- 09 Jan, 2020 5 commits
-
-
Chris Wilson authored
Since we now allow the intel_context_unpin() to run unserialised, we risk our operations under the intel_context_lock_pinned() being run as the context is unpinned (and thus invalidating our state). We can atomically acquire the pin, testing to see if it is pinned in the process, thus ensuring that the state remains consistent during the course of the whole operation. Fixes: 84135022 ("drm/i915/gt: Drop mutex serialisation between context pin/unpin") Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by:
Mika Kuoppala <mika.kuoppala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200109085142.871563-1-chris@chris-wilson.co.uk
-
Tvrtko Ursulin authored
This reverts commit 08fff7ae. For some yet unexplained reason not having this improves stability of some media workloads. Promise is that the media hang will be root caused properly and in the meantime absence of this workaround is unlikely to cause problems. Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Sudeep Dutt <sudeep.dutt@intel.com> Cc: Francesco Balestrieri <francesco.balestrieri@intel.com> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Cc: Tony Ye <tony.ye@intel.com> Acked-by:
-
Chris Wilson authored
Access through the GGTT (iomap) into the vma does require the device to be awake. However, we often take the i915_vma_pin_iomap() as an early preparatory step that is long before we use the iomap. Asserting that the device is awake at pin time does not protect us, and is merely a nuisance. Signed-off-by:
-
Chris Wilson authored
If we have no fence and desire no fence on the vma, return before we try and take the vm->mutex. Signed-off-by:
-
Chris Wilson authored
intel_timeline_enter() has been decoupled from intel_timeline_pin() and both enter/exit & pin/unpin are allowed [read expected] to run concurrently with one another. The assertion that they had better not is stale. Closes: https://gitlab.freedesktop.org/drm/intel/issues/940 References: a6edbca7 ("drm/i915/gt: Close race between engine_park and intel_gt_retire_requests") Signed-off-by:
-
- 08 Jan, 2020 3 commits
-
-
José Roberto de Souza authored
Capturing the restrictions of the BSpec pages bellow: SKL and CNL do not support MST in DDI E, DDI E only support 2 lanes and it is mostly used to support a 4 lanes eDP panel together with DDI A. ICL's DDI E support MST just like other ports but DDI A is still eDP and MIPI only. TGL supports MST in any DDI, including DDI A but TGL has it's own ddi_pre_enable_dp function already without any warning. [ 215.579791] ------------[ cut here ]------------ [ 215.579794] WARN_ON(is_mst && (port == PORT_A || port == PORT_E)) [ 215.579875] WARNING: CPU: 0 PID: 268 at drivers/gpu/drm/i915/display/intel_ddi.c:3576 intel_ddi_pre_enable+0x124/0xea0 [i915] [ 215.579878] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic i915 btusb btrtl btbcm btintel bluetooth prime_numbers snd_hda_intel snd_intel_dspcfg snd_hda_codec e1000e snd_hwdep snd_hda_core asix mei_hdcp cdc_ether x86_pkg_temp_thermal mei_me snd_pcm r8152 coretemp usbnet mei crct10dif_pclmul mii ptp ecdh_generic crc32_pclmul i2c_i801 ecc pps_core ghash_clmulni_intel thunderbolt [ 215.579905] CPU: 0 PID: 268 Comm: kworker/0:2 Tainted: G W 5.4.0-rc8-zeh+ #1307 [ 215.579907] Hardware name: Intel Corporation Ice Lake Client Platform/IceLake U DDR4 SODIMM PD RVP TLC, BIOS ICLSFWR1.R00.3201.A00.1905140358 05/14/2019 [ 215.579912] Workqueue: events_long drm_dp_mst_link_probe_work [ 215.579975] RIP: 0010:intel_ddi_pre_enable+0x124/0xea0 [i915] [ 215.579978] Code: ff 8b 7c 24 10 89 44 24 30 85 ff 74 1f f7 44 24 18 fb ff ff ff 75 15 48 c7 c6 98 fa 48 a0 48 c7 c7 d3 df 4a a0 e8 cf d5 d0 e0 <0f> 0b 0f b6 4c 24 2c 41 8b b5 04 06 00 00 4c 89 e7 41 0f b6 95 0c [ 215.579980] RSP: 0018:ffffc90001a5f990 EFLAGS: 00010286 [ 215.579984] RAX: 0000000000000000 RBX: ffff88848356a000 RCX: 0000000000000000 [ 215.579986] RDX: 0000000000001df1 RSI: ffff88849340c998 RDI: ffffffff821489c5 [ 215.579989] RBP: ffff88848356a000 R08: 00000000c021a419 R09: 