The bspec is confusing on the nature of the upper 32bits of the LRC
descriptor. Once upon a time, it said that it uses the upper 32b to
decide if it should perform a lite-restore, and so we must ensure that
each unique context submitted to HW is given a unique CCID [for the
duration of it being on the HW]. Currently, this is achieved by using
a small circular tag, and assigning every context submitted to HW a
new id. However, this tag is being cleared on repinning an inflight
context such that we end up re-using the 0 tag for multiple contexts.
To avoid accidentally clearing the CCID in the upper 32bits of the LRC
descriptor, split the descriptor into two dwords so we can update the
GGTT address separately from the CCID.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1796
Fixes: 2935ed5339 ("drm/i915: Remove logical HW ID")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: <stable@vger.kernel.org> # v5.5+
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200428184751.11257-1-chris@chris-wilson.co.uk
(cherry picked from commit 2632f174a2)
(cherry picked from commit a4b70fcc587860f4b972f68217d8ebebe295ec15)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
The bspec is confusing on the nature of the upper 32bits of the LRC
descriptor. Once upon a time, it said that it uses the upper 32b to
decide if it should perform a lite-restore, and so we must ensure that
each unique context submitted to HW is given a unique CCID [for the
duration of it being on the HW]. Currently, this is achieved by using
a small circular tag, and assigning every context submitted to HW a
new id. However, this tag is being cleared on repinning an inflight
context such that we end up re-using the 0 tag for multiple contexts.
To avoid accidentally clearing the CCID in the upper 32bits of the LRC
descriptor, split the descriptor into two dwords so we can update the
GGTT address separately from the CCID.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1796
Fixes: 2935ed5339 ("drm/i915: Remove logical HW ID")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: <stable@vger.kernel.org> # v5.5+
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200428184751.11257-1-chris@chris-wilson.co.uk
The history of i915_vma_close() is confusing, as is its use. As the
lifetime of the i915_vma is currently bounded by the object it is
attached to, we needed a means of identify when a vma was no longer in
use by userspace (via the user's fd). This is further complicated by
that only ppgtt vma should be closed at the user's behest, as the ggtt
were always shared.
Now that we attach the vma to a lut on the user's context, the open
count does indicate how many unique and open context/vm are referencing
this vma from the user. As such, we can and should just use the
open_count to track when the vma is still in use by userspace.
It's a poor man's replacement for reference counting.
Closes: https://gitlab.freedesktop.org/drm/intel/issues/1193
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Andi Shyti <andi.shyti@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200422190558.30509-1-chris@chris-wilson.co.uk
UAPI Changes:
- drm/i915/perf: introduce global sseu pinning
Allow userspace to request at perf/OA open full SSEU configuration
on the system to be able to benchmark 3D workloads, at the cost of not
being able to run media workloads. (Lionel)
Userspace changes: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4021
- drm/i915/perf: add new open param to configure polling of OA buffer
Let application choose how often the OA buffer should be checked on
the CPU side for data availability for choosig between CPU overhead
and realtime nature of data.
Userspace changes: https://patchwork.freedesktop.org/series/74655/
(i915 perf recorder is a tool to capture i915 perf data for viewing
in GPUVis.)
- drm/i915/perf: remove generated code
Removal of the machine generated perf/OA test configurations from i915.
Used by Mesa v17.1-18.0, and shortly replaced by userspace supplied OA
configurations. Removal of configs causes affected Mesa versions to
fall back to earlier kernel behaviour (potentially missing metrics).
(Lionel)
Cross-subsystem Changes:
- Backmerge of drm-next
- Includes tag 'topic/phy-compliance-2020-04-08' from
git://anongit.freedesktop.org/drm/drm-misc
Driver Changes:
- Fix for GitLab issue #27: Support 5k tiled dual DP display on SKL (Ville)
- Fix https://github.com/thesofproject/linux/issues/1719: Broken audio after
S3 resume on JSL platforms. (Kai)
- Add new Tigerlake PCI IDs (Swathi D.)
- Add missing Tigerlake W/As (Matt R.)
