Switch to per-device debugs in the hdcp code so we see at least which
device is involved. Should proably also print the connector/encoder/etc.
in there, but left that for the future.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240208151720.7866-9-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
for_each_old_global_obj_in_state() gives us the old state, not the
new state. Correct the name of the macro argument.
Note that while the argument was misnamed the macro did work
correctly regardless.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240208151720.7866-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Commit bd077259d0 ("drm/i915/vdsc: Add function to read any PPS
register") defines a new macro to calculate the DSC PPS register
addresses with PPS number as an input. This macro correctly calculates
the addresses till PPS 11 since the addresses increment by 4. So in that
case the following macro works correctly to give correct register
address:
_MMIO(_DSCA_PPS_0 + (pps) * 4)
However after PPS 11, the register address for PPS 12 increments by 12
because of RC Buffer memory allocation in between. Because of this
discontinuity in the address space, the macro calculates wrong addresses
for PPS 12 - 16 resulting into incorrect DSC PPS parameter value
read/writes causing DSC corruption.
This fixes it by correcting this macro to add the offset of 12 for PPS
>=12.
v3: Add correct paranthesis for pps argument (Jani Nikula)
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/10172
Fixes: bd077259d0 ("drm/i915/vdsc: Add function to read any PPS register")
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Cc: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Cc: Animesh Manna <animesh.manna@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Sean Paul <sean@poorly.run>
Cc: Drew Davenport <ddavenport@chromium.org>
Signed-off-by: Manasi Navare <navaremanasi@chromium.org>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240205204619.1991673-1-navaremanasi@chromium.org
ALPM Entry Check represents the number of lines needed to put the main link
to sleep and keep it in the sleep state before it can be taken out of the
SLEEP state (eDP requires the main link to be in the SLEEP state for a
minimum of 5us).
Bspec: 71477
v2: move display version check into _lnl_compute_alpm_param
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Arun R Murthy <arun.r.murthy@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240130111130.3298779-4-jouni.hogander@intel.com
Annotate a few more of the failure paths on the initial
BIOS fb takeover to avoid having to guess why things
aren't working the way we expect.
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-17-ville.syrjala@linux.intel.com
On MTL the GOP (for whatever reason) likes to bind its framebuffer
high up in the ggtt address space. This can conflict with whatever
ggtt_reserve_guc_top() is trying to do, and the result is that
ggtt_reserve_guc_top() fails and then we proceed to explode when
trying to tear down the driver. Thus far I haven't analyzed what
causes the actual fireworks, but it's not super important as even
if it didn't explode we'd still fail the driver load and the user
would be left with an unusable GPU.
To remedy this (without having to figure out exactly what
ggtt_reserve_guc_top() is trying to achieve) we can attempt to
relocate the BIOS framebuffer to a lower ggtt address. We can do
this at this early point in driver init because nothing else is
supposed to be clobbering the ggtt yet. So we simply change where
in the ggtt we pin the vma, the original PTEs will be left as is,
and the new PTEs will get written with the same dma addresses.
The plane will keep on scanning out from the original PTEs until
we are done with the whole process, and at that point we rewrite
the plane's surface address register to point at the new ggtt
address.
Since we don't need a specific ggtt address for the plane
(apart from needing it to land in the mappable region for
normal stolen objects) we'll just try to pin it without a fixed
offset first. It should end up at the lowest available address
(which really should be 0 at this point in the driver init).
If that fails we'll fall back to just pinning it exactly to the
origianal address.
