- 05 Feb, 2019 3 commits
-
-
Tvrtko Ursulin authored
Timeline barrier allows serialization between different timelines. After calling i915_timeline_set_barrier with a request, all following submissions on this timeline will be set up as depending on this request, or barrier. Once the barrier has been completed it automatically gets cleared and things continue as normal. This facility will be used by the upcoming context SSEU code. v2: * Assert barrier has been retired on timeline_fini. (Chris Wilson) * Fix mock_timeline. v3: * Improved comment language. (Chris Wilson) v4: * Maintain ordering with previous barriers set on the timeline. v5: * Rebase. Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Suggested-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by:
-
Lionel Landwerlin authored
If some of the contexts submitting workloads to the GPU have been configured to shutdown slices/subslices, we might loose the NOA configurations written in the NOA muxes. One possible solution to this problem is to reprogram the NOA muxes when we switch to a new context. We initially tried this in the workaround batchbuffer but some concerns where raised about the cost of reprogramming at every context switch. This solution is also not without consequences from the userspace point of view. Reprogramming of the muxes can only happen once the powergating configuration has changed (which happens after context switch). This means for a window of time during the recording, counters recorded by the OA unit might be invalid. This requires userspace dealing with OA reports to discard the invalid values. Minimizing the reprogramming could be implemented by tracking of the last programmed configuration somewhere in GGTT and use MI_PREDICATE to discard some of the programming commands, but the command streamer would still have to parse all the MI_LRI instructions in the workaround batchbuffer. Another solution, which this change implements, is to simply disregard the user requested configuration for the period of time when i915/perf is active. On most platforms there are no issues with this apart from a performance penality for some media workloads that benefit from running on a partially powergated GPU. We already prevent RC6 from affecting the programming so it doesn't sound completely unreasonable to hold on powergating for the same reason. On Icelake however there would a functional problem if the slices not- containing the VME block were left enabled with a running media workload which explicitly disabled them. To avoid a GPU hang in this case, on Icelake we lock the enablement to only slices which contain VME blocks. Downside is that it means degraded GPU performance when OA is active but there is no known alternative solution for this. v2: Leave RPCS programming in intel_lrc.c (Lionel) v3: Update for s/union intel_sseu/struct intel_sseu/ (Lionel) More to_intel_context() (Tvrtko) s/dev_priv/i915/ (Tvrtko) Tvrtko Ursulin: v4: * Rebase for make_rpcs changes. v5: * Apply OA restriction from make_rpcs directly. v6: * Rebase for context image setup changes. v7: * Move stream assignment before metric enable. v8-9: * Rebase. v10: * Squashed with ICL support patch. Bspec: 21140 Co-developed-by:Lionel Landwerlin <lionel.g.landwerlin@intel.com> Signed-off-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190205095032.22673-2-tvrtko.ursulin@linux.intel.com
-
Lionel Landwerlin authored
We want to expose the ability to reconfigure the slices, subslice and eu per context and per engine. To facilitate that, store the current configuration on the context for each engine, which is initially set to the device default upon creation. v2: record sseu configuration per context & engine (Chris) v3: introduce the i915_gem_context_sseu to store powergating programming, sseu_dev_info has grown quite a bit (Lionel) v4: rename i915_gem_sseu into intel_sseu (Chris) use to_intel_context() (Chris) v5: More to_intel_context() (Tvrtko) Switch intel_sseu from union to struct (Tvrtko) Move context default sseu in existing loop (Chris) v6: s/intel_sseu_from_device_sseu/intel_device_default_sseu/ (Tvrtko) Tvrtko Ursulin: v7: * Pass intel_sseu by pointer instead of value to make_rpcs. * Rebase for make_rpcs changes. v8: * Rebase for RPCS edit on pin. v9: * Rebase for context image setup changes. v10: * Rename dev_priv to i915. (Chris Wilson) v11: * Rebase. v12: * Rebase for IS_GEN changes. v13: * Rebase for RUNTIME_INFO. v14: * Rebase for intel_context_init. v15: * Rebase for drm-tip changes. v16: * Moved struct intel_sseu definition to i915_gem_context.h. Signed-off-by:
-
- 04 Feb, 2019 2 commits
-
-
Chris Wilson authored
