Currently skl_pcode_try_request function doesn't
properly handle return value it gets from
snb_pcode_rw, but treats status != 0 as success,
returning true, which basically doesn't allow
to use retry/timeout mechanisms if PCode happens
to be busy and returns EGAIN or some other status
code not equal to 0.
We saw this on real hw and also tried simulating this
by always returning -EAGAIN from snb_pcode_rw for 6 times, which
currently will just result in false success, while it should
have tried until timeout is reached:
[ 22.357729] i915 0000:00:02.0: [drm:intel_cdclk_dump_config [i915]] Changing CDCLK to
307200 kHz, VCO 614400 kHz, ref 38400 kHz, bypass 19200 kHz, voltage level 0
[ 22.357831] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 1
[ 22.357892] i915 0000:00:02.0: [drm:skl_pcode_request [i915]] Success, exiting
[ 22.357936] i915 0000:00:02.0: [drm] ERROR Failed to inform PCU about cdclk change (err -11, freq 307200)
We see en error because higher level api, still notices that status was wrong,
however we still did try only once.
We fix it by requiring _both_ the status to be 0 and
request/reply match for success(true) and function
should return failure(false) if either status turns
out to be EAGAIN, EBUSY or whatever or reply/request
masks do not match.
So now we see this in the logs:
[ 22.318667] i915 0000:00:02.0: [drm:intel_cdclk_dump_config [i915]] Changing CDCLK to
307200 kHz, VCO 614400 kHz, ref 38400 kHz, bypass 19200 kHz, voltage level 0
[ 22.318782] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 1
[ 22.318849] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 2
[ 22.319006] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 3
[ 22.319091] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 4
[ 22.319158] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 5
[ 22.319224] i915 0000:00:02.0: [drm:__snb_pcode_rw [i915]] Returning EAGAIN retry 6
Reviewed-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220408125200.9069-2-stanislav.lisovskiy@intel.com
Sync up with v5.18-rc1, in particular to get 5e3094cfd9
("drm/i915/xehpsdv: Add has_flat_ccs to device info").
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
The intent of the version check in the mmap ioctl was to maintain
support for existing platforms (i.e., ADL/RPL and earlier), but drop
support on all future igpu platforms. As we've seen on the dgpu side,
the hardware teams are using a more fine-grained numbering system for IP
version numbers these days, so it's possible the version number
associated with our next igpu could be some form of "12.xx" rather than
13 or higher. Comparing against the full ver.release number will ensure
the intent of the check is maintained no matter what numbering the
hardware teams settle on.
Fixes: d3f3baa356 ("drm/i915: Reinstate the mmap ioctl for some platforms")
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
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/20220407161839.1073443-1-matthew.d.roper@intel.com
(cherry picked from commit 8e7e5c077c)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The intent of the version check in the mmap ioctl was to maintain
support for existing platforms (i.e., ADL/RPL and earlier), but drop
support on all future igpu platforms. As we've seen on the dgpu side,
the hardware teams are using a more fine-grained numbering system for IP
version numbers these days, so it's possible the version number
associated with our next igpu could be some form of "12.xx" rather than
13 or higher. Comparing against the full ver.release number will ensure
the intent of the check is maintained no matter what numbering the
hardware teams settle on.
Fixes: d3f3baa356 ("drm/i915: Reinstate the mmap ioctl for some platforms")
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
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/20220407161839.1073443-1-matthew.d.roper@intel.com
Freq caps (i.e. RP0, RP1 and RPn frequencies) are read from HW. However the
formats (bit positions, widths, registers and units) of these vary for
different generations with even more variations arriving in the future. In
order not to have to do identical computation for these caps in multiple
places, here we centralize the computation of these caps. This makes the
code cleaner and also more extensible for the future.
v2: Clarify that caps are in "hw units" in comments (Lucas De Marchi)
v3: Minor checkpatch fix
v4: s/intel_rps_get_freq_caps/gen6_rps_get_freq_caps/ (Badal Nilawar)
v5: Changes comments to kernel doc (Anshuman Gupta)
Cc: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Reviewed-by: Badal Nilawar <badal.nilawar@intel.com>
Acked-by: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Anshuman Gupta <anshuman.gupta@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220406191848.20895-1-ashutosh.dixit@intel.com
Add an usage for kernel submissions. Waiting for those are mandatory for
dynamic DMA-bufs.
