The current implementation of the async flip wm0/ddb optimization
does not work at all. The biggest problem is that we skip the
whole intel_pipe_update_{start,end}() dance and thus never actually
complete the commit that is trying to do the wm/ddb change.
To fix this we need to move the do_async_flip flag to the crtc
state since we handle commits per-pipe, not per-plane.
Also since all planes can now be included in the first/last
"async flip" (which gets converted to a sync flip to do the
wm/ddb mangling) we need to be more careful when checking if
the plane state is async flip comptatible. Only planes doing
the async flip should be checked and other planes are perfectly
fine not adhereing to any async flip related limitations.
However for subsequent commits which are actually going do the
async flip in hardware we want to make sure no other planes
are in the state. That should never happen assuming we did our
job correctly, so we'll toss in a WARN to make sure we catch
any bugs here.
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Fixes: c3639f3be4 ("drm/i915: Use wm0 only during async flips for DG2")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220214105532.13049-4-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
(cherry picked from commit 2e08437160)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Since the async flip state check is done very late and
thus it can see potentially all the planes in the state
(due to the wm/ddb optimization) we need to move the
"can the requested plane do async flips at all?" check
much earlier. For this purpose we introduce
intel_async_flip_check_uapi() that gets called early during
the atomic check.
And for good measure we'll throw in a couple of basic checks:
- is the crtc active?
- was a modeset flagged?
- is+was the plane enabled?
Though atm all of those should be guaranteed by the fact
that the async flip can only be requested through the legacy
page flip ioctl.
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Fixes: c3639f3be4 ("drm/i915: Use wm0 only during async flips for DG2")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220214105532.13049-3-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
(cherry picked from commit b0b2bed2a1)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Currently we are observing occasional screen flickering when
PSR2 selective fetch is enabled. More specifically glitch seems
to happen on full frame update when cursor moves to coords
x = -1 or y = -1.
According to Bspec SF Single full frame should not be set if
SF Partial Frame Enable is not set. This happened to be true for
ADLP as PSR2_MAN_TRK_CTL_ENABLE is always set and for ADL_P it's
actually "SF Partial Frame Enable" (Bit 31).
Setting "SF Partial Frame Enable" bit also on full update seems to
fix screen flickering.
Also make code more clear by setting PSR2_MAN_TRK_CTL_ENABLE
only if not on ADL_P. Bit 31 has different meaning in ADL_P.
Bspec: 49274
v2: Fix Mihai Harpau email address
v3: Modify commit message and remove unnecessary comment
Tested-by: Lyude Paul <lyude@redhat.com>
Fixes: 7f6002e580 ("drm/i915/display: Enable PSR2 selective fetch by default")
Reported-by: Lyude Paul <lyude@redhat.com>
Cc: Mihai Harpau <mharpau@gmail.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Bugzilla: https://gitlab.freedesktop.org/drm/intel/-/issues/5077
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: 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/20220225070228.855138-1-jouni.hogander@intel.com
(cherry picked from commit 8d5516d18b)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
If the vm doesn't request async binding, like for example with the dpt,
then we should be able to skip the async path and avoid calling
i915_vm_lock_objects() altogether. Currently if we have a moving fence
set for the BO(even though it might have signalled), we still take the
async patch regardless of the bind_async setting, and then later still
end up just doing i915_gem_object_wait_moving_fence() anyway.
Alternatively we would need to add dummy scratch object which can be
locked, just for the dpt.
Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220304095934.925036-2-matthew.auld@intel.com
vms are not getting properly closed. Rather than fixing that,
Remove the vm open count and instead rely on the vm refcount.
The vm open count existed solely to break the strong references the
vmas had on the vms. Now instead make those references weak and
ensure vmas are destroyed when the vm is destroyed.