Limit the NEWCLIENT boost to only give its small priority boost to fresh clients only that have no dependencies. The idea for using NEWCLIENT boosting, commit b16c7651 ("drm/i915: Priority boost for new clients"), is that short-lived streams are often interactive and require lower latency -- and that by executing those ahead of the long running hogs, the short-lived clients do little to interfere with the system throughput by virtue of their short-lived nature. However, we were only considering the client's own timeline for determining whether or not it was a fresh stream. This allowed for compositors to wake up before their vblank and bump all of its client streams. However, in testing with media-bench this results in chaining all cooperating contexts together preventing us from being able to reorder contexts to reduce bubbles (pipeline stalls), overall increasing latency, and reducing system throughput. The exact opposite of our intent. The compromise of applying the NEWCLIENT boost to strictly fresh clients (that do not wait upon anything else) should maintain the "real-time response under load" characteristics of FQ_CODEL, without locking together the long chains of dependencies across the system. References: b16c7651 ("drm/i915: Priority boost for new clients") Signed-off-by:
-
Chris Wilson authored
When first enabling preemption, we hesitated from making it a free-for-all where every higher priority client would force a preempt-to-idle cycle and take over from all lower priority clients. We hesitated because we were uncertain just how well preemption would work in practice, whether the preemption latency itself would detract from the latency gains for higher priority tasks and whether it would work at all. Since introducing preemption, we have been enabling it for more common tasks, even giving normal clients a small preemptive boost when they first start (to aide fairness and improve interactivity). Now lets take one step further and give permission for all normal (priority:0) clients to preempt any idle (priority:<0) task so that users running long compute jobs do not overly impact other jobs (i.e. their desktop) and the system remains responsive under such idle loads. References: f6322edd ("drm/i915/preemption: Allow preemption between submission ports") References: b16c7651 ("drm/i915: Priority boost for new clients") Signed-off-by:
Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Reviewed-by:
-
- 02 Feb, 2019 1 commit
-
-
Rodrigo Vivi authored
Signed-off-by:Rodrigo Vivi <rodrigo.vivi@intel.com>
-
- 31 Jan, 2019 1 commit
-
-
Rodrigo Vivi authored
While cross checking PCI IDs from Intel Media SDK and kernel Dmitry noticed this gap. So we checked the spec and this new ID had been recently added. v2: Adding new H_GT1 entry to i915_pci.c (Jose) Reported-by: Dmitry Rogozhkin<dmitry.v.rogozhkin@intel.com> Cc: Dmitry Rogozhkin<dmitry.v.rogozhkin@intel.com> Cc: José Roberto de Souza <jose.souza@intel.com> Signed-off-by:
José Roberto de Souza <jose.souza@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190201235049.27206-1-rodrigo.vivi@intel.com
-
- 01 Feb, 2019 5 commits
-
-
https://github.com/intel/gvt-linuxRodrigo Vivi authored
gvt-next-2019-02-01 - new VFIO EDID region support (Henry) Signed-off-by:
-
Hans de Goede authored
We really want to have fastboot enabled by default to avoid an ugly modeset during boot. Currently we are enabling fastboot by default on gen9+ (Skylake and newer). The intention is to enable it on older generations after it has seen more testing on gen9+. VLV and CHV devices are still being sold in stores today, as such it is desirable to also enable fastboot by default on these now. I've extensively tested fastboot=1 support on over 50 different Bay- and Cherry-Trail devices. Testing DSI and eDP panels as well as HDMI output (and even DP over Type-C on one device). All 50 devices work fine with fastboot=1. On 2 devices their DSI panel turns black as soon as the i915 driver loads when fastboot=0, so having fastboot enabled is required for these 2 to work properly (for lack of a better fix). Signed-off-by:
Hans de Goede <hdegoede@redhat.com> Reviewed-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190129142237.8684-1-hdegoede@redhat.com
-
Talha Nassar authored
Enables blend optimization for floating point RTs This restores the workaround that was reverted in c358514b ("Revert "drm/i915/icl: WaEnableFloatBlendOptimization""). The revert was due to the register write seemingly not sticking, but the HW team has confirmed that this is because the register is WO and that the workaround is indeed required. Here the wa is added with a mask of 0 since the register is WO. References: https://hsdes.intel.com/resource/1408134172 References: https://bugs.freedesktop.org/show_bug.cgi?id=107338 Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by:
Talha Nassar <talha.nassar@intel.com> Reviewed-by:
-
Tvrtko Ursulin authored
