- 14 Oct, 2016 24 commits
-
-
Michał Winiarski authored
Since "Dynamic page table allocations" were introduced, our page tables can grow (being dynamically allocated) with address space range usage. Unfortunately, their lifetime is bound to vm. This is not a huge problem when we're not using softpin - drm_mm is creating an upper bound on used range by causing addresses for our VMAs to eventually be reused. With softpin, long lived contexts can drain the system out of memory even with a single "small" object. For example: bo = bo_alloc(size); while(true) offset += size; exec(bo, offset); Will cause us to create new allocations until all memory in the system is used for tracking GPU pages (even though almost all PTEs in this vm are pointing to scratch). Let's free unused page tables in clear_range to prevent this - if no entries are used, we can safely free it and return this information to the caller (so that higher-level entry is pointing to scratch). v2: Document return value and free semantics (Joonas) v3: No newlines in vars block (Joonas) v4: Drop redundant local 'reduce' variable v5: Handle CI fail with enable_ppgtt=2 Cc: Michel Thierry <michel.thierry@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Reviewed-by:Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by:
Michał Winiarski <michal.winiarski@intel.com> Signed-off-by:
-
Michał Winiarski authored
Let's use more top-down approach, where each gen8_ppgtt_clear_* function is responsible for clearing the struct passed as an argument and calling relevant clear_range functions on lower-level tables. Doing this rather than operating on PTE ranges makes the implementation of shrinking page tables quite simple. v2: Drop min when calculating num_entries, no negation in 48b ppgtt check, no newlines in vars block (Joonas) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Michel Thierry <michel.thierry@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by:
-
Michał Winiarski authored
We never used any invalid ptes, those were put in place for a possibility of doing gpu faults. However our batchbuffers are not restricted in length, so everything needs to be pointing to something and thus out-of-bounds is pointing to scratch. Remove the valid flag as it is always true. v2: Expand commit msg, patch reorder (Mika) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michel Thierry <michel.thierry@intel.com> Reviewed-by:
Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by:
-
Imre Deak authored
Currently the display INIT power domain disabling/enabling happens in a mismatched way in the suspend/resume_early hooks respectively. This can leave display power wells incorrectly disabled in the resume hook if the suspend sequence is aborted for some reason resulting in the suspend/resume hooks getting called but the suspend_late/resume_early hooks being skipped. In particular this change fixes "Unclaimed read from register 0x1e1204" on BYT/BSW triggered from i915_drm_resume()-> intel_pps_unlock_regs_wa() when suspending with /sys/power/pm_test set to devices. Fixes: 85e90679 ("drm/i915: disable power wells on suspend") Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: David Weinehall <david.weinehall@intel.com> Signed-off-by:
Imre Deak <imre.deak@intel.com> Reviewed-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476358446-11621-1-git-send-email-imre.deak@intel.com
-
Tvrtko Ursulin authored
Saves 1416 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by:
David Weinehall <david.weinehall@linux.intel.com> Acked-by:
Daniel Vetter <daniel.vetter@ffwll.ch> Acked-by:
Jani Nikula <jani.nikula@linux.intel.com> Acked-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476352990-2504-1-git-send-email-tvrtko.ursulin@linux.intel.com
-
Tvrtko Ursulin authored
Saves 968 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) v3: Rebase. Signed-off-by:
-
Tvrtko Ursulin authored
Saves 944 bytes of .rodata strings and 128 bytes of .text. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 864 bytes of .rodata strings and ~100 of .text. v2: Add parantheses around dev_priv. (Ville Syrjala) v3: Rebase. Signed-off-by:
-
Tvrtko Ursulin authored
Saves 472 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 472 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 1392 bytes of .rodata strings. Also change a few function/macro prototypes in i915_gem_gtt.c from dev to dev_priv where it made more sense to do so. v2: Add parantheses around dev_priv. (Ville Syrjala) v3: Mention function prototype changes. (David Weinehall) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 1016 bytes of .rodata strings and couple hundred of .text. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 1320 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 2432 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 1808 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 848 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) v3: Rebase. Signed-off-by:
-
Tvrtko Ursulin authored