- Extended Wa_2006604312 to EHL (Matt A)
- Add DPCD link_rate quirk for Apple 15" MBP 2017 (v3) (Mario)
- Make Wa_14010229206 apply to all Tigerlake steppings (Swathi d)
- Extend hotplug detect retry on TypeC connectors to 5 seconds (Imre)
- Yield the timeslice if caught waiting on a user semaphore (Chris)
- Limit the residual W/A batch to Haswell due to instability on IVB/BYT (Chris)
- TBT AUX should use TC power well ops on Tigerlake (Matt R)
- Update PMINTRMSK holding fw to make it effective for RPS (Francisco, Chris)
- Add YUV444 packed format support for skl+ (Stanislav)
- Invalidate OA TLB when closing perf stream to avoid corruption (Umesh)
- HDCP: fix Ri prime check done during link check (Oliver)
- Rearm heartbeat on sysfs interval change (Chris)
- Fix crtc nv12 etc. plane bitmasks for DPMS off (Ville)
- Treat idling as a RPS downclock event (Chris)
- Leave rps->cur_freq on unpark (Chris)
- Ignore short pulse when EDP panel powered off (Anshuman)
- Keep the engine awake until the next jiffie, to avoid ping-pong on
moderate load (Chris)
- Select the deepest available parking mode for rc6 on IVB (Chris)
- Optimizations to direct submission execlist path (Chris)
- Avoid NULL pointer dereference at intel_read_infoframe() (Chris)
- Fix mode private_flags comparison at atomic_check (Uma, Ville)
- Use forced codec wake on all gen9+ platforms (Kai)
- Schedule oa_config after modifying the contexts (Chris, Lionel)
- Explicitly reset both reg and context runtime on GPU reset (Chris)
- Don't enable DDI IO power on a TypeC port in TBT mode (Imre)
- Fixes to TGL, ICL and EHL vswing tables (Jose)
- Fill all the unused space in the GGTT (Chris, imre)
- Ignore readonly failures when updating relocs (Chris)
- Attempt to find free space earlier for non-pinned VMAs (Chris)
- Only wait for GPU activity before unbinding a GGTT fence (Chris)
- Avoid data loss on small userspace perf OA polling (Ashutosh)
- Watch out for unevictable nodes during eviction (Matt A)
- Reinforce the barrier after GTT updates for Ironlake (Chris)
- Convert various parts of driver to use drm_device based logging (Wambui, Jani)
- Avoid dereferencing already closed context for engine (Chris)
- Enable non-contiguous pipe fusing (Anshuman)
- Add HW readout of Gamma LUT on ICL (Swati S.)
- Use explicit flag to mark unreachable intel_context (Chris)
- Cancel a hung context if already closed (Chris)
- Add DP VSC/HDR SDP data structures and write routines (Gwan-gyeong)
- Report context-is-closed prior to pinning at execbuf (Chris)
- Mark timeline->cacheline as destroyed after rcu grace period (Chris)
- Avoid live-lock with i915_vma_parked() (Chris)
- Avoid gem_context->mutex for simple vma lookup (Chris)
- Rely on direct submission to the queue (Chris)
- Configure DSI transcoder to operate in TE GATE command mode (Vandita)
- Add DI vblank calculation for command mode (Vandita)
- Disable periodic command mode if programmed by GOP (Vandita)
- Use private flags to indicate TE in cmd mode (Vandita)
- Make fences a nice-to-have for FBC on GEN9+ (Jose)
- Fix work queuing issue with mixed virtual engine/physical engine
submissions (Chris)
- Drop final few uses of drm_i915_private.engine (Chris)
- Return early after MISSING_CASE for write_dp_sdp (Chris)
- Include port sync state in the state dump (Ville)
- ELSP workaround switching back to a completed context (Chris)
- Include priority info in trace_ports (Chris)
- Allow for different modes of interruptible i915_active_wait (Chris)
- Split eb_vma into its own allocation (Chris)
- Don't read perf head/tail pointers outside critical section (Lionel)
- Pause CS flow before execlists reset (Chris)
- Make fence revocation unequivocal (Chris)
- Drop cached obj->bind_count (Chris)