Unfortunately if the vm destructor and the object destructor both
wants to destroy a vma, that may lead to a race in that the vm
destructor just unbinds the vma and leaves the actual vma destruction
to the object destructor. However in order for the object destructor
to ensure the vma is unbound it needs to grab the vm mutex. In order
to keep the vm mutex alive until the object destructor is done with
it, somewhat hackishly grab a vm_resv refcount that is released late
in the vma destruction process, when the vm mutex is no longer needed.
v2: Address review-comments from Niranjana
- Clarify that the struct i915_address_space::skip_pte_rewrite is a hack
and should ideally be replaced in an upcoming patch.
- Remove an unneeded continue in clear_vm_list and update comment.
v3:
- Documentation update
- Commit message formatting
Co-developed-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220304082641.308069-2-thomas.hellstrom@linux.intel.com
The current implementation of i915 prime mmap only works when initializing
drm_i915_gem_object with shmem_region.
When using LMEM, drm_i915_gem_object is initialized with ttm_system_region.
In order to make prime mmap work even this case, when using LMEM
(when using ttm in i915), dma_buf_ops.mmap callback function calls
drm_gem_prime_mmap(). drm_gem_prime_mmap() of drm core calls internally
i915_gem_mmap() so that prime mmap can perform normally.
The fake offset is processed inside drm_gem_prime_mmap().
Testcase: igt/prime_mmap
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Gwan-gyeong Mun <gwan-gyeong.mun@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220225131316.1433515-3-gwan-gyeong.mun@intel.com
Toss a bunch if constants into .rodata drom the stack. Also
shrink the types of some of the arrays to reduce the size.
bloat-o-meter -c intel_dpll_mgr.o:
add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-86 (-86)
Function old new delta
icl_get_dplls 3393 3372 -21
skl_get_dpll 2069 2004 -65
Total: Before=28029, After=27943, chg -0.31%
add/remove: 0/0 grow/shrink: 0/0 up/down: 0/0 (0)
Data old new delta
Total: Before=17, After=17, chg +0.00%
add/remove: 2/0 grow/shrink: 0/2 up/down: 28/-129 (-101)
RO Data old new delta
dco_central_freq - 24 +24
div1_vals - 4 +4
odd_dividers 28 7 -21
even_dividers 144 36 -108
Total: Before=3600, After=3499, chg -2.81%
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220301173128.6988-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Now that we track framestart_delay in the crtc state with readout
and state checker support we can remove the explicit framestart_delay
sanitation code.
Also I'm not convinced reprogramming this while the pipe is running
is even valid. CHICKEN_TRANS (hsw+) and TRANS_CHICKEN2 (cpt+) docs
at least make no mention of double buffering which seems to imply
that live reprogramming is not supported. On older platforms
PIPECONF and PCH_TRANSCONF (ibx) are double buffered though, so
might be that we could do this on the older platforms. But doesn't
really make sense to special case old platforms for this.
So from now on if the BIOS has misprogrammed this we shall simply do
a full modeset at boot to fix it up. Such systems will of course lose
fastboot, but I think less code (and less uncertainty what
reprogramming this on a running pipe will even do) outweighs that.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220221110356.5532-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
In the past we've always assumed that an RCS engine is present on every
platform. However now that we have compute engines there may be
platforms that have CCS engines but no RCS, or platforms that are
designed to have both, but have the RCS engine fused off.
Various engine-centric initialization that only needs to be done a
single time for the group of RCS+CCS engines can't rely on being setup
with the RCS now; instead we add a I915_ENGINE_FIRST_RENDER_COMPUTE flag
that will be assigned to a single engine in the group; whichever engine
has this flag will be responsible for some of the general setup
(RCU_MODE programming, initialization of certain workarounds, etc.).
