- 16 Aug, 2016 1 commit
-
-
Chris Wilson authored
Daniel Vetter proposed a new challenge to the serialisation inside the busy-ioctl that exposed a flaw that could result in us reporting the wrong engine as being busy. If the request is reallocated as we test its busyness and then reassigned to this object by another thread, we would not notice that the test itself was incorrect. We are faced with a choice of using __i915_gem_active_get_request_rcu() to first acquire a reference to the request preventing the race, or to acknowledge the race and accept the limitations upon the accuracy of the busy flags. Note that we guarantee that we never falsely report the object as idle (providing userspace itself doesn't race), and so the most important use of the busy-ioctl and its guarantees are fulfilled. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Daniel Vetter <daniel.vetter@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
Mika Kuoppala <mika.kuoppala@intel.com> Reviewed-by:
Daniel Vetter <daniel.vetter@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471337440-16777-1-git-send-email-chris@chris-wilson.co.uk
-
- 15 Aug, 2016 39 commits
-
-
Chris Wilson authored
Reported-by: 0day kbuild test robot Fixes: 2bd160a1 ("drm/i915: Reduce i915_gem_objects to only show...") Signed-off-by:
-
Chris Wilson authored
Just another useful register to inspect following a GPU hang. v2: Remove partial decoding of RING_MODE to userspace, be consistent and use GEN > 2 guards around RING_MODE everywhere. Signed-off-by:
-
Chris Wilson authored
There is no other state pertaining to the completed requests in the hang, other than gleamed through the ringbuffer, so including the expired requests in the list of outstanding requests simply adds noise. Signed-off-by:
Matthew Auld <matthew.auld@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471254551-25805-31-git-send-email-chris@chris-wilson.co.uk
-
Chris Wilson authored
It is useful when looking at captured error states to check the recorded BBADDR register (the address of the last batchbuffer instruction loaded) against the expected offset of the batch buffer, and so do a quick check that (a) the capture is true or (b) HEAD hasn't wandered off into the badlands. Signed-off-by:
-
Chris Wilson authored
Since contexts are not currently shared between userspace processes, we have an exact correspondence between context creator and guilty batch submitter. Therefore we can save some per-batch work by inspecting the context->pid upon error instead. Note that we take the context's creator's pid rather than the file's pid in order to better track fd passed over sockets. Signed-off-by:
-
Chris Wilson authored
This little helper only exists to safely discard the upper unused 32bits of the general 64-bit VMA address - as we know that all Global GTT currently are less than 4GiB in size and so that the upper bits must be zero. In many places, we use a u32 for the global GTT offset and we want to document where we are discarding the full VMA offset. Signed-off-by:
-
Chris Wilson authored
Treat the VMA as the primary struct responsible for tracking bindings into the GPU's VM. That is we want to treat the VMA returned after we pin an object into the VM as the cookie we hold and eventually release when unpinning. Doing so eliminates the ambiguity in pinning the object and then searching for the relevant pin later. v2: Joonas' stylistic nitpicks, a fun rebase. Signed-off-by:
-
Chris Wilson authored
In a few places, we repeat a call to clear a pointer to a vma whilst unpinning and releasing a reference to its owner. Refactor those into a common function. Signed-off-by:
-
Chris Wilson authored
Signed-off-by:
-
Chris Wilson authored
Signed-off-by:
-
Chris Wilson authored
Signed-off-by:
-
Chris Wilson authored
Signed-off-by:
-
Chris Wilson authored
Since the intel_engine_init_seqno() is shared by all engine submission backends, move it out of the legacy intel_ringbuffer.c and into the new home for common routines, intel_engine_cs.c Signed-off-by:
-
Chris Wilson authored