0000000000000000 [ 215.579991] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88848356a118 [ 215.579994] R13: ffff88847f39c000 R14: ffff88847fe70000 R15: ffff88848356a000 [ 215.579996] FS: 0000000000000000(0000) GS:ffff88849f800000(0000) knlGS:0000000000000000 [ 215.579999] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 215.580001] CR2: 000055d3d5a26bc0 CR3: 0000000480ba6005 CR4: 0000000000760ef0 [ 215.580004] PKRU: 55555554 [ 215.580006] Call Trace: [ 215.580014] ? drm_dp_mst_topology_put_port+0x6f/0x130 [ 215.580072] intel_mst_pre_enable_dp+0x14b/0x170 [i915] [ 215.580129] intel_encoders_pre_enable+0x76/0x90 [i915] [ 215.580191] haswell_crtc_enable+0x84/0x880 [i915] [ 215.580266] intel_update_crtc+0x1e4/0x200 [i915] [ 215.580333] skl_commit_modeset_enables+0x287/0x420 [i915] [ 215.580405] intel_atomic_commit_tail+0x332/0x14e0 [i915] [ 215.580410] ? queue_work_on+0x41/0x70 [ 215.580489] intel_atomic_commit+0x31e/0x350 [i915] [ 215.580500] drm_client_modeset_commit_atomic+0x18b/0x220 [ 215.580523] drm_client_modeset_commit_force+0x4d/0x180 [ 215.580531] drm_fb_helper_restore_fbdev_mode_unlocked+0x46/0xa0 [ 215.580538] drm_fb_helper_set_par+0x27/0x50 [ 215.580543] drm_fb_helper_hotplug_event.part.0+0xa7/0xc0 [ 215.580549] drm_kms_helper_hotplug_event+0x21/0x30 [ 215.580553] process_one_work+0x25b/0x5b0 [ 215.580566] worker_thread+0x4b/0x3b0 [ 215.580578] kthread+0x100/0x140 [ 215.580581] ? process_one_work+0x5b0/0x5b0 [ 215.580585] ? kthread_park+0x80/0x80 [ 215.580591] ret_from_fork+0x24/0x50 [ 215.580603] irq event stamp: 1393930 [ 215.580606] hardirqs last enabled at (1393929): [<ffffffff8112a013>] vprintk_emit+0x143/0x330 [ 215.580609] hardirqs last disabled at (1393930): [<ffffffff81001cfa>] trace_hardirqs_off_thunk+0x1a/0x20 [ 215.580613] softirqs last enabled at (1393434): [<ffffffff81c00389>] __do_softirq+0x389/0x47f [ 215.580618] softirqs last disabled at (1393423): [<ffffffff810b7199>] irq_exit+0xa9/0xc0 [ 215.580621] ---[ end trace afd44ea9caa6373e ]--- BSpec: 4217 BSpec: 14004 BSpec: 20584 BSpec: 50583 Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by:
José Roberto de Souza <jose.souza@intel.com> Reviewed-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200107170922.153612-2-jose.souza@intel.com
-
José Roberto de Souza authored
Talked with HW team and this is a left over, driver should not program clockgating, mg or dekel firmware is reponsible for any clockgating programing. Also removing the register and bits definition related to clockgating. v2: Added WARN_ON v3: Only calling icl_phy_set_clock_gating() on intel_ddi_pre_enable_hdmi for GEN11 v4: ICL should also not program clockgating (thanks Matt for catching this) BSpec issue: 20885 BSpec: 49292 BSpec: 21735 Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Reviewed-by:
Matt Roper <matthew.d.roper@intel.com> Signed-off-by:
-
José Roberto de Souza authored
Recent improvements in the state tracking in i915 caused PSR to not be enabled when reusing firmware/BIOS modeset, this is due to all initial commits returning ealier in intel_atomic_check() as needs_modeset() is always false. To fix that here forcing the state compute phase in CRTC that is driving the eDP that supports PSR once. Enable or disable PSR do not require a fullmodeset, so user will still experience glitch free boot process plus the power savings that PSR brings. It was tried to set mode_changed in intel_initial_commit() but at this point the connectors are not registered causing a crash when computing encoder state. v2: - removed function return - change arguments to match intel_hdcp_atomic_check v3: - replaced drm includes in intel_psr.h by forward declaration(Jani) Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=112253 Reported-by: <s.zharkoff@gmail.com> Cc: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com> Cc: Jani Nikula <jani.nikula@intel.com> Signed-off-by:
Gwan-gyeong Mun <gwan-gyeong.mun@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200106152128.195171-1-jose.souza@intel.com
-
- 07 Jan, 2020 13 commits
-
-
Chris Wilson authored