No functional or code size change - just notice we can compact the source by re-using a single helper for adding workarounds. Signed-off-by:
-
Tvrtko Ursulin authored
Top comment in intel_workarounds.c says common code should come first so lets respect that. Also, by moving the common code together opportunities to reduce duplication will become more obvious. Signed-off-by:
-
- 31 Jan, 2019 9 commits
-
-
Imre Deak authored
VBT may include incorrect information about the presence of port F. Work around this on SKUs where we know the port is not present. v2: - Fix IS_ICL_WITH_PORT_F, so it's useable from any context. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108915 Cc: Mika Kahola <mika.kahola@intel.com> Cc: Jani Nikula <jani.nikula@linux.intel.com> Signed-off-by:
Imre Deak <imre.deak@intel.com> Reviewed-by:
Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20181220155211.31456-1-imre.deak@intel.com
-
Imre Deak authored
We can't safely probe Type C ports, whether they are a legacy or a USB/Thunderbolt DP Alternate Type C port. This would require performing the TypeC connect sequence - as described by the specification - but that may have unwanted side-effects. These side-effects include at least - without completeness - timeouts during AUX power well enabling and subsequent PLL enabling errors. To safely identify these ports we really need VBT, which has the proper flag for this (ddi_vbt_port_info::supports_typec_usb, supports_tbt). Based on the above disable Type C ports if we can't load VBT for some reason. v2: - Notice that we disable TypeC ports completely and simplify accordingly (Jose). - Add code comment explaining why we disabled the ports. (Jani) Cc: Jani Nikula <jani.nikula@intel.com> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Jose Roberto de Souza <jose.souza@intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by:
-
Ville Syrjälä authored
commit 5b0bd14d ("drm/i915/icl: keep track of unused pll while looping") inadvertently (I presume) changed the code to pick the last unused dpll rather than the first unused one like we did before. While there should most likely be no harm in changing the order let's change back just to avoid a change in the behaviour. At least it might reduce the confusion when staring at logs (took me a while to figure out why DPLL1 being picked over DPLL0 when the latter was most definitely available). Cc: Lucas De Marchi <lucas.demarchi@intel.com> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190130181359.20693-1-ville.syrjala@linux.intel.comReviewed-by:
Paulo Zanoni <paulo.r.zanoni@intel.com> Acked-by:
Lucas De Marchi <lucas.demarchi@intel.com>
-
Ville Syrjälä authored
The code managing the dbuf slices is borked and needs some real work to fix. In the meantime let's just stop using the second slice. v2: Drop the change to intel_enabled_dbuf_slices_num() (Mahesh) Cc: Mahesh Kumar <mahesh1.sh.kumar@gmail.com> Reviewed-by: Imre Deak <imre.deak@intel.com> #v1 Signed-off-by:
Mahesh Kumar <mahesh1.sh.kumar@gmail.com>
-
Hang Yuan authored
Implement VFIO EDID region for vgpu. Support EDID blob update and notify guest on link state change via hotplug event. v3: move struct edid_region to kvmgt.c <zhenyu> v2: add EDID sanity check and size update <zhenyu> Tested-by:
Gerd Hoffmann <kraxel@redhat.com> Reviewed-by:
Hang Yuan <hang.yuan@linux.intel.com> Signed-off-by:
Zhenyu Wang <zhenyuw@linux.intel.com>
-
Hang Yuan authored
Add function to emulate hotplug interrupt for SKL/KBL platforms Signed-off-by:
-
Hang Yuan authored
These functions will get default resolution according to vgpu type. Signed-off-by:
-
Matt Roper authored
Use of the new DRM_COLOR_LUT_NON_DECREASING test was a bit over-zealous; it doesn't actually need to be applied to the degamma on "bdw-style" platforms. Likewise, we overlooked the fact that CHV should have that test applied to the gamma LUT as well as the degamma LUT. Rather than adding more complicated platform checking to intel_color_check(), let's just store the appropriate set of LUT validation flags for each platform in the intel_device_info structure. v2: - Shuffle around LUT size tests so that the hardware-specific tests won't be applied to legacy gamma tables. (Ville) - Add a debug message so that it will be easier to understand why an atomic transaction involving incorrectly-sized LUT's got rejected by the driver. v3: - Switch size_t's to int's. (Ville) Fixes: 85e2d61e ("drm/i915: Validate userspace-provided color management LUT's (v4)") References: https://lists.freedesktop.org/archives/intel-gfx/2019-January/187634.html Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
Matt Roper <matthew.d.roper@intel.com> Reviewed-by:
-
Matt Roper authored