Saves 4472 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 944 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
Saves 1520 bytes of .rodata strings. Signed-off-by:
-
Tvrtko Ursulin authored
Saves 960 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
More .rodata string saving by avoid __I915__ magic inside WARNs. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
This saves 1872 bytes of .rodata strings. v2: * Rebase. * Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Tvrtko Ursulin authored
This saves 3248 bytes of .rodata strings. v2: Add parantheses around dev_priv. (Ville Syrjala) Signed-off-by:
-
Akash Goel authored
With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by:
Akash Goel <akash.goel@intel.com> Reviewed-by:
-
- 13 Oct, 2016 10 commits
-
-
Chris Wilson authored
gen4/vlv/chv all use the same bits in pipestat to enable the vblank interrupt, so they can share the same callbacks to enable/disable. Signed-off-by:
-
Dan Carpenter authored
We want to read 3 bytes here, but because the parenthesis are in the wrong place we instead read: sizeof(intel_dp->edp_dpcd) == sizeof(intel_dp->edp_dpcd) which is one byte. Fixes: fe5a66f9 ("drm/i915: Read PSR caps/intermediate freqs/etc. only once on eDP") Reviewed-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Chris Wilson authored
If the user requests a mappable binding to the global GTT, we will first unbind an existing mapping if it doesn't match. We will unbind even if there is no possibility that the object can fit in the mappable aperture. This may lead to a ping-pong migration of the object, for example igt/gem_exec_big. v2: Comment upon the reasoning, or lack thereof!, behind the choice of magic numbers. Testcase: igt/gem_exec_big Signed-off-by:
-
Chris Wilson authored
'\n' is supposed to be at the end of the line, not in the middle. Fixes: cdb324bd ("drm/i915: Show bounds of active request in the ring...") Signed-off-by:
-
Chris Wilson authored
Mika wanted to know what requests were pending at the time of a hang as we now track which requests we have submitted to the hardware. Signed-off-by:
-
Tvrtko Ursulin authored
Make struct video_levels and struct tv_mode use data types of sufficient width to save approximately one kilobyte in the .rodata section. v2: Do not align struct members. (Jani Nikula, Joonas Lahtinen) Signed-off-by:
-
Tvrtko Ursulin authored
Use types of more appropriate size in struct intel_watermark_params to save 512 bytes of .rodata. Signed-off-by:
-
Tvrtko Ursulin authored
Pack the struct _sdvo_cmd_name to save 736 bytes of .rodata. This is fine since the name pointers are used only for debug. Signed-off-by:
-
Tvrtko Ursulin authored
unsigned long is too wide - use smaller types in struct cxsr_latency to save 800-something bytes of .rodata. v2: All data even fits in u16 for even more saving. (Ville Syrjala) v3: Move bitfields to the end of the struct. (Joonas Lahtinen) Signed-off-by:
-
Imre Deak authored
Currently resuming on HSW from S3 pm_test/devices state leads to an unrecoverable GPU hang. Resetting the GPU during suspend fixes this. For a full S3 cycle this change only means the reset happens earlier (before reaching S3). For S4 the reset will happen now both during the freeze and quiesce phases, which is a benefit since it will guarantee that the GPU is idle before creating and loading the hibernation image. Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Suggested-by:
-
- 12 Oct, 2016 6 commits
-
-
Chris Wilson authored
Treat a framebuffer reference with the same priority as an active reference whilst shrinking. Framebuffers are likely to be reused and typically cost more to migrate to and from GPU memory (on LLC architectures we need to clflush), so defer the temptation to purge them during a kswapd run until we have run out of cheap buffers. Signed-off-by:
John Harrison <john.c.harrison@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161012124824.23521-1-chris@chris-wilson.co.uk
-
Chris Wilson authored
The current meaning of whether an object has a GGTT vma is very ill-defined (and note we don't check for any partials either), it just means that at some point it was in the GGTT but it may not be now. The information we really care about here is whether it is taking up precious mappable aperture space. This is the obj->fault_mappable flag. We have a redundant long form reprinting of this information, so remove that in favour of the compact flag. Signed-off-by:
-
Chris Wilson authored
