xf86-video-ati-7.6.1

[Fernando de Oliveira]

The xf86-video-ati 7.6.0 release supports xserver versions 1.8-1.18. [So should 7.6.1, I think.]

​http://xorg.freedesktop.org/archive/individual/driver/xf86-video-ati-7.6.1.tar.bz2

​http://xorg.freedesktop.org/archive/individual/driver/xf86-video-ati-7.6.1.tar.bz2.sig

​http://lists.x.org/archives/xorg-announce/2015-November/002658.html

MD5: ede86cd3d1b1d8882f0aea61d9e924ed xf86-video-ati-7.6.1.tar.bz2

Also included announcement for 7.6.0:

​http://lists.x.org/archives/xorg-announce/2015-November/002657.html

[ANNOUNCE] xf86-video-ati 7.6.1
Michel Dänzer michel at daenzer.net
Wed Nov 11 20:48:47 PST 2015

This is a brown paper bag release fixing a double-free bug in the 7.6.0
release.

I could swear I'd tested this appropriately yesterday, but apparently
not. :(


Michel Dänzer (3):
    ◦ Post 7.6.0 release version bump
    ◦ Remove duplicate free(output_ids) call
    ◦ Set version for 7.6.1 release

-- 
Earthling Michel Dänzer               |               http://www.amd.com
Libre software enthusiast             |             Mesa and X developer


[ANNOUNCE] xf86-video-ati 7.6.0
Michel Dänzer michel at daenzer.net
Wed Nov 11 20:18:21 PST 2015

The xf86-video-ati 7.6.0 release supports xserver versions 1.8-1.18.

New features in this release:

• DRI3 and Present extension support, allowing more efficient and secure
  OpenGL compositing
• Option "TearFree" [0] to eliminate tearing (currently only effective
  for non-rotated outputs)
• Option "ShadowPrimary" [0] which may deliver better performance for
  some 2D use cases, at the potential expense of other use cases (only
  available with glamor)
• DisplayPort 1.2 MST support
• Implemented support for Xorg -background none with glamor and fixed it
  to work better with EXA
• The contents of newly allocated buffers are always initialized before
  scanout, reducing intermittent artifacts

[0] Enabling Option "TearFree" or Option "ShadowPrimary" currently
    disables DRI page flipping.


Plus support for a few more Bonaire and Oland GPUs, and lots of other
small improvements and fixes. Thanks to everybody who contributed to
this release!


Adam Jackson (1):
    ◦ Remove dead accelDFS flag

Alex Deucher (3):
    ◦ radeon: bump version post release
    ◦ add new bonaire pci id
    ◦ add new OLAND pci id

Dave Airlie (6):
    ◦ radeon: stop caching mode resources
    ◦ radeon: move output name creation to its own function
    ◦ radeon: add support for DP 1.2 display hotplug (v2)
    ◦ radeon: adopt for new X server dirty tracking APIs.
    ◦ radeon: cleanup the entity rec
    ◦ radeon: move radeon_pixmap forward declaration into other block

David Heidelberger (2):
    ◦ radeon/vdpau: don't report VDPAU for < r300
    ◦ Handle tiling in radeon_set_shared_pixmap_backing

Emil Velikov (1):
    ◦ Do not link radeon_drv.so against libpciaccess

Jerome Glisse (2):
    ◦ Avoid leaking memory on output.
    ◦ Proper leak fix, previous leak fix was bogus.

Mario Kleiner (6):
    ◦ present: Move check for async flips
    ◦ present: Fix present notify timestamps and counts.
    ◦ Allow/Fix use of multiple ZaphodHead outputs per x-screen. (v2)
    ◦ Don't set TILE_SPLIT flags if surface.tile_split == 0.
    ◦ Fix inconsistent default eg_tile_split in evergreen_accel.c
    ◦ Make selection between DRI2 and DRI3 consistent with other
      drivers. (v2)