We don't yet allow userspace to control the CRTC background color, but we should manually program the color to black to ensure the BIOS didn't leave us with some other color. We should also set the pipe gamma and pipe CSC bits so that the background color goes through the same color management transformations that a plane with black pixels would. v2: Rename register to SKL_BOTTOM_COLOR to more closely follow bspec naming. (Ville) Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
-
- 30 Jan, 2019 10 commits
-
-
Rodrigo Vivi authored
Signed-off-by: -
Ville Syrjälä authored
IS_GLK||IS_BXT == IS_GEN9_LP Signed-off-by:
-
Ville Syrjälä authored
0*whatever==0 so this check is pointless. Remove it. Signed-off-by:
-
Ville Syrjälä authored
The spec doesn't use a definite article in front of SAGV. The rules regarding articles and initialisms are super fuzzy, but at least to my ears it sounds much more natural to not have the article. Perhaps because I tend to pronounce it as "sag-vee" instead of spelling out the letters one at a time. Actually I might still prefer to leave out the article if I did spell them out. Signed-off-by:
-
Ville Syrjälä authored
skl_needs_memory_bw_wa() doesn't look at the passed in state at all. Possibly it should, but for now let's make life simpler by just passing in dev_priv. Signed-off-by:
-
Ville Syrjälä authored
On icl+ bspec tells us to calculate a separate minimum ddb allocation from the blocks watermark. Both have to be checked against the actual ddb allocation, but since we do things the other way around we'll just calculat the minimum acceptable ddb allocation by taking the maximum of the two values. We'll also replace the memcmp() with a full trawl over the the watermarks so that it'll ignore the min_ddb_alloc because we can't directly read that out from the hw. I suppose we could reconstruct it from the other values, but I was too lazy to do that now. Signed-off-by:
-
Ville Syrjälä authored
Bspec says we have to reject the watermark if it's >= the ddb allocation. Fix the code to reject the == case as it should. For transition watermarks we can just use >=, for the rest we'll do +1 when calculating the minimum ddb allocation size. Signed-off-by:
-
Ville Syrjälä authored
The spec used to say "8bpp" which someone took to mean 8 bytes per pixel when in fact it was supposed to be 8 bits per pixel. The spec has been updated to make it more clear now. Fix the code to match. Signed-off-by:
Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> Reviewed-by:
-
Ville Syrjälä authored
I thought we could remove all the early latency==0 checks and rely on skl_wm_method{1,2}() checking for it. But skl_compute_plane_wm() applies a bunch of workarounds to bump up the latency before calling those guys so clearly it won't end up doing the right thing. Also not sure if the calculations based on the method1/2 results are safe agaisnt overflows so it might not work all that well in any case. Let's put the early check back. Signed-off-by: -
Ville Syrjälä authored
On glk+ the level 0 lines watermark actually matters. Do not ignore it. And while at it let's change things so that we always program a consistnet 0 to the register when the lines watermarks is ignored by the hardware. Signed-off-by:
-
- 29 Jan, 2019 9 commits
-
-
Lucas De Marchi authored
Instead of looping again on the range of plls, just keep track of one unused one and use it later. Signed-off-by:
-
Lucas De Marchi authored
We should not pass DPLL_ID_ICL_DPLL0 or DPLL_ID_ICL_DPLL1 to this function because the path is only taken for non-combophy ports. Let the warning trigger if improper value is given. While at it, rename the function to match the register name we are trying to program. v2: fix typo in comment Signed-off-by:
-
Lucas De Marchi authored
Even if we don't have the correct clock and get a warning, we should not skip the return. v2: improve commit message (from Joonas) Fixes: 1fa11ee2 ("drm/i915/icl: start adding the TBT pll") Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: <stable@vger.kernel.org> # v4.19+ Signed-off-by:
Mika Kahola <mika.kahola@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190125222444.19926-3-lucas.demarchi@intel.com
-
Lucas De Marchi authored
Fix the TODO leftover in the code by changing the argument in MG_PLL macros. The MG_PLL ids used to access the register values can be converted from tc_port rather than port. All these registers can use the TC port to calculate the right offsets because they are only available for TC ports. The range (PORT_C onwards) may not be stable and change from platform to platform. So by using the TC id directly we avoid having to check for the platform in the "leaf functions" and thus passing dev_priv around. The helper functions were also renamed to use "tc" as prefix to make them more generic. v2: Improve commit message and fix checkpatch warning (from Paulo) Signed-off-by:
-
Chris Wilson authored
