- 29 Mar, 2022 7 commits
-
-
Ville Syrjälä authored
All the other intel_panel functions take struct intel_connector, so might as well make init()/fini() take one as well. Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220323182935.4701-2-ville.syrjala@linux.intel.comReviewed-by:
Jani Nikula <jani.nikula@intel.com>
-
Ville Syrjälä authored
Move most of the pipe+output CSC programming to the .color_commit_noarm() hook which runs before vblank evasion. Only PIPE_CSC_MODE (the arming register) needs to remain in inside the critical section. A test case that just updates the CTM in a loop produces the following i915_update_info numbers on ilk (w/o lockdep): old new Updates: 10012 Updates: 10008 | | 1us |** 1us |********** |************* |************* 4us |********* 4us |* |* |** 16us | 16us | | | 66us | 66us | | | 262us | 262us | | | 1ms | 1ms | | | 4ms | 4ms | | | 17ms | 17ms | | | Min update: 1345ns Min update: 1268ns Max update: 16672ns Max update: 15656ns Average update: 3914ns Average update: 2185ns Overruns > 100us: 0 Overruns > 100us: 0 And here is tgl (forced to update both pipe CSC and output CSC, and with lockdep enabled): old new Updates: 10012 Updates: 10012 | | 1us | 1us | | | 4us |* 4us |** |** |********** 16us |************* 16us |************* |* | 66us | 66us | | | 262us | 262us | | | 1ms | 1ms | | | 4ms | 4ms | | | 17ms | 17ms | | | Min update: 5204ns Min update: 5176ns Max update: 176038ns Max update: 186685ns Average update: 23931ns Average update: 16654ns Overruns > 250us: 0 Overruns > 250us: 0 Signed-off-by:Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
-
Ville Syrjälä authored
To reduce the amount of registers written during the vblank evade critical section let's also split the .color_commit() hook to noarm+arm pair. The noarm hook runs before the vblank evasion with the arm hook staying inside the critical section. Just the framework here, actually moving stuff out into the noarm hook will follow. Signed-off-by:
-
Ville Syrjälä authored
The ilk+ panel fitter register are sitting nicely on their own cacheline, so no need for global serialization via uncore.lock. Signed-off-by:
-
Ville Syrjälä authored
All the skl+ scaler registers are suitably confined to their own cachelines so we don't need the uncore.lock to globally serialize access to these registers. We actually already dropped some of this in commit 14ad1529 ("drm/i915: Make skl+ universal plane registers unlocked") as the plane scaler enabling/reconfiguration became lockless. So let's complete that and remove the rest of the locks from the scaler programming as well. Signed-off-by:
-
José Roberto de Souza authored
intel_drrs_enable() and intel_drrs_disable() were renamed to intel_drrs_activate() and intel_drrs_deactivate() in commit 54903c7a ("drm/i915: s/enable/active/ for DRRS") and it is causing warnings when generating the kernel documentation. But as for a while DRRS has its own file, so here just let the tool generate the documentation for all exported and documented functions in intel_drrs.c. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by:
José Roberto de Souza <jose.souza@intel.com> Reviewed-by:
-
Imre Deak authored
At least some DELL monitors (P2715Q) with DPCD_REV 1.2 return corrupted DPCD register values when reading from the 0xF0000- LTTPR range with an AUX transaction block size bigger than 1. The DP standard requires 0 to be returned - as for any other reserved/invalid addresses - but these monitors return the DPCD_REV register value repeated in each byte of the read buffer. This will in turn corrupt the values returned by the LTTPRs between the source and the monitor: LTTPRs must adjust the values they read from the downstream DPRX, for instance right-shift/init the downstream DP_PHY_REPEATER_CNT value. Since the value returned by the monitor's DPRX is non-zero the adjusted values will be corrupt. Reading the LTTPR registers one-by-one instead of reading all of them with a single AUX transfer works around the issue. According to the DP standard's 0xF0000 register description: "LTTPR-related registers at DPCD Addresses F0000h through F02FFh are valid only for DPCD r1.4 (or higher)." While it's unclear if DPCD r1.4 refers to the DPCD_REV or to the LT_TUNABLE_PHY_REPEATER_FIELD_DATA_STRUCTURE_REV register (tickets filed at the VESA site to clarify this haven't been addressed), one possibility is that it's a restriction due to non-compliant monitors described above. Disabling the non-transparent LTTPR mode for all such monitors is not a viable solution: the transparent LTTPR mode has its own issue causing link training failures and this would affect a lot of monitors in use with DPCD_REV < 1.4. Instead this patch works around the problem by reading the LTTPR common and PHY cap registers one-by-one for any monitor with a DPCD_REV < 1.4. The standard requires the DPCD capabilities to be read after the LTTPR common capabilities are read, so re-read the DPCD capabilities after the LTTPR common and PHY caps were read out. v2: - Use for instead of a while loop. (Ville) - Add to code comment the monitor model with the problem. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4531 Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
Imre Deak <imre.deak@intel.com> Reviewed-by:
-
- 28 Mar, 2022 3 commits
-
-
Uma Shankar authored