Michel Dänzer (78):
    ◦ Remove duplicate OPTION_PAGE_FLIP entry
    ◦ PRIME: Don't advertise offload capabilities when acceleration is
      disabled
    ◦ Prefer drmModeSetCursor2 over drmModeSetCursor
    ◦ Require at least xserver 1.8
    ◦ Move xorg_list backwards compatibility to new radeon_list.h header
    ◦ Add DRM event queue helpers v2
    ◦ DRI2: Simplify blit fallback handling for scheduled swaps
    ◦ DRI2: Remove superfluous assignments to *_info->frame
    ◦ DRI2: Move radeon_dri2_flip_event_handler
    ◦ DRI2: Use helper functions for DRM event queue management v3
    ◦ DRI2: Split out helper for getting UST and MSC of a specific CRTC
    ◦ Move #include "radeon_glamor.h" from radeon.h to where it's needed
    ◦ Fold radeon_glamor_flush into radeon_cs_flush_indirect
    ◦ Add support for SYNC extension fences v2
    ◦ Add support for the Present extension v2
    ◦ Add DRI3 support v2
    ◦ Simplify includes in radeon_{dri3,present}.c
    ◦ Add xorg_list_for_each_entry_safe fallback in radeon_list.h
    ◦ Initialize boolean variable before calling xf86GetOptValBool
    ◦ DRI3: Use open hook instead of open_client
    ◦ Always include misync.h before other misync headers
    ◦ glamor: Add glamor_fd_from_pixmap define for standalone glamor
      tree
    ◦ dri3: Use screen->CreatePixmap instead of fbCreatePixmap directly
    ◦ EXA: Return NULL from radeon_get_pixmap_bo if there is no driver
      private
    ◦ configure.ac: Check for misyncshm.h again
    ◦ Adapt radeon_sync.c for misyncshm.h vs misync.h as well
    ◦ Make radeon_do_pageflip take a BO handle directly
    ◦ Present: Add radeon_present_get_pixmap_handle helper
    ◦ glamor: Avoid generating GEM flink names for BOs shared via DRI3
      v2
    ◦ glamor: Handle GLAMOR_* flags removed from xserver
    ◦ Increase robustness against DRM page flip ioctl failures v3
    ◦ Simplify radeon_do_pageflip() error handling slightly more
    ◦ Move radeon_drm_handler/abort_proc fields to drmmode_flipdata_rec
    ◦ Move get_pixmap_handle helper to radeon_bo_helper.c
    ◦ DRI2: Use radeon_get_pixmap_handle
    ◦ Only enable SYNC extension fences and the Present extension along
      with DRI3
    ◦ Add radeon_get_pixmap_tiling_flags helper
    ◦ present: Don't flip between BOs with different tiling parameters
    ◦ Rename scanout_pixmap_x field to prime_pixmap_x
    ◦ Split out struct drmmode_scanout for rotation shadow buffer
      information
    ◦ Add RADEON_CREATE_PIXMAP_SCANOUT flag
    ◦ glamor: Remove unused function radeon_glamor_pixmap_is_offscreen
    ◦ glamor: Add radeon_pixmap parameter to
      radeon_glamor_create_textured_pixmap
    ◦ glamor: Add wrappers for the X server rendering hooks
    ◦ glamor: Add Option "ShadowPrimary" v2
    ◦ glamor: Remove the stride member of struct radeon_pixmap
    ◦ Add Option "TearFree" v4
    ◦ DRI2: Clear old->devPrivate.ptr in fixup_glamor
    ◦ Only copy fbcon BO contents if bgNoneRoot is TRUE
    ◦ Defer initial drmmode_copy_fb call until root window creation
    ◦ Defer initial modeset until the first BlockHandler invocation
    ◦ Update scanout pixmap contents before setting a mode with it
    ◦ Make drmmode_copy_fb() work with glamor as well
    ◦ On screen resize, clear the new buffer before displaying it
    ◦ Don't call radeon_get_pixmap_private() when using EXA
    ◦ Skip disabled CRTCs in drmmode_set_desired_modes() even if set_hw
      == FALSE
    ◦ glamor: Deal with glamor_glyphs_init being removed from xserver
    ◦ Replace SyncFence typedef with declaration of struct _SyncFence
    ◦ present: Remove get_drmmode_crtc helper
    ◦ glamor: Only wait for GPU writes in
      radeon_glamor_prepare_access_cpu_ro
    ◦ DRI2: Fix handling of drmmode_crtc_get_ust_msc return code
    ◦ Don't attempt a DRI2/Present page flip while the other one is
      flipping
    ◦ Skip disabled CRTCs in radeon_scanout_(do_)update
    ◦ present: Fall back to modeset for unflip operation
    ◦ present: Look at all CRTCs to determine if we can flip
    ◦ present: Handle DPMS off in radeon_present_get_ust_msc
    ◦ Remove unused local variable pRADEONEnt
    ◦ glamor: Move declaration of struct radeon_pixmap out of #if/#else
      blocks
    ◦ Only call drmmode_copy_fb (at most) once on server startup
    ◦ Wait for scanout BO initialization to finish before setting mode
    ◦ DRI2: Don't ignore rotated CRTCs in amdgpu_dri2_drawable_crtc
    ◦ DRI2: Keep MSC monotonic when moving window between CRTCs
    ◦ Only align screen / scanout pixmap height where necessary
    ◦ Call xf86CrtcRotate from initial drmmode_set_desired_modes call
    ◦ Don't advertise rotation support without hardware acceleration v2
    ◦ Eliminate redundant data parameter from
      drmmode_crtc_scanout_create
    ◦ Properly handle drmModeAddFB failure in
      drmmode_crtc_scanout_allocate
    ◦ Bump version for 7.6.0 release