To make sure we don't accidentlally pin it partially over the
original ggtt range (as that would corrupt the original PTEs)
we reserve the original range temporarily during this process.
v2: Try to pin explicitly to ggtt offset 0 as otherwise DG2 puts it
even higher (atm we have no PIN_LOW flag to force it low)
v3: "fix" xe
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-16-ville.syrjala@linux.intel.com
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Currently we assume that we bind the BIOS fb exactly into the same
ggtt address where the BIOS left it. That is about to change, and
in order to keep intel_reuse_initial_plane_obj() working as intended
we need to compare the original ggtt offset (called 'base' here)
as opposed to the actual vma ggtt offset we selected. Otherwise
the first plane could change the ggtt offset, and then subsequent
planes would no longer notice that they are in fact using the same
ggtt offset that the first plane was already using. Thus the reuse
check will fail and we proceed to turn off these subsequent planes.
TODO: would probably make more sense to do the pure readout first
for all the planes, then check for fb reuse, and only then proceed
to pin the object into the final location in the ggtt...
v2: "fix" xe
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-15-ville.syrjala@linux.intel.com
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
The "io" address of an object is its dma address minus the
region.start. Subtract the latter to make smem_start correct.
The current code happens to work for genuine LMEM objects
as LMEM region.start==0, but for LMEMBAR stolen objects
region.start!=0.
TODO: perhaps just set smem_start=0 always as our .fb_mmap()
implementation no longer depends on it? Need to double check
it's not needed for anything else...
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-14-ville.syrjala@linux.intel.com
There's no reason the caller of intel_initial_plane_config() should
have to loop over the CRTCs. Pull the loop into the function to
make life simpler for the caller.
v2: "fix" xe
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-13-ville.syrjala@linux.intel.com
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Declutter initial_plane_vma() a bit by pulling the lmem and smem
readout paths into their own functions.
TODO: the smem path should still be fixed to get and validate
the dma address from the pte as well
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-12-ville.syrjala@linux.intel.com
On MTL the stolen region starts at offset 8MiB from the start of
LMEMBAR. The dma addresses are thus also offset by 8MiB. However the
mm_node/etc. is zero based, and i915_pages_create_for_stolen() will
add the appropriate region.start into the sg dma address. So when
we do the readout we need to convert the dma address read from
the PTE to be zero based as well.
Note that currently we don't take this path on MTL, but we should
and thus this needs to be fixed. For lmem this works correctly
already as the lmem region.start==0.
While at it let's also make sure the address points to somewhere within
the memory region. We don't need to check the size as
i915_gem_object_create_region_at() should later fail if the object size
exceeds the region size.
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-9-ville.syrjala@linux.intel.com
When multiple pipes are enabled by the BIOS we try to read out each
in turn. But we do the readout for the second only after the inherited
vma for the first has been rebound into its original place (and thus
the PTEs have been rewritten). Unlike the BIOS we set some high caching
bits in the PTE on MTL which confuses the readout for the second plane.
Filter out the non-address bits from the PTE value appropriately to
fix this.
I suppose it might also be possible that the BIOS would already set
some caching bits as well, in which case we'd run into this same
issue already for the first plane.
TODO:
- should abstract the PTE decoding to avoid details leaking all over
- should probably do the readout for all the planes before
we touch anything (including the PTEs) so that we truly read
out the BIOS state
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Acked-by: Nirmoy Das <nirmoy.das@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-8-ville.syrjala@linux.intel.com
mem->region is a struct resource, but mem->io_start and
mem->io_size are not for whatever reason. Let's unify this
and convert the io stuff into a struct resource as well.
Should make life a little less annoying when you don't have
juggle between two different approaches all the time.
Mostly done using cocci (with manual tweaks at all the
places where we mutate io_size by hand):
@@
struct intel_memory_region *M;
expression START, SIZE;
@@
- M->io_start = START;
- M->io_size = SIZE;
+ M->io = DEFINE_RES_MEM(START, SIZE);
@@
struct intel_memory_region *M;
@@
- M->io_start
+ M->io.start
@@
struct intel_memory_region M;
@@
- M.io_start
+ M.io.start
@@
expression M;
@@
- M->io_size
+ resource_size(&M->io)
@@
expression M;
@@
- M.io_size
+ resource_size(&M.io)
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Acked-by: Nirmoy Das <nirmoy.das@intel.com>
Tested-by: Paz Zcharya <pazz@chromium.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240202224340.30647-2-ville.syrjala@linux.intel.com
AFAICS there is no hardware restriction on where in ggtt
the hdcp gsc message object needs to be bound. And as it's
a regular shmem object we don't need it be in the mappabe
range either. So pin it high to make avoid needlessly
wasting the precious mappable range for it.
