- 14 Jul, 2015 9 commits
-
-
Peter Antoine authored
This change adds the programming of the MOCS registers to the gen 9+ platforms. The set of MOCS configuration entries introduced by this patch is intended to be minimal but sufficient to cover the needs of current userspace - i.e. a good set of defaults. It is expected to be extended in the future to provide further default values or to allow userspace to redefine its private MOCS tables based on its demand for additional caching configurations. In this setup, userspace should only utilize the first N entries, higher entries are reserved for future use. It creates a fixed register set that is programmed across the different engines so that all engines have the same table. This is done as the main RCS context only holds the registers for itself and the shared L3 values. By trying to keep the registers consistent across the different engines it should make the programming for the registers consistent. v2: -'static const' for private data structures and style changes.(Matt Turner) v3: - Make the tables "slightly" more readable. (Damien Lespiau) - Updated tables fix performance regression. v4: - Code formatting. (Chris Wilson) - re-privatised mocs code. (Daniel Vetter) v5: - Changed the name of a function. (Chris Wilson) v6: - re-based - Added Mesa table entry (skylake & broxton) (Francisco Jerez) - Tidied up the readability defines (Francisco Jerez) - NUMBER of entries defines wrong. (Jim Bish) - Added comments to clear up the meaning of the tables (Jim Bish) Signed-off-by:
Peter Antoine <peter.antoine@intel.com> v7 (Francisco Jerez): - Don't write L3-specific MOCS_ESC/SCC values into the e/LLC control tables. Prefix L3-specific defines consistently with L3_ and e/LLC-specific defines with LE_ to avoid this kind of confusion in the future. - Change L3CC WT define back to RESERVED (matches my hardware documentation and the original patch, probably a misunderstanding of my own previous comment). - Drop Android tables, define new minimal tables more suitable for the open source stack. - Add comment that the MOCS tables are part of the kernel ABI. - Move intel_logical_ring_begin() and _advance() calls one level down (Chris Wilson). - Minor formatting and style fixes. v8 (Francisco Jerez): - Add table size sanity check to emit_mocs_control/l3cc_table() (Chris Wilson). - Add comment about undefined entries being implicitly set to uncached for forwards compatibility. v9 (Francisco Jerez): - Minor style fixes. Signed-off-by:
Francisco Jerez <currojerez@riseup.net> Acked-by:
Damien Lespiau <damien.lespiau@intel.com> Signed-off-by:
Daniel Vetter <daniel.vetter@ffwll.ch>
-
Daniel Vetter authored
Totatlly forgotten that we have these when nuking all the UMS code. Signed-off-by:
Daniel Vetter <daniel.vetter@intel.com> Reviewed-by:
Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by:
-
Maarten Lankhorst authored
Nothing depends on this outside initial hw readout, so keep this struct on the stack instead. Changes since v1: - Remove unrelated changes. Signed-off-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Reviewed-by:
Daniel Stone <daniels@collabora.com> Signed-off-by:
-
Maarten Lankhorst authored
The src and crtc rectangles were never set, resulting in the primary plane being made invisible on first atomic update. Signed-off-by:
-
Maarten Lankhorst authored
Instead of doing ad-hoc checks we already have a way of checking if the state is compatible or not. Use this to force a modeset. Only during modesets, or with PIPE_CONFIG_QUIRK_INHERITED_MODE we should check if a full modeset is really needed. Fastboot will allow the adjust parameter to ignore some stuff too, and it will fix up differences in state that are ignored by the compare function. Changes since v1: - Increase the value of the lowest m/n to prevent truncation. - Dump pipe config when fastboot's used, without a modeset. - Add adjust parameter to intel_compare_link_m_n, which is used to adjust m2_n2 if it's a multiple of m_n. - Add exact parameter intel_compare_m_n. Signed-off-by:
-
Maarten Lankhorst authored
Use the atomic state instead, this allows removing plane_config from the crtc after the full hw readout is completed. The size can be found in the fb, no need for the plane_config. Signed-off-by:
-
Maarten Lankhorst authored
There's not much point for calculating the changes for the old state. Instead just disable all scalers when disabling. It's probably good enough to just disable the crtc_scaler, but just in case there's a bug disable all scalers. This means intel_atomic_setup_scalers is only called in the crtc check function now, so all the transitional code can be removed. Signed-off-by:
-
Maarten Lankhorst authored
This is probably hard to hit right now because in most cases all atomic locks are taken, but after conversion to atomic this will make it more likely to corrupt the crtc->config pointer, resulting in hard to find bugs. Signed-off-by:
-
Akash Goel authored
Ring frequency table programming is not required on BXT. Added separate checks to enable the programming only for SKL & skip for BXT. v2: Removed the BXT check from gen6_update_ring_freq function Issue: VIZ-5144 Signed-off-by:
Akash Goel <akash.goel@intel.com> Reviewed-by: Rodrigo Vivi <rodrigo.vivi at intel.com> Signed-off-by:
-
- 13 Jul, 2015 11 commits
-
-
Patrik Jakobsson authored