Piotr Redlewski (2):
    ◦ Disable CRTCs when disabling the outputs
    ◦ Do not try to enable already enabled CRTCs in DPMS hook

Rolf Eike Beer (1):
    ◦ radeon: remove unneeded inclusion of sarea.h

Stephen Chandler Paul (1):
    ◦ Handle failures in setting a CRTC to a DRM mode properly

Tom St Denis (5):
    ◦ dri2: Avoid calculation with undefined msc value
    ◦ Simplify pick best crtc to fold two loops into one
    ◦ Clean up allocation in RADEONInitVideo()
    ◦ Simplify drmmode_set_mode_major() and avoid leaking memory.
    ◦ Clean up radeon_dri2_create_buffer2()

-- 
Earthling Michel Dänzer               |               http://www.amd.com
Libre software enthusiast             |             Mesa and X developer

[bdubbs@…]

I was able to download this and it builds correctly against xorg-server-1.18.0, however I can't test it.

Ken, can you do this one?

[ken@…]

I replied to the list yesterday, but I think that mail might have got lost in the ether. At that time, I said my available radeon machine would be doing other things. Those have now been deferred for a few days (I need to read up on things, and get a new drive), so I can pick this up when I've updated my blfs scripts (they are now about 10 days old).

That machine has blfs-7.8, so I'll start by dropping this driver into 7.8 with whichever version of the 1.17 server I installed. If that is ok, I'll update libdrm, mesa, xorg-server, all the installed drivers.

[bdubbs@…]

I would think that the only updates needed is to xorg-server-1.18.0 and rebuild any other drivers you are using in addition to the ati driver.

I am thinking about installing the latest git master for the intel and nouveau drivers. I've tested nouveau and it works for me and I've built the intel driver and it builds without error, but I've not tested it yet.

If you test the ati driver and it is OK, I'll update all three drivers and xorg-server at the same time.

[bdubbs@…]

Just following up here although it is about the Intel driver. I installed the git version of the Intel driver along with xorg-1.18.0 on my laptop and it worked nicely. We have a section in the Intel driver about problems with SandyBridge New Acceleration (SNA) and directions to use the old UXA (UMA Acceleration Architecture).

The Intel driver now works well for me (start time was very fast) with both SNA and UXA.

As soon as this ATI driver is tested I'll update to xorg-1.18.0 and the ati, intel, and nouveau drivers at the same commit.

[ken@…]

I've now got it running with 1.18, using all the versions in the book (I didn't notice that mtdev had become recommended, until configure complained). Seems to work fine, as it did with the old 7.8 versions. Those are both on the kaveri (the radeonsi part of mesa). It is also working ok on an older R600 using versions of everything else from June, so it seems good. I'll pass you the ticket.

[bdubbs@…]

Got it. It's a little late for tonight, but I'll get it in the book tomorrow.

[bdubbs@…]

Fixed at revision 16660.