HSW+ platforms are able to send out HDR Metadata SDP DIP packet as GMP. Hence, extending the support for HDR on DP encoders for the same. v2: Limited to non eDP ports on hsw/bdw and removed it for lspcon as it is done separately (suggested by Ville) v3: Added helper and limited eDP restriction to port A (Ville) v4: Dropped some redundant checks (Ville) Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5389 Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
Uma Shankar <uma.shankar@intel.com> Reviewed-by:
-
Ankit Nautiyal authored
Print I/O voltage and process info for each combo phy ports. v2: Used drm_dbg_kms for logs. (Jani) Added names for different voltage levels. (Imre) v3: Used const char * for names. (Jani) v4: Dropped the procom values and changed commit msg (Imre) Suggested-by:
Ankit Nautiyal <ankit.k.nautiyal@intel.com> Reviewed-by:
-
Ankit Nautiyal authored
The low voltage sku check can be ignored as OEMs need to consider that when designing the board and then put any limits in VBT. Same is now changed in Bspec pages. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5272Signed-off-by:
-
- 21 Mar, 2022 12 commits
-
-
Ville Syrjälä authored
Rename the DRRS functiosn to say "(de)activate" rather than "enable/disable". This let's us differentiate between the logically enabled vs. actually currently active cases. v2: Fix kernel doc for intel_drrs_deactivate() Signed-off-by:
-
Ville Syrjälä authored
Make sure the CDCLK is high enough to support the so called "maximum pipe read bandwidth" limitation. Specified as 51.2 x CDCLK. Signed-off-by:
-
Ville Syrjälä authored
Make the dbuf bandwidth min cdclk calculations match the spec more closely. Supposedly the arbiter can only guarantee an equal share of the total bandwidth of the slice to each active plane on that slice. So we take the max bandwidth of any of the planes on each slice and multiply that by the number of active planes on the slice to get a worst case estimate on how much bandwidth we require. Signed-off-by:
-
Ville Syrjälä authored
The current code also forgets to call intel_atomic_lock_global_state() when other stuff besides the final min_cdlck changes in the state. That means we may throw away data which actually has changed, and thus we can't be at all sure what the code ends up doing during subsequent commits. Do the write lock properly. Signed-off-by:
-
Ville Syrjälä authored
We should round up when doing bandwidth calculations to make sure our estimates don't fall short of the actual number. Signed-off-by:
-
Ville Syrjälä authored
intel_bw_calc_min_cdclk() is entirely pointless. All it manages to do is somehow conflate the per-pipe min cdclk with dbuf min cdclk. There is no (at least documented) dbuf min cdclk limit on pre-skl so let's just get rid of all this confusion. Reviewed-by:
-
Ville Syrjälä authored
There's really no need to maintain these total[] arrays to track the size of each plane's ddb allocation. We just stick the results straight into the crtc_state ddb tracking structures. The main annoyance with all this is the mismatch between wm_uv vs. ddb_y on pre-icl. If only the hw was consistent in what it considers the primary source of information we could avoid some of the uglyness. But since that is not the case we need a bit of special casing for planar formats. v2: Keep the ddb entry zeroed when the plane is disabled Reviewed-by:
-
Ville Syrjälä authored
Handle the plane relative data rate in exactly the same way as we already handle the real data rate. Ie. pre-calculate it during intel_plane_atomic_check_with_state(), and assign/clear it for the Y plane as needed. This should guarantee that the tracking is 100% consistent, and makes me have to think less when the same apporach is used by both types of data rate. We might even want to consider replacing the relative data rate with the real data rate entirely, but it's not clear if that will produce less optimal plane ddb allocations. So for now lets keep using the current approach. v2: Rebase due to async flip wm optimization Reviewed-by:
-
Ville Syrjälä authored
Split the currently combined plane data_rate into the proper Y vs. CbCr components. This matches how we now track the plane dbuf allocations, and thus will make the dbuf bandwidth calculations actually produce the correct numbers for each dbuf slice. Reviewed-by:
-
Ville Syrjälä authored
Let's store the plane allocation in a manner which more closely matches how the hw operates. That is, we store the packed/CbCr ddb in one struct, and the Y ddb in another. Currently we're storing packed/Y in one struct, CbCr in the other. This also works pretty well for icl+ where the UV plane is the main plane and the Y plane is subservient to it. Although in this case we do not even use ddb_y as we do the ddb allocation in terms of hw planes. v2: Rebase Reviewed-by:
-
José Roberto de Souza authored
A few more updates in the alderlake-P voltage swing tables. eDP HBR3 table was the same as icelake one but now it has changes for voltage 0 and pre-emphasis 2 line. And DP tables also had one line change in each. Bspec: 49291 Signed-off-by:
-
Jani Nikula authored
The kernel preference is to use the __packed macro instead of the direct __attribute__. Signed-off-by:
-
- 18 Mar, 2022 8 commits
-
-
Ville Syrjälä authored