It is highly unlikely, but still conceivable, that we submit a context with the same GGTT address as last active on the HW. In this case, with a matching LRCA, the HW would not restore the new context image causing a potential violation of our context isolation. Signed-off-by:
-
Matthew Auld authored
Attempt to split i915_gem_gtt.[ch] into more manageable chunks. Suggested-by:
Matthew Auld <matthew.auld@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
-
Chris Wilson authored
In order to avoid a double cleanup on error, take ownership of engine->release past the point of no [error] return. Reported-by:
-
José Roberto de Souza authored
This loop was added directly to intel_atomic_check() to be used by all other features that have external pipe dependencies, so using it and removing intel_atomic_check_synced_crtcs(). After this changes is_trans_port_sync_master() it not used anywhere, so removing it. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Manasi Navare <manasi.d.navare@intel.com> Signed-off-by:
Manasi Navare <manasi.d.navare@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200106142823.145260-1-jose.souza@intel.com
-
Dan Carpenter authored
There is a cut and paste bug so we return the wrong error code. Fixes: a603f5bd ("drm/i915/dp: Make sure all tiled connectors get added to the state with full modeset") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
-
YueHaibing authored
Fix build error: ./drivers/gpu/drm/i915/selftests/i915_random.h: In function i915_prandom_u32_max_state: ./drivers/gpu/drm/i915/selftests/i915_random.h:48:23: error: implicit declaration of function mul_u32_u32; did you mean mul_u64_u32_div? [-Werror=implicit-function-declaration] return upper_32_bits(mul_u32_u32(prandom_u32_state(state), ep_ro)); Reported-by:
Hulk Robot <hulkci@huawei.com> Fixes: 7ce5b685 ("drm/i915/selftests: Use mul_u32_u32() for 32b x 32b -> 64b result") Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Reviewed-by:
-
Dhinakaran Pandiyan authored
Detect the modifier corresponding to media compression to enable display decompression for YUV and xRGB packed formats. A new modifier is added so that the driver can distinguish between media and render compressed buffers. Unlike render decompression, plane 6 and plane 7 do not support media decompression. v2: Fix checkpatch warnings on code style (Lucas) From DK: Separate modifier array for planes that cannot decompress media (Ville) v3: Support planar formats v4: Switch plane order v5: - Use format block descriptors to get CCS subsampling calculation right everywhere. - Extend the plane state normal view array to accommodate 4 color planes. - Use helpers to convert between main and CCS planes. v6: Add missing packed YUV formats to the MC format list. (Yang) v7: Align UV planes to tile-row size. Cc: Nanley G Chery <nanley.g.chery@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Yang A Shi <yang.a.shi@intel.com> Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com> Signed-off-by:
Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Signed-off-by:
Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by:
Imre Deak <imre.deak@intel.com> Reviewed-by:
Radhakrishna Sripada <radhakrishna.sripada@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191231233756.18753-8-imre.deak@intel.com
-
Dhinakaran Pandiyan authored
addfb() uAPI has supported four planes for a while now, make format_info compatible with that. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Mika Kahola <mika.kahola@intel.com> Cc: dri-devel@lists.freedesktop.org Signed-off-by:
Mika Kahola <mika.kahola@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191231233756.18753-7-imre.deak@intel.com
-
Dhinakaran Pandiyan authored
Gen-12 display can decompress surfaces compressed by the media engine, add a new modifier as the driver needs to know the surface was compressed by the media or render engine. v2: Update code comment describing the color plane order for YUV semiplanar formats. Cc: Nanley G Chery <nanley.g.chery@intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Mika Kahola <mika.kahola@intel.com> Cc: dri-devel@lists.freedesktop.org Signed-off-by: -