- Peek at the next submission for error interrupts (Chris)
- Utilize rcu iteration of context engines (Chris)
- Keep a per-engine request pool for power management ops (Chris)
- Refactor port sync code into normal modeset flow (Ville)
- Check current i915_vma.pin_count status first on unbind (Chris)
- Free request pool from virtual engines (Chris)
- Flush all the reloc_gpu batch (Chris)
- Make exclusive awaits on i915_active optional and allow async waits (Chris)
- Wait until the context is finally retired before releasing engines (Chris)
- Prefer '%ps' for printing function symbol names (Chris)
- Allow setting generic data pointer on intel GT debugfs (Andi)
- Constify DP link computation code more (Ville)
- Simplify MST master transcoder computation (Ville)
- Move TRANS_DDI_FUNC_CTL2 programming where it belongs (Ville)
- Move icl_get_trans_port_sync_config() into the DDI code (Ville)
- Add definitions for VRR registers and bits (Aditya)
- Refactor hardware fence code (Chris)
- Start passing latency as parameter to WM calculation (Stanislav)
- Kernel selftest and debug tracing improvements (Matt A, Chris, Mika)
- Fixes to CI found corner cases and lockdep splats (Chris)
- Overall fixes and refactoring to GEM code (Chris)
- Overall fixes and refactoring to display code (Ville)
- GuC/HuC code improvements (Daniele, Michal Wa)
- Static code checker fixes (Nathan, Ville, Colin, Chris)
- Fix spelling mistake (Chen)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200417111548.GA15033@jlahtine-desk.ger.corp.intel.com
Software is not expected to populate engine context except when using
restore inhibit bit or golden state to initialize it for the first time.
Therefore, if a newly submitted guest context is the same as the last
shadowed one, no need to populate its engine context from guest again.
Currently using lrca + ring_context_gpa to identify whether two guest
contexts are the same.
The reason of why context id is not included as an identifier is that
i915 recently changed the code and context id is only unique for a
context when OA is enabled. And when OA is on, context id is generated
based on lrca. Therefore, in that case, if two contexts are of the same
lrca, they have identical context ids as well.
(This patch also works with old guest kernel like 4.20.)
for guest context, if its ggtt entry is modified after last context
shadowing, it is also deemed as not the same context as last shadowed one.
v7:
-removed local variable "valid". use the one in s->last_ctx diretly
v6:
-change type of lrca of last ctx to be u32. as currently it's all
protected by vgpu lock (Kevin Tian)
-reset valid of last ctx to false once it needs to be repopulated before
population completes successfully (Kevin Tian)
v5:
-merge all 3 patches into one patch (Zhenyu Wang)
v4:
- split the series into 3 patches.
- don't turn on optimization until last patch in this series (Kevin Tian)
- define lrca to be atomic in this patch rather than update its type in
the second patch (Kevin Tian)
v3: updated commit message to describe engine context and context id
clearly (Kevin Tian)
v2: rebased to 5.6.0-rc4+Signed-off-by: Yan Zhao <yan.y.zhao@intel.com>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Cc: Kevin Tian <kevin.tian@intel.com>
Suggested-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Yan Zhao <yan.y.zhao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200417091334.32628-1-yan.y.zhao@intel.com
UAPI Changes:
On i915 we have a new UAPI to allow userspace to specify CS ring buffer size on
construction (I915_CONTEXT_PARAM_RINGSIZE) and also new sysfs entries exposing
various engine properties
GVT Changes:
VFIO edid getting expanded to all platforms and a big cleanup around attr
group, unused vblank complete, kvmgt, Intel engine and dev_priv usages.