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Cc: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231215110933.9188-3-ville.syrjala@linux.intel.com
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
intel_hdcp_component_init()->...->intel_hdcp_gsc_initialize_message()
will allocate ggtt address space for some hdcp gsc message thing.
That is currently being done way too early as we haven't even
taken over the BIOS fb yet. So this has the potential of corrupting
ggtt PTEs that need to be preserved until the BIOS fb takover
is done.
Only call intel_hdcp_component_init() once all the BIOS fb takeover,
and full ggtt init (which currently also needs to reserve very
specific ranges of ggtt, thus assuming that no one else has stolen
them yet) is done.
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Cc: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231215110933.9188-2-ville.syrjala@linux.intel.com
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Only display workarounds 0391 and 0475 call for disabling
FBC with render compression, and those are listed only for
pre-prod SKL steppings. So it should be safe to enable
FB+CCS on production hardware.
AFAIK CCS is limited to 50% bandwidth reduction (perhaps
clear color can do better?). FBC can exceed that number
by quite a bit, given the right kind of framebuffer
contents. So piling on both kinds of compressions could
still make sense.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/10125
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240123090244.30025-1-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Pull all the state swap stuff into its own function to declutter
intel_atomic_commit() a bit.
Note that currently the state swap is spread across both
sides of the unprepare branch in intel_atomic_commit(), but
we can pull all of it ahead a bit since we bail on the first
error, and thus there is no change in behaviour from the
reordering.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231219130756.25986-4-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Instead of injecting extra crtc commits to serialize the global
state let's hand roll a bit of commit machinery to take care of
the hardware synchronization.
Rather than basing everything on the crtc commits we track these
as their own thing. I think this makes more sense as the hardware
blocks we are working with are not in any way tied to the pipes,
so the completion should not be tied in with the vblank machinery
either.
The difference to the old behaviour is that:
- we no longer pull extra crtcs into the commit which should
make drm_atomic_check_only() happier
- since those crtcs don't get pulled in we also don't end up
reprogamming them and thus don't need to wait their vblanks
to pass/etc. So this should be tad faster as well.
TODO: perhaps have each global object complete its own commit
once the post-plane update phase is done?
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6728
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231219130756.25986-3-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
drm_atomic_check_only() gets upset if we try to add extra crtcs
to any commit that isn't flagged with DRM_MODE_ATOMIC_ALLOW_MODESET.
This conflicts with how SAGV watermarks work on pre-ADL as we
need to manually switch over the SAGV watermarks before we can
safely enable SAGV.
So in order to make SAGV usage possible we need to compute each
pipe's use of SAGV watermarks as if there aren't any other
active pipes. Ie. if the current pipe isn't the one blocking
SAGV then we make it use the SAGV watermarks, even if some
other pipe prevents SAGV from actually being used. Otherwise
we could end up with a pipes using the normal watermarks (but
not blocking SAGV), and some other pipe in parallel enabling
SAGV, which would likely cause underruns.
The alternative approach of preventing SAGV usage until all
pipes simultanously end up using SAGV watermarks would only
really work if userspace always adds all pipes to every
commits, which isn't the case typically.
The downside of this is that we will end up using the less
optimal SAGV watermarks even if some other pipe prevents
SAGV from actually being enabled. In which case the system
won't achieve the minimum possible power consumption.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231219130756.25986-2-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Commit 8015bee0bf ("drm/i915/display: Add framework to add parameters
specific to display") added the file intel_display_debugfs_params.c,
which calls the functions "debugfs_create_{bool, ulong, str}" -- all of
which are defined in <linux/debugfs.h>. The missing inclusion of this
header file is breaking the ChromeOS build -- add an explicit include
to fix that.