During rpm resume we restore the fences, but we do not have the protection of struct_mutex. This rules out updating the activity tracking on the fences, and requires us to rely on the rpm as the serialisation barrier instead. [ 350.298052] [drm:intel_runtime_resume [i915]] Resuming device [ 350.308606] [ 350.310520] =============================== [ 350.315560] [ INFO: suspicious RCU usage. ] [ 350.320554] 4.8.0-rc8-bsw-rapl+ #3133 Tainted: G U W [ 350.327208] ------------------------------- [ 350.331977] ../drivers/gpu/drm/i915/i915_gem_request.h:371 suspicious rcu_dereference_protected() usage! [ 350.342619] [ 350.342619] other info that might help us debug this: [ 350.342619] [ 350.351593] [ 350.351593] rcu_scheduler_active = 1, debug_locks = 0 [ 350.358952] 3 locks held by Xorg/320: [ 350.363077] #0: (&dev->mode_config.mutex){+.+.+.}, at: [<ffffffffa030589c>] drm_modeset_lock_all+0x3c/0xd0 [drm] [ 350.375162] #1: (crtc_ww_class_acquire){+.+.+.}, at: [<ffffffffa03058a6>] drm_modeset_lock_all+0x46/0xd0 [drm] [ 350.387022] #2: (crtc_ww_class_mutex){+.+.+.}, at: [<ffffffffa0305056>] drm_modeset_lock+0x36/0x110 [drm] [ 350.398236] [ 350.398236] stack backtrace: [ 350.403196] CPU: 1 PID: 320 Comm: Xorg Tainted: G U W 4.8.0-rc8-bsw-rapl+ #3133 [ 350.412457] Hardware name: Intel Corporation CHERRYVIEW C0 PLATFORM/Braswell CRB, BIOS BRAS.X64.X088.R00.1510270350 10/27/2015 [ 350.425212] 0000000000000000 ffff8801680a78c8 ffffffff81332187 ffff88016c5c5000 [ 350.433611] 0000000000000001 ffff8801680a78f8 ffffffff810ca6da ffff88016cc8b0f0 [ 350.442012] ffff88016cc80000 ffff88016cc80000 ffff880177ad0000 ffff8801680a7948 [ 350.450409] Call Trace: [ 350.453165] [<ffffffff81332187>] dump_stack+0x67/0x90 [ 350.458931] [<ffffffff810ca6da>] lockdep_rcu_suspicious+0xea/0x120 [ 350.466002] [<ffffffffa039e8dd>] fence_update+0xbd/0x670 [i915] [ 350.472766] [<ffffffffa039efe2>] i915_gem_restore_fences+0x52/0x70 [i915] [ 350.480496] [<ffffffffa0368f42>] vlv_resume_prepare+0x72/0x570 [i915] [ 350.487839] [<ffffffffa0369802>] intel_runtime_resume+0x102/0x210 [i915] [ 350.495442] [<ffffffff8137f26f>] pci_pm_runtime_resume+0x7f/0xb0 [ 350.502274] [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40 [ 350.509883] [<ffffffff814401c5>] __rpm_callback+0x35/0x70 [ 350.516037] [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40 [ 350.523646] [<ffffffff81440224>] rpm_callback+0x24/0x80 [ 350.529604] [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40 [ 350.537212] [<ffffffff814417bd>] rpm_resume+0x4ad/0x740 [ 350.543161] [<ffffffff81441aa1>] __pm_runtime_resume+0x51/0x80 [ 350.549824] [<ffffffffa03889c8>] intel_runtime_pm_get+0x28/0x90 [i915] [ 350.557265] [<ffffffffa0388a53>] intel_display_power_get+0x23/0x50 [i915] [ 350.565001] [<ffffffffa03ef23d>] intel_atomic_commit_tail+0xdfd/0x10b0 [i915] [ 350.573106] [<ffffffffa034b2e9>] ? drm_atomic_helper_swap_state+0x159/0x300 [drm_kms_helper] [ 350.582659] [<ffffffff81615091>] ? _raw_spin_unlock+0x31/0x50 [ 350.589205] [<ffffffffa034b2e9>] ? drm_atomic_helper_swap_state+0x159/0x300 [drm_kms_helper] [ 350.598787] [<ffffffffa03ef8a5>] intel_atomic_commit+0x3b5/0x500 [i915] [ 350.606319] [<ffffffffa03061dc>] ? drm_atomic_set_crtc_for_connector+0xcc/0x100 [drm] [ 350.615209] [<ffffffffa0306b49>] drm_atomic_commit+0x49/0x50 [drm] [ 350.622242] [<ffffffffa034dee8>] drm_atomic_helper_set_config+0x88/0xc0 [drm_kms_helper] [ 350.631419] [<ffffffffa02f94ac>] drm_mode_set_config_internal+0x6c/0x120 [drm] [ 350.639623] [<ffffffffa02fa94c>] drm_mode_setcrtc+0x22c/0x4d0 [drm] [ 350.646760] [<ffffffffa02f0f19>] drm_ioctl+0x209/0x460 [drm] [ 350.653217] [<ffffffffa02fa720>] ? drm_mode_getcrtc+0x150/0x150 [drm] [ 350.660536] [<ffffffff810c984a>] ? __lock_is_held+0x4a/0x70 [ 350.666885] [<ffffffff81202303>] do_vfs_ioctl+0x93/0x6b0 [ 350.672939] [<ffffffff8120f843>] ? __fget+0x113/0x200 [ 350.678797] [<ffffffff8120f735>] ? __fget+0x5/0x200 [ 350.684361] [<ffffffff81202964>] SyS_ioctl+0x44/0x80 [ 350.690030] [<ffffffff81001deb>] do_syscall_64+0x5b/0x120 [ 350.696184] [<ffffffff81615ada>] entry_SYSCALL64_slow_path+0x25/0x25 Note we also have to remember the lesson from commit 4fc788f5 ("drm/i915: Flush delayed fence releases after reset") where we have to flush any changes to the fence on restore. v2: Replace call to release user mmaps with an assertion that they have already been zapped. Fixes: 49ef5294 ("drm/i915: Move fence tracking from object to vma") Reported-by: -
Chris Wilson authored
Our error states are quickly growing, pinning kernel memory with them. The majority of the space is taken up by the error objects. These compress well using zlib and without decode are mostly meaningless, so encoding them does not hinder quickly parsing the error state for familiarity. v2: Make the zlib dependency optional Signed-off-by:
-
Chris Wilson authored
Leave all the pretty printing to userspace and simplify the error capture to only have a single common object printer. It makes the kernel code more compact, and the refactoring allows us to apply more complex transformations like compressing the output. Signed-off-by:
-
Chris Wilson authored
Since the GTT provides universal access to any GPU page, we can use it to reduce our plethora of read methods to just one. It also has the important characteristic of being exactly what the GPU sees - if there are incoherency problems, seeing the batch as executed (rather than as trapped inside the cpu cache) is important. Signed-off-by:
-