i915 Changes:
- new UAPI to allow userspace to specify CS ring buffer size on construction
(I915_CONTEXT_PARAM_RINGSIZE) - (Chris)
- New sysfs entries exposing various engine properties (Chris)
- Tiger Lake is out of require_force_probe protection (Jose)
- Changes in many places around active requests, reset and heartbeat (Chris)
- Stop assigning drm-dev_private pointer (Jani)
- Many code refactor in many places, including intel_modeset_init,
increasing use of intel_uncore_*, vgpu, and gvt stuff (Jani)
- Fixes around display pipe iterators (Anshuman)
- Tigerlake enabling work (Matt Ropper, Matt Atwood, Ville, Lucas, Daniele,
Jose, Anusha, Vivek, Swathi, Caz. Kai)
- Code clean-up like reducing use of drm/i915_drv.h, removing unused
registers, removing garbage warns, and some other code polishing (Jani, Lucas,
Ville)
- Selftests fixes, improvements and additions (Chris, Dan, Aditya, Matt Auld)
- Fix plane possible_crtcs bit mask (Anshuman)
- Fixes and cleanup on GLK pre production identification and w/a (Ville)
- Fix display orientation on few cases (Hans, Ville)
- dbuf clean-up and improvements for slice arrays handling (Ville)
- Improvement around min cdclk calculation (Stanislav)
- Fixes and refactor around display PLLs (Imre)
- Other execlists and perf fixes (Chris)
- Documentation fixes (Jani, Chris)
- Fix build issue (Anshuman)
- Many more fixes around the locking mechanisms (Chris)
- Other fixes and debugability info around preemption (Chris, Tvrtko)
- Add mechanism to submit a context WA on ring submission (Mika)
- Clear all Eu/L3 resitual context (Prathap)
- More changes around local memory (Abdiel, Matt, Chris)
- Fix RPS (Chris)
- DP MST fix (Lyude)
- Display FBC fixes (Jose, RK)
- debugfs cleanup (Tvrtko)
- More convertion towards drm_debive based loggin (Wambui, Ram)
- Avoid potential buffer overflow (Takashi)
- Ice Lake and Elkhart Lake workarounds (Matt Roper)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200314001535.GA2969344@intel.com
According to the vbt provided by GVT-g, the display output type should
be display port, instead of DVI. Currently, GVT-g display model only
supports display port emulation, not DVI or HDMI. This patch fixes this
by telling guest i915 the right output type supported by GVT-g.
v3: make port_b and port_c use dp type connector too. (Zhenyu)
v2: refine the commit message. (Zhenyu)
Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200317093615.10538-1-tina.zhang@intel.com
Those regs are added in order to slove the following complains:
[70811.201818] gvt: vgpu(1) Invalid FORCE_NONPRIV write 2341 at offset 24d8
[70811.201825] gvt: vgpu(1) Invalid FORCE_NONPRIV write 2351 at offset 24dc
[70811.201831] gvt: vgpu(1) Invalid FORCE_NONPRIV write 10000d82 at offset 24e0
[70811.201837] gvt: vgpu(1) Invalid FORCE_NONPRIV write 10064844 at offset 24e4
So solve them by adding the required regs to the whitelist.
Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200302083130.17831-1-tina.zhang@intel.com
The emulated vbt doesn't tell its size correctly. According to the
intel_vbt_defs.h, vbt_header.vbt_size should the size of VBT (VBT Header,
BDB Header and data blocks), and bdb_header.bdb_size should be the size
of BDB (BDB Header and data blocks).
This patch fixes the issue and lets vbt provided by GVT-g pass the guest
i915's sanity test.
v2: refine the commit message. (Zhenyu)
Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200305131600.29640-1-tina.zhang@intel.com
Drm specific drm_WARN* calls include device information in the
backtrace, so we know what device the warnings originate from.
Covert all the calls of WARN* with device specific drm_WARN*
variants in functions where drm_device struct pointer is readily
available.
The conversion was done automatically with below coccinelle semantic
patch. checkpatch errors/warnings are fixed manually.
@@
identifier func, T;
@@
func(struct intel_vgpu *T,...) {
+struct drm_i915_private *i915 = T->gvt->dev_priv;
<+...
(
-WARN(
+drm_WARN(&i915->drm,
...)
|
-WARN_ON(
+drm_WARN_ON(&i915->drm,
...)
|
-WARN_ONCE(
+drm_WARN_ONCE(&i915->drm,
...)
|
-WARN_ON_ONCE(
+drm_WARN_ON_ONCE(&i915->drm,
...)
)
...+>
}
Signed-off-by: Pankaj Bharadiya <pankaj.laxminarayan.bharadiya@intel.com>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200220165507.16823-9-pankaj.laxminarayan.bharadiya@intel.com