Signed-off-by: Paz Zcharya <pazz@chromium.org>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240131204658.795278-1-pazz@chromium.org
Some registers for DDI A/B moved to PICA and now follow the same format
as the ones for the PORT_TC ports. The wrapper here deals with 2 issues:
- Share the implementation between xe2lpd and previous
platforms: there are minor layout changes, it's mostly the
register location that changed
- Handle offsets after TC ports
v2:
- Explain better the trick to use just the second range (Matt Roper)
- Add missing conversions after rebase (Matt Roper)
- Use macro instead of inline function, avoiding includes in the
header (Jani)
- Prefix old macros with underscore so they don't get used by mistake,
and name the new ones using the previous names
v3: Use the same logic for the recently-introduced XELPDP_PORT_MSGBUS_TIMER
(Gustavo)
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240126224638.4132016-3-lucas.demarchi@intel.com
Bits to enable/disable and check state for D2D moved from
XELPDP_PORT_BUF_CTL1 to DDI_BUF_CTL (now named DDI_CTL_DE in the spec).
Make the functions mtl_ddi_disable_d2d() and mtl_ddi_enable_d2d generic
to work with multiple reg location and bitfield layout.
v2: Set/Clear XE2LPD_DDI_BUF_D2D_LINK_ENABLE in saved_port_bits when
enabling/disabling D2D so DDI_BUF_CTL is correctly programmed in
other places without overriding these bits (Clint)
v3: Leave saved_port_bits alone as those bits are not meant to be
modified outside of the port initialization. Rather propagate the
additional bit in DDI_BUF_CTL to be set when that register is
written again after D2D is enabled.
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240126224638.4132016-2-lucas.demarchi@intel.com
TC ports have both the MG/TC and TBT PLLs selected simultanously (so
that we can switch from MG/TC to TBT as a fallback). This doesn't play
well with the state checker that assumes that the old PLL shouldn't
have the pipe in its pipe_mask anymore. Suppress that check for these
PLLs to avoid spurious WARNs when you disconnect a TC port and a
non-disabling modeset happens before actually disabling the port.
v2: Only suppress when one of the PLLs is the TBT PLL and the
other one is not
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9816
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240123093137.9133-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
On HSW non-ULT (or at least on Dell Latitude E6540) external displays
start to flicker when we enable PSR on the eDP. We observe a much higher
SR and PC6 residency than should be possible with an external display,
and indeen much higher than what we observe with eDP disabled and
only the external display enabled. Looks like the hardware is somehow
ignoring the fact that the external display is active during PSR.
I wasn't able to redproduce this on my HSW ULT machine, or BDW.
So either there's something specific about this particular laptop
(eg. some unknown firmware thing) or the issue is limited to just
non-ULT HSW systems. All known registers that could affect this
look perfectly reasonable on the affected machine.
As a workaround let's unmask the LPSP event to prevent PSR entry
except while in LPSP mode (only pipe A + eDP active). This
will prevent PSR entry entirely when multiple pipes are active.
The one slight downside is that we now also prevent PSR entry
when driving eDP with pipe B or C, but I think that's a reasonable
tradeoff to avoid having to implement a more complex workaround.
Cc: stable@vger.kernel.org
Fixes: 783d8b8087 ("drm/i915/psr: Re-enable PSR1 on hsw/bdw")
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/10092
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240118212131.31868-1-ville.syrjala@linux.intel.com
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
(cherry picked from commit 94501c3ca6)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Currently icl_compute_tc_phy_dplls() assumes that the active
PLL will be the TC PLL (as opposed to the TBT PLL). The actual
PLL will be selected during the modeset enable sequence, but
we need to put *something* into the crtc_state->shared_dpll
already during compute_config().