As a precaution this patch also changes all occurrences where fences are
added as part of memory management in TTM, VMWGFX and i915 to use the
new value because it now becomes possible for drivers to ignore fences
with the WRITE usage.
v2: use "must" in documentation, fix whitespaces
v3: separate out some driver changes and better document why some
changes should still be part of this patch.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20220407085946.744568-5-christian.koenig@amd.com
Instead of distingting between shared and exclusive fences specify
the fence usage while adding fences.
Rework all drivers to use this interface instead and deprecate the old one.
v2: some kerneldoc comments suggested by Daniel
v3: fix a missing case in radeon
v4: rebase on nouveau changes, fix lockdep and temporary disable warning
v5: more documentation updates
v6: separate internal dma_resv changes from this patch, avoids to
disable warning temporary, rebase on upstream changes
v7: fix missed case in lima driver, minimize changes to i915_gem_busy_ioctl
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20220407085946.744568-3-christian.koenig@amd.com
This change adds the dma_resv_usage enum and allows us to specify why a
dma_resv object is queried for its containing fences.
Additional to that a dma_resv_usage_rw() helper function is added to aid
retrieving the fences for a read or write userspace submission.
This is then deployed to the different query functions of the dma_resv
object and all of their users. When the write paratermer was previously
true we now use DMA_RESV_USAGE_WRITE and DMA_RESV_USAGE_READ otherwise.
v2: add KERNEL/OTHER in separate patch
v3: some kerneldoc suggestions by Daniel
v4: some more kerneldoc suggestions by Daniel, fix missing cases lost in
the rebase pointed out by Bas.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20220407085946.744568-2-christian.koenig@amd.com
Since gen6 we use FPGA_DBG register to detect unclaimed MMIO registers.
This register is in the display engine IP and can only ever detect
unclaimed accesses to registers in this area. However sometimes there
are reports of this triggering for registers in other areas, which
should not be possible.
Right now we always warn after the read/write of registers going through
unclaimed_reg_debug(). However places using __raw_uncore_* may be
triggering the unclaimed access and those being later accounted to a
different register. Let's warn both before and after the read/write
with a slightly different message, so it's clear if the register
reported in the warning is actually the culprit.
Commit dda960335e ("drm/i915: Just clear the mmiodebug before a
register access") attempted to solve the same issue by removing the
warning when if FPGA_DBG flags before the mmio read/write. However, it
doesn't solve it completely as FPGA_DBG may remain set when reading
registers outside display. So in the end the check after the mmio
read/write triggers the warning pointing to the wrong register.
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/20220405001149.2675226-1-lucas.demarchi@intel.com
Some functions defined in the intel-gtt module are used in several
areas, but is only supported on x86 platforms.
By separating these calls and their static underlying functions to
another area, we are able to compile out these functions for
non-x86 builds and provide stubs for the non-x86 implementations.
In addition to the problematic calls, we are moving the gmch-related
functions to the new area.
Signed-off-by: Casey Bowman <casey.g.bowman@intel.com>
Acked-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220330234809.1218210-2-casey.g.bowman@intel.com
Accessing the DDI_BUF_CTL register without the port's DDI clock being
enabled (to set/clear the TypeC PHY ownership for the port) can lead to
a corrupted value read during any i915 register access right after the
DDI clock is enabled.
The root cause is the way clock synchronization works for this register,
controlled by the CHICKEN_DCPR_1 DDI_CLOCK_REG_ACCESS flag. Correctly
this flag should be cleared on ADLP (see the Bspec link below), however
after bootup the flag is set.
One easily reproducible issue is an unclaimed register access of the
PWR_WELL_CTL_DDI2 register, programmed right after DDI clock enabling to
enable the port's DDI_IO power well (see the HSDES, VLK links below).
With the correct setting above this problem can't be reproduced.
Bspec: 49189
HSDES: 18019028154
VLK: 28328, 28655
Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Arthur J Runyan <arthur.j.runyan@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Acked-by: Arthur J Runyan <arthur.j.runyan@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220323201749.288566-1-imre.deak@intel.com
Instead of exit PSR when a frontbuffer invalidation happens, we can
enable the PSR2 selective fetch continuous full frame, that will keep
the panel updated like PSR was disabled but without keeping PSR active.
So as soon as the frontbuffer flush happens we can disable the
continuous full frame and start to do selective fetches much quicker
than the path that would enable PSR, that will wait a few frames
to actually activate PSR.
Also this approach has proven to fix some glitches found in Alderlake-P
when there are a lot of invalidations happening together with page
flips.