Imre Deak authored
As intel_fb_plane_get_subsampling() returns the subsampling factor wrt. its main plane, for a CCS plane we need to apply both the main and the CCS plane's subsampling factor on the FB's dimensions to get the CCS plane's dimensions. Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Cc: Mika Kahola <mika.kahola@intel.com> Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Matt Roper <matthew.d.roper@intel.com> Signed-off-by:
-
Imre Deak authored
Print a debug message if the FB plane[0] offset is not 0 as expected, to help understainding an add FB IOCTL fail. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Mika Kahola <mika.kahola@intel.com> Signed-off-by:
-
Imre Deak authored
Currently the GGTT offset of a UV plane in a semiplanar YUV FB is tile size (4kB) aligned. I noticed, that enforcing only this alignment leads oddly to random memory corruptions on TGL while scanning out Y-tiled FBs. This issue can be easily reproduced with a UV plane offset that is not aligned to the plane's tile row size. Some experiments showed the correct alignment to be tile row size indeed. This also makes sense, since the de-tiling fence created for the object - with its own stride and so "left" and "right" edge - applies to all the planes in the FB, so each tile row of all planes should be tile row aligned. In fact BSpec requires this alignment since SKL. On SKL we may enforce this due to the AUX plane x,y coords check, but on ICL and TGL we don't. For now enforce this only on TGL; I can follow up with any necessary change for ICL after more tests. BSpec requires a stricter alignment for linear UV planes too (kind of a tile row alignment), but it's unclear whether that's really needed (couldn't be explained with the de-tiling fence as above) and enforcing that could break existing user space; so avoid that too for now until more tests. v2: - Clarify the commit log wrt. the address space the alignment applies to. (Chris) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
-
Imre Deak authored
At least one framebuffer plane on TGL - the UV plane of YUV semiplanar FBs - requires a non-power-of-2 alignment, so add support for this. This new alignment restriction applies only to an offset within an FB, so the GEM buffer itself containing the FB must still be power-of-2 aligned. Add a check for this (in practice plane 0, since the plane 0 offset must be 0). v2: - Fix WARN check for alignment=0. v3: - Return error for alignment programming bugs. (Chris) Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
-
- 06 Jan, 2020 8 commits
-
-
Kai Vehmanen authored
Revert changes done in commit f6ec9483 ("drm/i915: extend audio CDCLK>=2*BCLK constraint to more platforms"). Audio drivers communicate with i915 over HDA bus multiple times during system boot-up and each of these transactions result in matching get_power/put_power calls to i915, and depending on the platform, a modeset change causing visible flicker. GLK is the only platform with minimum CDCLK significantly lower than BCLK, and thus for GLK setting a higher CDCLK is mandatory. For other platforms, minimum CDCLK is close but below 2*BCLK (e.g. on ICL, CDCLK=176.4kHz with BCLK=96kHz). Spec-wise the constraint should be set, but in practise no communication errors have been reported and the downside if set is the flicker observed at boot-time. Revert to old behaviour until better mechanism to manage probe-time clocks is available. The full CDCLK>=2*BCLK constraint is still enforced at pipe enable time in intel_crtc_compute_min_cdclk(). Bugzilla: https://gitlab.freedesktop.org/drm/intel/issues/913 Fixes: f6ec9483 ("drm/i915: extend audio CDCLK>=2*BCLK constraint to more platforms") Signed-off-by:
Kai Vehmanen <kai.vehmanen@linux.intel.com> Reviewed-by:
-
Chris Wilson authored
Be sure to initialise the uabi_instance on the virtual engine to the special invalid value, just in case we ever peek at it from the uAPI. Reported-by:
-
Chris Wilson authored