Missed breadcrumb detection is defunct due to the tight coupling with dma_fence signaling and the myriad ways we may signal fences from everywhere but from an interrupt, i.e. we frequently signal a fence before we even see its interrupt. This means that even if we miss an interrupt for a fence, it still is signaled before our breadcrumb hangcheck fires, so simplify the breadcrumb hangchecking by moving it into the GPU hangcheck and forgo fake interrupts. Signed-off-by:
-
Chris Wilson authored
A few years ago, see commit 688e6c72 ("drm/i915: Slaughter the thundering i915_wait_request herd"), the issue of handling multiple clients waiting in parallel was brought to our attention. The requirement was that every client should be woken immediately upon its request being signaled, without incurring any cpu overhead. To handle certain fragility of our hw meant that we could not do a simple check inside the irq handler (some generations required almost unbounded delays before we could be sure of seqno coherency) and so request completion checking required delegation. Before commit 688e6c72, the solution was simple. Every client waiting on a request would be woken on every interrupt and each would do a heavyweight check to see if their request was complete. Commit 688e6c72 introduced an rbtree so that only the earliest waiter on the global timeline would woken, and would wake the next and so on. (Along with various complications to handle requests being reordered along the global timeline, and also a requirement for kthread to provide a delegate for fence signaling that had no process context.) The global rbtree depends on knowing the execution timeline (and global seqno). Without knowing that order, we must instead check all contexts queued to the HW to see which may have advanced. We trim that list by only checking queued contexts that are being waited on, but still we keep a list of all active contexts and their active signalers that we inspect from inside the irq handler. By moving the waiters onto the fence signal list, we can combine the client wakeup with the dma_fence signaling (a dramatic reduction in complexity, but does require the HW being coherent, the seqno must be visible from the cpu before the interrupt is raised - we keep a timer backup just in case). Having previously fixed all the issues with irq-seqno serialisation (by inserting delays onto the GPU after each request instead of random delays on the CPU after each interrupt), we can rely on the seqno state to perfom direct wakeups from the interrupt handler. This allows us to preserve our single context switch behaviour of the current routine, with the only downside that we lose the RT priority sorting of wakeups. In general, direct wakeup latency of multiple clients is about the same (about 10% better in most cases) with a reduction in total CPU time spent in the waiter (about 20-50% depending on gen). Average herd behaviour is improved, but at the cost of not delegating wakeups on task_prio. v2: Capture fence signaling state for error state and add comments to warm even the most cold of hearts. v3: Check if the request is still active before busywaiting v4: Reduce the amount of pointer misdirection with list_for_each_safe and using a local i915_request variable inside the loops v5: Add a missing pluralisation to a purely informative selftest message. References: 688e6c72 ("drm/i915: Slaughter the thundering i915_wait_request herd") Signed-off-by:
-
Chris Wilson authored
The global seqno is defunct and so we have no meaningful indicator of forward progress for an engine. You need to listen to the request signaling tracepoints instead. Signed-off-by:
-
Ville Syrjälä authored
Let's switch the pipe into interlaced mode and switch off the TV encoder vertical filter if the pipe vdisplay matches the TV YSIZE exactly. While I didn't measure it I presume this might reduce the power consumption a little bit, and the pixel rate is halved as the pipe will now fetching in interlaced mode rather than in progressive mode (effectively the same difference as between IF-ID vs. PF-ID pfit modes on more modern hardware) so a bit easier on the memory bandwidth. Cc: Imre Deak <imre.deak@intel.com> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch>
-
Ville Syrjälä authored
intel_tv_mode_to_mode() assumes the pipe will be in progressive fetch mode, and thus when programming the pipe into interlaced mode we have to halve the calculated dotclock to get the correct field duration. This becomes more important when we start to program the pipe into interlaced mode on i965gm as we depend on the timestamps to get accurate frame counter values. Withot halving the clock our guesstimated frame counter would tick at twice the expected speed. Cc: Imre Deak <imre.deak@intel.com> Fixes: 690157f0 ("drm/i915/tv: Fix >1024 modes on gen3") Signed-off-by:
-