Some may ask why it is writing to CURSURFLIVE(), it is because
that is the way that hardware team provided us to poke display to
handle PSR updates, and it is being used since display 9.
v2:
- handling possible race conditions between frontbuffer rendering and
page flips
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Cc: Khaled Almahallawy <khaled.almahallawy@intel.com>
Cc: Shawn C Lee <shawn.c.lee@intel.com>
Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405155344.47219-3-jose.souza@intel.com
Frontbuffer rendering and page flips can race with each other
and this can potentialy cause issues with PSR2 selective fetch.
And because pipe/crtc updates are time sentive we can't grab the
PSR lock after intel_pipe_update_start() and before
intel_pipe_update_end().
So here adding the lock and unlock functions and calls, the
proper PSR2 selective fetch handling will come in a separated patch.
v2:
- fixed new functions documentation
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405155344.47219-2-jose.souza@intel.com
Following up what was done in commit 804f468853 ("drm/i915/psr: Set
"SF Partial Frame Enable" also on full update") and also setting
partial frame enable when psr_force_hw_tracking_exit() is called.
Also as PSR2_MAN_TRK_CTL is a double buffered registers do a RMW
is not a good idea so here also setting the man_trk_ctl_enable_bit()
that is required in TGL and only doing a register write.
v2:
- not doing a rmw
v3:
- removing the inline from functions that return PSR2_MAN_TRK_CTL
bits
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Cc: Jouni Högander <jouni.hogander@intel.com>
Cc: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405155344.47219-1-jose.souza@intel.com
The request to aqquire gem resources is failing for DSB in rare
scenario where it is busy and the register programming will be done
through mmio fallback path.
DSB has extra advantage of faster register programming which may
go away through mmio path. Adding wait for gem resource also may
not be right as anyways losing time.
To make the CI execution happy replaced drm_err() to drm_info()
for printing debug info during dsb buffer preparation.
v1: Initial version.
v2: Added print for mmio fallback at out label. [Nirmoy]
v3: Improved debug message. [Nirmoy]
Cc: Nirmoy Das <nirmoy.das@linux.intel.com>
Signed-off-by: Animesh Manna <animesh.manna@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220325161140.11906-1-animesh.manna@intel.com
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Track context active (on hardware) status together with the start
timestamp.
This will be used to provide better granularity of context
runtime reporting in conjunction with already tracked pphwsp accumulated
runtime.
The latter is only updated on context save so does not give us visibility
to any currently executing work.
As part of the patch the existing runtime tracking data is moved under the
new ce->stats member and updated under the seqlock. This provides the
ability to atomically read out accumulated plus active runtime.
v2:
* Rename and make __intel_context_get_active_time unlocked.
v3:
* Use GRAPHICS_VER.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com> # v1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220401142205.3123159-6-tvrtko.ursulin@linux.intel.com
Tracking DRM clients more explicitly will allow later patches to
accumulate past and current GPU usage in a centralised place and also
consolidate access to owning task pid/name.
Unique client id is also assigned for the purpose of distinguishing/
consolidating between multiple file descriptors owned by the same process.
v2:
Chris Wilson:
* Enclose new members into dedicated structs.
* Protect against failed sysfs registration.
v3:
* sysfs_attr_init.
v4:
* Fix for internal clients.
v5:
* Use cyclic ida for client id. (Chris)
* Do not leak pid reference. (Chris)
* Tidy code with some locals.
v6:
* Use xa_alloc_cyclic to simplify locking. (Chris)
* No need to unregister individial sysfs files. (Chris)
* Rebase on top of fpriv kref.
* Track client closed status and reflect in sysfs.
v7:
* Make drm_client more standalone concept.
v8:
* Simplify sysfs show. (Chris)
* Always track name and pid.
v9:
* Fix cyclic id assignment.
v10:
* No need for a mutex around xa_alloc_cyclic.
* Refactor sysfs into own function.
* Unregister sysfs before freeing pid and name.
* Move clients setup into own function.
v11:
* Call clients init directly from driver init. (Chris)
v12:
* Do not fail client add on id wrap. (Maciej)
v13 (Lucas): Rebase.
v14:
* Dropped sysfs bits.
v15:
* Dropped tracking of pid/ and name.
* Dropped RCU freeing of the client object.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> # v11
Reviewed-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com> # v11
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220401142205.3123159-2-tvrtko.ursulin@linux.intel.com