Put memset_p to use to clear the array of pointers used for tracking the ELSP. Signed-off-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200106114234.2529613-6-chris@chris-wilson.co.uk
-
Chris Wilson authored
The last remaining reason for serialising the pin/unpin of the intel_context is to ensure that our preallocated wakerefs are not consumed too early (i.e. the unpin of the previous phase does not emit the idle barriers for this phase before we even submit). All of the other operations within the context pin/unpin are supposed to be atomic... Therefore, we can reduce the serialisation to being just on the i915_active.preallocated_barriers itself and drop the nested pin_mutex from intel_context_unpin(). Signed-off-by:
-
Chris Wilson authored
Convert the few remaining GEM_TRACE() used for debugging over to the appropriate GT_TRACE or RQ_TRACE. References: 639f2f24 ("drm/i915: Introduce new macros for tracing") Signed-off-by:
-
Chris Wilson authored
As we already have a flags field buried within i915_request, reuse it! Signed-off-by:
-
Chris Wilson authored
Avoid spinning indefinitely waiting for the request to be submitted, and instead apply a timeout. A secondary benefit is that the error message will show which suspect is blocked. Signed-off-by:
-
Chris Wilson authored
The local var does not need the __user as it exists on the kernel stack and not a pointer into the __user address space. drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:989:9: warning: dereference of noderef expression drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c:990:13: warning: dereference of noderef expression Signed-off-by:
-
- 05 Jan, 2020 2 commits
-
-
Ramalingam C authored
When LMEM is supported, dumb buffer preferred to be created from LMEM. v2: Parameters are reshuffled. [Chris] v3: s/region_id/mem_type v4: use the i915_gem_object_create_region [chris] Signed-off-by:
Ramalingam C <ramalingam.c@intel.com> cc: Matthew Auld <matthew.auld@intel.com> Reviewed-by:
-
Ramalingam C authored
Lookup function to retrieve the pointer to a memory region of a mem_type. v2: for_each_memory_region is used. Signed-off-by:
-
- 04 Jan, 2020 3 commits
-
-
Chris Wilson authored
Check that the user writes into their mmap are visible on the GPU. Signed-off-by:
Abdiel Janulgue <abdiel.janulgue@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20200103204137.2131004-3-chris@chris-wilson.co.uk
-
Abdiel Janulgue authored
Instead of testing individually our new fault handlers, iterate over all memory regions and test all from one interface. Signed-off-by:
-
Abdiel Janulgue authored
Local memory objects are similar to our usual scatterlist, but instead of using the struct page stored therein, we need to use the sg->dma_address. Signed-off-by:
-
- 03 Jan, 2020 6 commits
-
-
Chris Wilson authored
Include the types used by the headers to they can be compiled standalone. Signed-off-by:
-
Chris Wilson authored
Move the definition of the igt_atomic_section[] into a C file, leaving the declaration in the header so as not to upset headertest! Signed-off-by:
-
Chris Wilson authored
Create a vmap for discontinguous lmem objects to support i915_gem_object_pin_map(). v2: Offset io address by region.start for fake-lmem Signed-off-by:
-
Chris Wilson authored
Keep scrubbing the kernel_context image with poison before we reset it in order to demonstrate that we will be resilient in the case where it is accidentally overwritten on idle. Suggested-by:
-
Chris Wilson authored
Before we idle, on parking, we switch to the kernel context such that we have a scratch context loaded while the GPU idle, protecting any precious user state. Be paranoid and assume that the idle state may have been trashed, and reset the kernel_context image after idling. Signed-off-by:
-
Chris Wilson authored
We leave the kernel_context on the HW as we suspend (and while idle). There is no guarantee that is complete in memory, so we try to inhibit restoration from the kernel_context. Reinforce the inhibition by scrubbing the context. Signed-off-by:
-