Name all the ICL_PCODE_SAGV_DE_MEM_SS_CONFIG request/response bits in a manner that we can actually understand what they're doing. Signed-off-by:
-
Ville Syrjälä authored
Use separate bitmasks for QGV vs. PSF GV points during the computation. Makes the whole thing a lot less confusing. Signed-off-by:
-
Ville Syrjälä authored
Don't just mask off all the PSF GV points when SAGV gets disabled. This should in fact cause the Pcode to reject the request since at least one PSF point must remain enabled at all times. Cc: stable@vger.kernel.org Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com> Fixes: 192fbfb7 ("drm/i915: Implement PSF GV point support") Signed-off-by:
-
Ville Syrjälä authored
Give the pre-icl SAGV control functions a skl_ prefix instead of the intel_ prefix to make it a bit more clear that they are not some kind of universal things that can be called on any platform. Also make the functions void since we never use the return value anyway. Signed-off-by:
-
Ville Syrjälä authored
If the mailbox returns an exceesively large SAGV block time let's just reject it. This avoids having to worry about overflows when we add the SAGV block time to the wm0 latency. We shall put the limit arbitrarily at U16_MAX. >65msec latency doesn't really make sense to me in any case. Signed-off-by:
-
Ville Syrjälä authored
Instead of leaving the SAGV enable/disable to the first commit let's try to disable it first thing to see if we can do it or not (disabling SAGV is a safe thing to at any time). This avoids running the code in this funny intermediate state where we don't know if SAGV is available or not. Signed-off-by:
-
Ville Syrjälä authored
I'd like to see the SAGV block time we got from the mailbox in the logs regardless of whether other factors prevent the use of SAGV. So let's adjust the code to always query the SAGV block time, log it, and then reset it if SAGV is not actually supported. Signed-off-by:
-
Ville Syrjälä authored
For modern platforms the spec explicitly states that a SAGV block time of zero means that SAGV is not supported. Let's extend that to all platforms. Supposedly there should be no systems where this isn't true, and it'll allow us to: - use the same code regardless of older vs. newer platform - wm latencies already treat 0 as disabled, so this fits well with other related code - make it a bit more clear when SAGV is used vs. not - avoid overflows from adding U32_MAX with a u16 wm0 latency value which could cause us to miscalculate the SAGV watermarks on tgl+ Cc: stable@vger.kernel.org Signed-off-by:
-
- 17 Mar, 2022 2 commits
-
-
Jani Nikula authored
Remove the uapi/drm/i915_drm.h include from drm/i915_drm.h, and stop being a proxy for uapi/drm/i915_drm.h. Include uapi/drm/i915_drm.h and drm/i915_drm.h only where needed. Signed-off-by:
-
Jani Nikula authored
Move i915_gem_object_needs_bit17_swizzle() to i915_gem_tiling.[ch] as a i915_gem_object function related to tiling. Also un-inline while at it; does not seem like this is a function needed in hot paths. v2: i915_gem_tiling.[ch] instead of intel_ggtt_fencing.[ch] (Chris) Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220316095018.137998-1-jani.nikula@intel.com
-
- 16 Mar, 2022 8 commits
-
-
Ville Syrjälä authored
ICL+ PLLs can't genenerate certain frequencies. Running the PLL algorithms through for all frequencies 25-594MHz we see a gap just above 500 MHz. Specifically 500-522.8MHZ for TC PLLs, and 500-533.2 MHz for combo PHY PLLs. Reject those frequencies hdmi_port_clock_valid() so that we properly filter out unsupported modes and/or color depths for HDMI. Cc: stable@vger.kernel.org Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5247Signed-off-by:
Mika Kahola <mika.kahola@intel.com>
-
Ville Syrjälä authored
DRRS operates on transcoder level, so we should only poke at it from the master crtc rather than letting every joined pipe give it potentially conflicting input. Signed-off-by:
Manasi Navare <manasi.d.navare@intel.com>
-
Ville Syrjälä authored
Let's just do a full DRRS disable/enable across all pipe updates. This guarantees that the DRRS work doesn't interfere with anything while the atomic commit is busy reprogramming the pipe. Needed so that we can start reprogramming M/N seamlessly during fastsets whenever possible. Also avoids the pre-bdw DRRS PIPECONF rmw racing with the potential PIPECONF write from the atomic commit (eg. due to GAMMA_MODE changes). v2: Include has_drrs in state dump (José) Reviewed-by:
-
Ville Syrjälä authored
Schedule the DRRS downclock work already from intel_drrs_enable() instead of waiting around for a frontbuffer flush that may or may not ever come. Signed-off-by:
-
Ville Syrjälä authored
Skip all the DRRS work cancel/schedule stuff if the pipe's frontbuffer bits were not among those affected by the frontbuffer rendering. Signed-off-by:
-
Ville Syrjälä authored
Pre-determine the frontbuffer_bits for the each pipe during intel_drrs_enable(). Will become useful for bigjoiner use cases soon. Signed-off-by:
-
Ville Syrjälä authored
Now that DRRS can operate on multiple pipes we need to make sure one pipe doesn't throw away the other pipe's frontbuffer_bits before said pipe can handle them. Signed-off-by:
-
Ville Syrjälä authored
The DRRS refresh rate should be indented by one tab like the other per-crtc DRRS stuff. Signed-off-by:
-