Since the scratch allocation and cleanup is shared by all engine submission backends, move it out of the legacy intel_ringbuffer.c and into the new home for common routines, intel_engine_cs.c Signed-off-by:
-
Chris Wilson authored
Signed-off-by:
-
Chris Wilson authored
Use the GGTT VMA as the primary cookie for handing ring objects as the most common action upon the ring is mapping and unmapping which act upon the VMA itself. By restructuring the code to work with the ring VMA, we can shrink the code and remove a few cycles from context pinning. v2: Move the flush of the object back to before the first pin. We use the am-I-bound? query to only have to check the flush on the first bind and so avoid stalling on active rings. Lots of little renames and small hoops. Signed-off-by:
-
Chris Wilson authored
Access through the GTT requires the device to be awake. Ideally i915_vma_pin_iomap() is short-lived and the pinning demarcates the access through the iomap. This is not entirely true, we have a mixture of long lived pins that exceed the wakelock (such as legacy ringbuffers) and short lived pin that do live within the wakelock (such as execlist ringbuffers). Signed-off-by:
-
Chris Wilson authored
We know that the only access to the context object is via the GPU, and the only time when it can be out of the GPU domain is when it is swapped out and unbound. Therefore we only need to clflush the object when binding, thus avoiding any potential stall on touching the domain on an active context. Signed-off-by:
-
Chris Wilson authored
When working with contexts, we most frequently want the GGTT VMA for the context state, first and foremost. Since the object is available via the VMA, we need only then store the VMA. v2: Formatting tweaks to debugfs output, restored some comments removed in the next patch Signed-off-by:
-
Chris Wilson authored
v2: Rename functions to suit their more active role Signed-off-by:
-
Chris Wilson authored
Lookup the GGTT vma once for the object assigned to the fence, and then derive everything from that vma. Signed-off-by:
-
Chris Wilson authored
Since the guc allocates and pins and object into the GGTT for its usage, it is more natural to use that pinned VMA as our resource cookie. v2: Embrace naming tautology v3: Rewrite comments for guc_allocate_vma() Signed-off-by:
-
Chris Wilson authored
The VMA are unreferenced, they belong to the object and live until they are closed. However, if we want to use the VMA as a cookie and use it to keep the object alive, we want to hold onto a reference to the object for the lifetime of the VMA cookie. To facilitate this, add a couple of simple wrappers for managing the reference count on the object owning the VMA. Signed-off-by:
-
Chris Wilson authored
A simple little macro to clear a pointer and return the old value. This is useful for writing value = *ptr; if (!value) return; *ptr = 0; ... free(value); in a slightly more concise form: value = fetch_and_zero(ptr); if (!value) return; ... free(value); with the idea that this establishes a pattern that may be extended for atomic use (using xchg or cmpxchg) i.e. atomic_fetch_and_zero() and similar to llist. Signed-off-by:
-
Chris Wilson authored
In many places, we wish to store the VMA in preference to the object itself and so being able to create the persistent VMA is useful. Signed-off-by:
-
Chris Wilson authored
Previously, we would only set the vma->pages pointer for GGTT entries. However, if we always set it, we can use it to prettify some code that may want to access the backing store associated with the VMA (as assigned to the VMA). Signed-off-by:
-
Chris Wilson authored
It's an outright programming error, so explode if it is ever hit. Signed-off-by:
-
Chris Wilson authored
No longer is knowing how much of the GTT (both mappable aperture and beyond) relevant, and the output clutters the real information - that is how many objects are allocated and bound (and by who) so that we can quickly grasp if there is a leak. v2: Relent, and rename pinned to indicate display only. Since the display objects are semi-static and are of variable size, they are the interesting objects to watch over time for aperture leaking. The other pins are either static (such as the scratch page) or very short lived (such as execbuf) and not part of the precious GGTT. Signed-off-by:
-
Chris Wilson authored
Only those objects pinned to the display have semi-permanent pins of a global nature (other pins are transient within their local vm). Simplify i915_gem_pinned to only show the pertinent information about the pinned objects within the GGTT. v2: i915_gem_gtt_info is still shared with debugfs/i915_gem_gtt, rename i915_gem_pinned to i915_gem_pin_display to better reflect its contents Signed-off-by:
-
Chris Wilson authored
These two files (i915_gem_active, i915_gem_inactive) no longer give pertinent information since active/inactive tracking is per-vm and so we need the information per-vm. They are obsolete so remove them. Signed-off-by:
-
Chris Wilson authored
With execlists, we have context objects everywhere, not just RCS. So store them for post-mortem debugging. This also has a secondary effect of removing one more unsafe list iteration with using preserved state from the hanging request. And now we can cross-reference the request's context state with that loaded by the GPU. Signed-off-by:
-
Chris Wilson authored
When capturing the error state, we do not need to know about every address space - just those that are related to the error. We know which context is active at the time, therefore we know which VM are implicated in the error. We can then restrict the VM which we report to the relevant subset. v2: s/i/count_active/ (and similar) Rewrite label generation for "Buffers" Signed-off-by: -
Chris Wilson authored
Not only does it make for good documentation and debugging aide, but it is also vital for when we want to unwind requests - such as when throwing away an incomplete request. Signed-off-by:
-
Chris Wilson authored
For convenience when debugging user issues show the autotuning RPS parameters in debugfs/i915_rps_boost_info. v2: Refine the presentation v3: Style Signed-off-by:
David Weinehall <david.weinehall@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1471250973-31277-1-git-send-email-chris@chris-wilson.co.uk
-
Daniel Vetter authored
Backmerge because too many conflicts, and also we need to get at the latest struct fence patches from Gustavo. Requested by Chris Wilson. Signed-off-by: -
git://anongit.freedesktop.org/drm-intelDave Airlie authored
- refactor ddi buffer programming a bit (Ville) - large-scale renaming to untangle naming in the gem code (Chris) - rework vma/active tracking for accurately reaping idle mappings of shared objects (Chris) - misc dp sst/mst probing corner case fixes (Ville) - tons of cleanup&tunings all around in gem - lockless (rcu-protected) request lookup, plus use it everywhere for non(b)locking waits (Chris) - pipe crc debugfs fixes (Rodrigo) - random fixes all over * tag 'drm-intel-next-2016-08-08' of git://anongit.freedesktop.org/drm-intel: (222 commits) drm/i915: Update DRIVER_DATE to 20160808 drm/i915: fix aliasing_ppgtt leak drm/i915: Update comment before i915_spin_request drm/i915: Use drm official vblank_no_hw_counter callback. drm/i915: Fix copy_to_user usage for pipe_crc Revert "drm/i915: Track active streams also for DP SST" drm/i915: fix WaInsertDummyPushConstPs drm/i915: Assert that the request hasn't been retired drm/i915: Repack fence tiling mode and stride into a single integer drm/i915: Document and reject invalid tiling modes drm/i915: Remove locking for get_tiling drm/i915: Remove pinned check from madvise ioctl drm/i915: Reduce locking inside swfinish ioctl drm/i915: Remove (struct_mutex) locking for busy-ioctl drm/i915: Remove (struct_mutex) locking for wait-ioctl drm/i915: Do a nonblocking wait first in pread/pwrite drm/i915: Remove unused no-shrinker-steal drm/i915: Tidy generation of the GTT mmap offset drm/i915/shrinker: Wait before acquiring struct_mutex under oom drm/i915: Simplify do_idling() (Ironlake vt-d w/a) ...