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220303223435.2793124-1-matthew.d.roper@intel.com
Cross-subsystem Changes:
- drm-next backmerge for buddy allocator changes
Driver Changes:
- Skip i915_perf init for DG2 as it is not yet enabled (Ram)
- Add missing workarounds for DG2 (Clint)
- Add 64K page/align support for platforms like DG2 that require it (Matt A, Ram, Bob)
- Add accelerated migration support for DG2 (Matt A)
- Add flat CCS support for XeHP SDV (Abdiel, Ram)
- Add Compute Command Streamer (CCS) engine support for XeHP SDV (Michel,
Daniele, Aravind, Matt R)
- Don't support parallel submission on compute / render (Matt B, Matt R)
- Disable i915 build on PREEMPT_RT until RT behaviour fixed (Sebastian)
- Remove RPS interrupt support for TGL+ (Jose)
- Fix S/R with PM_EARLY for non-GTT mappable objects on DG2 (Matt, Lucas)
- Skip stolen memory init if it is fully reserved (Jose)
- Use iosys_map for GuC data structures that may be in LMEM BAR or SMEM (Lucas)
- Do not complain about stale GuC reset notifications for banned contexts (John)
- Move context descriptor fields to intel_lrc.h
- Start adding support for small BAR (Matt A)
- Clarify vma lifetime (Thomas)
- Simplify subplatform detection on TGL (Jose)
- Correct the param count for unset GuC SLPC param (Vinay, Umesh)
- Read RP_STATE_CAP correctly on Gen12 with GuC SLPC (Vinay)
- Initialize GuC submission locks and queues early (Daniele)
- Fix GuC flag query helper function to not modify state (John)
- Drop fake lmem support now we have real hardware available (Lucas)
- Move misplaced W/A to their correct locations (Srinivasan)
- Use get_reset_domain() helper (Tejas)
- Move context descriptor fields to intel_lrc.h (Matt R)
- Selftest improvements (Matt A)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YiBzY1dM7bKwMQ3H@jlahtine-mobl.ger.corp.intel.com
The LRC descriptor pool is going away. Further, the function that was
populating it was also doing a bunch of logic about the context
registration sequence. So, split that code apart into separate state
setup and try to register functions. Note that some of those 'try to
register' code paths actually undo the state setup and leave it to be
redone again later (with potentially different values). This is
inefficient. The next patch will correct this.
Also, move a comment about ignoring return values to the place where
the return values are actually ignored.
v2: Move some more splitting from a later patch (and do it correctly).
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220302003357.4188363-5-John.C.Harrison@Intel.com
The current implementation of the async flip wm0/ddb optimization
does not work at all. The biggest problem is that we skip the
whole intel_pipe_update_{start,end}() dance and thus never actually
complete the commit that is trying to do the wm/ddb change.
To fix this we need to move the do_async_flip flag to the crtc
state since we handle commits per-pipe, not per-plane.
Also since all planes can now be included in the first/last
"async flip" (which gets converted to a sync flip to do the
wm/ddb mangling) we need to be more careful when checking if
the plane state is async flip comptatible. Only planes doing
the async flip should be checked and other planes are perfectly
fine not adhereing to any async flip related limitations.
However for subsequent commits which are actually going do the
async flip in hardware we want to make sure no other planes
are in the state. That should never happen assuming we did our
job correctly, so we'll toss in a WARN to make sure we catch
any bugs here.
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Fixes: c3639f3be4 ("drm/i915: Use wm0 only during async flips for DG2")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220214105532.13049-4-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Since the async flip state check is done very late and
thus it can see potentially all the planes in the state
(due to the wm/ddb optimization) we need to move the
"can the requested plane do async flips at all?" check
much earlier. For this purpose we introduce
intel_async_flip_check_uapi() that gets called early during
the atomic check.
And for good measure we'll throw in a couple of basic checks:
- is the crtc active?
- was a modeset flagged?
- is+was the plane enabled?
Though atm all of those should be guaranteed by the fact
that the async flip can only be requested through the legacy
page flip ioctl.
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Fixes: c3639f3be4 ("drm/i915: Use wm0 only during async flips for DG2")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220214105532.13049-3-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
