[Spice-devel] [PATCH 0/8] Win32 driver fixes for 0.6.0

[alexl at redhat.com]

From: Alexander Larsson <alexl at redhat.com>

This is a bunch of changes I think we we need before the 0.6.0 release.

They do:

* Make SurfaceInfo global to fix potential memory corruption (#29053)
* Make malloc_sem global since it protects a the global mspace data
* always free all queued resources before allocating to keep down
  fragmentation
* Speed up surface id allocation

I additionally have some outstanding questions for the driver:

There are a lot of data that are not protected by any locks that
change during runtime. Both in the pdev (things like
e.g. FPUSave, update_surface_id and update_area) and in the DevRes
(like the caches and surfaces_used). With GDI apparently allowing
multiple concurrent calls to different surfaces this looks pretty
unsafe.

With surfaces_info global we avoid the problem with late resource
freeing accessing the wrong surface_info[id]. However, isn't there
also a possible problem where the pdev is not only inactive, but also
disabled. DrvDisablePDEV frees the pdev, which any outstanding
surface_info points to, so the late free would access a freed
pointer. Is there some way to have the surface resources ref the pdev
to make it not be disabled? Or is something else guaranteeing this
already?

Alexander Larsson (8):
  Move SurfaceInfo structure
  Add helper functions for surface info <-> id mapping
  Move SurfaceInfo to global data
  Always release free resources when allocating
  Make malloc_sem global
  Move surface in use check to helper function
  Store surfaces_used in a bit-array
  Add simple cache to speed up surface allocation

 display/driver.c  |   40 ++++----------------
 display/qxldd.h   |   36 +++++++++---------
 display/res.c     |  103 +++++++++++++++++++++++++++++++----------------------
 display/surface.c |   10 +++---
 display/surface.h |   92 +++++++++++++++++++++++++++++++++++++++--------
 5 files changed, 168 insertions(+), 113 deletions(-)

-- 
1.7.2.1