-
git://anongit.freedesktop.org/drm-intelDave Airlie authored
- more fence destaging and cleanup (Gustavo&Sumit) - DRIVER_LEGACY to untangle from DRIVER_MODESET - drm_mm refactor (Chris) - fbdev-less compile fies - clipped plane src/dst rects (Ville) - + a few mediatek patches that build on top of that (Bibby+Daniel) - small stuff all over really * tag 'topic/drm-misc-2016-08-12' of git://anongit.freedesktop.org/drm-intel: (43 commits) dma-buf/fence: kerneldoc: remove spurious section header dma-buf/fence: kerneldoc: remove unused struct members Revert "gpu: drm: omapdrm: dss-of: add missing of_node_put after calling of_parse_phandle" drm: Protect fb_defio in drivers with CONFIG_KMS_FBDEV_EMULATION drm/radeon|amgpu: Make fbdev emulation optional drm/vmwgfx: select CONFIG_FB drm: Remove superflous linux/fb.h includes drm/fb-helper: Add a dummy remove_conflicting_framebuffers dma-buf/sync_file: only enable fence signalling on poll() Documentation: add doc for sync_file_get_fence() dma-buf/sync_file: add sync_file_get_fence() dma-buf/sync_file: refactor fence storage in struct sync_file dma-buf/fence-array: add fence_is_array() drm/dp_helper: Rate limit timeout errors from drm_dp_i2c_do_msg() drm/dp_helper: Print first error received on failure in drm_dp_dpcd_access() drm: Add ratelimited versions of the DRM_DEBUG* macros drm: Make sure drm_vblank_no_hw_counter isn't abused drm/mediatek: Fix mtk_atomic_complete for runtime_pm drm/mediatek: plane: Use FB's format's cpp to compute x offset drm/mediatek: plane: Merge mtk_plane_enable into mtk_plane_atomic_update ...
-
git://git.pengutronix.de/git/pza/linuxDave Airlie authored
imx-drm updates and encoder atomic_mode_set helper callback - add pixel clock and DE polarity configuration from device tree using display timing bindings for parallel and LVDS output - cleanup/remove trivial functions - cleanup and fixes in preparation for capture support - add atomic_mode_set helper and use it in imx-ldb - this is an alternative to the encoder mode_set callback that passes the crtc and connector state instead of just the mode. It allows drivers to get information from the attached connector without having to iterate over all connectors - add drm_bridge support to imx-ldb, for bridges attached via LVDS * tag 'imx-drm-next-2016-08-12' of git://git.pengutronix.de/git/pza/linux: drm/imx-ldb: Add support to drm-bridge drm/imx: imx-ldb: use encoder atomic_mode_set callback drm/atomic-helper: Add atomic_mode_set helper callback drm/imx: Remove imx_drm_handle_vblank() gpu: ipu-v3: Add missing IDMAC channel names gpu: ipu-v3: rename CSI client device gpu: ipu-v3: Fix IRT usage gpu: ipu-v3: Fix CSI data format for 16-bit media bus formats gpu: ipu-v3: set correct full sensor frame for PAL/NTSC gpu: ipu-v3: Add VDI input IDMAC channels gpu: ipu-v3: Add ipu_get_num() gpu: ipu-cpmem: Add ipu_cpmem_get_burstsize() gpu: ipu-cpmem: Add ipu_cpmem_set_uv_offset() drm/imx: Remove imx_drm_crtc_id() drm/imx: Remove imx_drm_crtc_vblank_get/_put() drm/imx: convey the pixelclk-active and de-active flags from DT to the ipu-di driver drm: add a helper function to extract 'de-active' and 'pixelclk-active' from DT
-
git://git.pengutronix.de/git/pza/linuxDave Airlie authored
mediatek-drm maintainers and gamma correction - add MAINTAINERS entry for mediatek-drm driver - add support for AAL and GAMMA engines - hook up gamma correction LUT - add support for temporal dithering to OD and GAMMA engines * tag 'mediatek-drm-next-2016-08-12' of git://git.pengutronix.de/git/pza/linux: drm/mediatek: set mt8173 dithering function drm/mediatek: Add gamma correction. drm/mediatek: Add GAMMA engine basic function drm/mediatek: Add AAL engine basic function drm: mediatek: add Maintainers entry for Mediatek DRM drivers
-