The downside of assuming one PLL or the other is that we'll
fail to fastset if the assumption doesn't match what was in
use previously. So let's instead keep the same PLL that was
in use previously (assuming there was one). This should allow
fastset to work again when using TBT PLL, at least in the
steady state.
Now, assuming we want keep the same PLL may not be entirely
correct either. But we should be covered by the type-c link
reset handling which will force a full modeset by flagging
connectors_changed=true which means the resulting modeset
can't be converted into a fastset even if the full crtc state
looks identical.
Cc: Imre Deak <imre.deak@intel.com>
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240118142436.25928-1-ville.syrjala@linux.intel.com
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
This reverts commit cfeff354f7.
A core design consideration with legacy cursor updates is that the
cursor must not touch any other plane, even if we were to force it
to take the slow path. That is the real reason why the cursor uses
a fixed ddb allocation, not because bspec says so.
Treating cursors as any other plane during ddb allocation
violates that, which means we can now pull other planes into
fully unsynced legacy cursor mailbox commits. That is
definitely not something we've ever considered when designing
the rest of the code. The noarm+arm register write split in
particular makes that dangerous as previous updates can get
disarmed pretty much at any random time, and not necessarily
in an order that is actually safe (eg. against ddb overlaps).
So if we were to do this then:
- someone needs to expend the appropriate amount of brain
cells thinking through all the tricky details
- we should do it for all skl+ platforms since all
of those have double buffered wm/ddb registers. The current
arbitrary mtl+ cutoff doesn't really make sense
For the moment just go back to the original behaviour where
the cursor's ddb alloation does not change outside of
modeset/fastset. As of now anything else isn't safe.
Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231213102519.13500-10-ville.syrjala@linux.intel.com
Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Our legacy cursor updates are actually mailbox updates.
Ie. the hardware latches things once per frame on start of
vblank, but we issue an number of updates per frame,
withough any attempt to synchronize against the vblank
in software. So in theory only the last update issued
during the frame will latch, and the previous ones are
discarded.
However this can lead to problems with maintaining the
ggtt/iommu mappings as we have no idea which updates
will actually latch.
The problem is exacerbated by the hardware's annoying disarming
behaviour; any non-arming register write will disarm an already
armed update, only to be rearmed later by the arming register
(CURBASE in case of cursors). If a disarming write happens
just before the start of vblank, and the arming write happens
after start of vblank we have effectively prevented the hardware
from latching anything. And if we manage to straddle multiple
sequential vblank starts in this manner we effectively prevent
the hardware from latching any new registers for an arbitrary
amount of time. This provides more time for the (potentially
still in use by the hardware) gtt/iommu mappings to be torn
down.
A partial solution, of course, is to use vblank evasion to
avoid the register writes from spreading on both sides of
the start of vblank.
I've previously highlighted this problem as a general issue
affecting mailbox updates. I even added some notes to the
{i9xx,skl}_crtc_planes_update_arm() to remind us that the noarm
and arm phases both need to pulled into the vblank evasion
critical section if we actually decided to implement mailbox
updates in general. But as I never impelemented the noarm+arm
split for cursors we don't have to worry about that for the
moment.
We've been lucky enough so far that this hasn't really caused
problems. One thing that does help is that Xorg generally
sticks to the same cursor BO. But igt seems pretty good at
hitting this on MTL now, so apparently we have to start
thinking about this.
v2: Wait for PSR exit to avoid the vblank evasion timeout (1ms)
tripping due to PSR exit latency (~5ms typically)
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240116204927.23499-1-ville.syrjala@linux.intel.com
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
There isn't really any reason to make the caller suffer through
checking the vblank evasion min/max scanlines. If we somehow
ended up with bogus values (which really shouldn't happen)
then just skip the actual vblank evasion loop but otherwise
plow ahead as normal.
The only "real" change is that we now get+put a vblank reference
even if the min/max values are bogus, previously we skipped
directly to the end.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231213102519.13500-7-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
