes, sizeof(m_CurrentModes));
>
> This should be no-op but better to explicit.
>
> > @@ -1684,7 +1685,10 @@ VOID QxlDod::DpcRoutine(VOID)
> > BOOLEAN QxlDod::InterruptRoutine(_In_ ULONG MessageNumber)
> > {
> > DbgPrint(TRACE_LEVEL_INFORMATION, ("<--> 0 %s\n", __FUNCTION__));
> > - return m_pHWDevice->InterruptRoutine(&m_DxgkInterface,
> MessageNumber);
> > + if (m_Flags.DriverStarted) {
> > + return m_pHWDevice ? m_pHWDevice->InterruptRoutine(
> &m_DxgkInterface,
> > MessageNumber) : FALSE;
> > + }
> > + return FALSE;
> > }
> >
> > VOID QxlDod::ResetDevice(VOID)
>
> What about
>
> if (m_Flags.DriverStarted && m_pHWDevice) {
> return m_pHWDevice->InterruptRoutine(&m_DxgkInterface,
> MessageNumber);
> }
> return FALSE;
>
> Rest is fine, I can do the change by myself if you agree or I'll merge as
> it otherwise.
>
I agree, go ahead :)
>
> Acked-by: Frediano Ziglio <fziglio at redhat.com>
>
> Frediano
>
--
Respectfully,
*Sameeh Jubran*
*Linkedin <https://il.linkedin.com/pub/sameeh-jubran/87/747/a8a>*
*Junior Software Engineer @ Daynix <http://www.daynix.com>.*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/spice-devel/attachments/20160926/29be70f5/attachment-0001.html>
More information about the Spice-devel
mailing list