Watermark calculations depend on the intel_crtc->active flag to be set properly. Suspend/resume is broken on SKL and we also get DDB mismatches without this patch. The regression was introduced in: commit eddfcbcd Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Date: Mon Jun 15 12:33:53 2015 +0200 drm/i915: Update less state during modeset. No need to repeatedly call update_watermarks, or update_fbc. Down to a single call to update_watermarks in .crtc_enable Signed-off-by:
Matt Roper <matthew.d.roper@intel.com> Tested-by(IVB): Matt Roper <matthew.d.roper@intel.com> Signed-off-by:
Patrik Jakobsson <patrik.jakobsson@linux.intel.com> Reviewed-by:
-
Deepak S authored
Currently we update the freq before masking the interrupts, which can allow new interrupts to occur before the frequency has changed. These extra interrupts might waste some cpu cycles. This patch corrects this by masking interrupts prior to updating the frequency. Note from Chris: "Well it won't waste CPU cycles as the interrupt is also masked by the threshold limits, but there should be no harm at all in reordering the patch so, and it does make a certain amount of sense." Signed-off-by:
Deepak S <deepak.s@intel.com> Signed-off-by:
Praveen Paneri <praveen.paneri@intel.com> Reviewed-by:
-
Thulasimani,Sivakumar authored
Update the hotplug documentation to explain that hotplug storm is not expected for Display port panels and hence is not handled in current code. v2: update the statements as recommended by Daniel Signed-off-by:
Sivakumar Thulasimani <sivakumar.thulasimani@intel.com> Signed-off-by:
-
Ville Syrjälä authored
Since commit e6292556 Author: Vandana Kannan <vandana.kannan@intel.com> Date: Wed Jul 1 17:02:57 2015 +0530 drm/i915/bxt: BUNs related to port PLL BXT DPLL can now generate frequencies in the 216-223 MHz range. Adjust the HDMI port clock checks to account for the reduced range of invalid frequencies. Cc: Vandana Kannan <vandana.kannan@intel.com> Cc: Imre Deak <imre.deak@intel.com> Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by:
Imre Deak <imre.deak@intel.com> Signed-off-by:
-
Ville Syrjälä authored
We do the exact same steps around the disp2d/pipe A power well enable/disable on VLV and CHV. Refactor the shared code into some helpers. Note that this means we now call vlv_power_sequencer_reset() before turning off the power well, whereas before we did it after. That doesn't matter though since vlv_power_sequencer_reset() just resets the power sequencer software tracking and doesn't touch the hardware at all. Signed-off-by:
-
Ville Syrjälä authored
The pipe A power well is the "disp2d" well on CHV and pipe B and C wells don't even exist. Thereforce we can remove the checks for pipe A vs. others and just assume it's always pipe A. Signed-off-by:
-
Ville Syrjälä authored
Drop the spurious 'A' from the VLV/CHV ref clock enable define, and add the "REF" to the VLV ref clock selection bit. Also s/CLOCK/CLK/ for extra consistency. Signed-off-by:
-
Ville Syrjälä authored
We disable the DPLL VGA mode when enabling the DPLL, but we enaable it again when disabling the DPLL. Having VGA mode enabled even in unused DPLLs can cause problems for CHV, so it seems wiser to always keep it disabled. And let's just do that on all GMCH platforms to keep things as similar as possible between them. Signed-off-by:
-
Akash Goel authored
Updated the i915_ring_freq_table debugfs function to support the read of ring frequency table, through Punit interface, for SKL also. Issue: VIZ-5144 Signed-off-by:
Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by:
-
Akash Goel authored
Ring frequency table programming changes for SKL. No need for a floor on ring frequency, as the issue of performance impact with ring running below DDR frequency, is believed to be fixed on SKL v2: Removed the check for avoiding ring frequency programming for BXT (Rodrigo) Issue: VIZ-5144 Signed-off-by:
-
Akash Goel authored
Read the efficient frequency (aka RPe) value through the the mailbox command (0x1A) from the pcode, as done on Haswell and Broadwell. The turbo minimum frequency softlimit is not revised as per the efficient frequency value. v2: Replaced the conditional expression operator with 'if' statement (Tom) v3: Corrected the derivation of efficient frequency & shifted the GEN9_FREQ_SCALER multiplications downwards (Ville) Issue: VIZ-5143 Signed-off-by:
-
- 09 Jul, 2015 9 commits
-
-
Rodrigo Vivi authored
This fbdev restore mode was another corner case that was now calling frontbuffer flip and flush and making we miss screen updates with PSR enabled. So let's also add the invalidate hack here while we don't have a reliable dirty fbdev op. v2: As pointed by Paulo: removed seg fault risk, used fb_helper when possible and put brackets on if. Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by:Paulo Zanoni <paulo.r.zanoni@intel.com> Testcase: igt/kms_fbcon_fbt/psr Signed-off-by:
-
Rodrigo Vivi authored
fbdev_set_par is called when fbcon is taking over control. In the past frontbuffer was being invalidated on set_to_gtt_domain, but it moved to set_domain fixing that case, but left this behind and broken in commit 031b698a Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Fri Jun 26 19:35:16 2015 +0200 drm/i915: Unconditionally do fb tracking invalidate in set_domain Note that even before this commit it wasn't perfect since the invalidate was omitted if the fbcon was already in the GTT domain, which it usually was. Since we are also invalidating in other fbdev cases this one was masked here. At least until now that I found this corner case: On boot with plymouth doing a splash screen when returning to the console frontbuffer wans't being invalidated causing missed screen updates with PSR enabled. So this patch fixes this issue. v2: Make invalidate directly and unconditionally and fix commit message indicating the set_domain fix as pointed out by Daniel. v3: Remove unecessary if(obj) added by mistake Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by:
-
Rodrigo Vivi authored
Idle frames the number of identical frames needed before panel can enter PSR. There are some panels that requires up to minimum of 4 idle frames available on the market. For these cases usually VBT should be used to configure the number of idle frames, but unfortunately this isn't always true and VBT isn't being set at all. Let's trust VBT when it is set + 1 and use minimum of 4 + 1 when VBT isn't set. "+1" covers the "of-by-one" case. Signed-off-by:
-
Rodrigo Vivi authored
By Spec we should only mask memup and hotplug detection for hardware tracking cases. However we always masked LPSP because with power well always enabled on audio PSR was never being activated and residency was always zeroed. Apparently audio driver is tying power well management and runtime PM for some reason. But with audio runtime PM working or with audio completely out of picture we should remove this mask, otherwise we have a high risk of miss screen updates as faced by Matthew. WARNING: With this patch if snd_intel_hda driver is running and not releasing power well properly PSR will constant Exit and Performance Counter will be 0. But the best thing of this patch is that with one more HW tracking working the risks of missed blank screen are minimized at most. This affects just core platforms where PSR exit are also helped by HW tracking: Haswell, Broadwell and Skylake for now. v2: Fix commit message explanation. It has nothing to do with runtime PM on i915 as previously advertised. Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Matthew Garrett <mjg59@srcf.ucam.org> Signed-off-by:
-
Paulo Zanoni authored
Reported by the kbuild test robot. Regression introduced by: commit fdbff928 Author: Daniel Vetter <daniel.vetter@ffwll.ch> Date: Thu Jun 18 11:23:24 2015 +0200 drm/i915: Clear fb_tracking.busy_bits also for synchronous flips (I reviewed this commit, so it's also my fault) Signed-off-by:
-
Paulo Zanoni authored
So make it static. Signed-off-by:
-
Paulo Zanoni authored
Reported by the kbuild test robot. Regression introduced by: commit de152b62 Author: Rodrigo Vivi <rodrigo.vivi@intel.com> Date: Tue Jul 7 16:28:51 2015 -0700 drm/i915: Add origin to frontbuffer tracking flush (I reviewed this commit, so it's also my fault) Signed-off-by:
-
Rodrigo Vivi authored
Let's do a frontbuffer flush on dirty fb. To be used for DIRTYFB drm ioctl. This patch solves the biggest PSR known issue, that is missed screen updates during boot, mainly when there is a splash screen involved like Plymouth. Previously PSR was being invalidated by fbdev and Plymounth was taking control with PSR yet invalidated and could get screen updates normally. However with some atomic modeset changes Pymouth modeset over ioctl was now causing frontbuffer flushes making PSR gets back to work while it cannot track the screen updates and exit properly. By adding this flush on dirtyfb we properly track frontbuffer writes and properly exit PSR. Actually all mmap_wc users should call this dirty callback in order to have a proper frontbuffer tracking. In the future it can be extended to return 0 if the whole screen has being flushed or the number of rects flushed as Chris suggested. v2: Remove ORIGIN_FB_DIRTY and use ORIGIN_GTT instead since dirty callback is just called after few screen updates and not on everyone as pointed by Daniel. v3: Use flush instead of invalidate since flush means invalidate + flush and dirty means drawn had finished and it can be flushed. v4: Remove PSR from subject since it is purely frontbuffer tracking change and that can be useful for FBC as well. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: -
Rodrigo Vivi authored
Since flush actually means invalidate + flush we need to force psr exit on PSR flush. On Core platforms there is no way to disable hw tracking and do the pure sw tracking so we simulate it by fully disable psr and reschedule a enable back. So a good idea is to minimize sequential disable/enable in cases we know that HW tracking like when flush has been originated by a flip. Also flip had just invalidated it already. It also uses origin to minimize the a bit the amount of disable/enabled, mainly when flip already had invalidated. With this patch in place it is possible to do a flush on dirty areas properly in a following patch. v2: Remove duplicated exit on HSW+Sprites as pointed out by Paulo. Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by:
-
- 08 Jul, 2015 9 commits
-
-
Rodrigo Vivi authored
This will be useful to PSR and FBC once we start making dirty fb calls to also flush frontbuffer. Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by:
-
Arun Siluvery authored
wa_ctx_emit() depends on the name of a local variable; if the name of that variable is changed then we get compile errors. In this case it is unlikely to be changed as this macro is only used in this set of functions but Kernel coding guidelines doesn't recommend doing this. It was my mistake as I should have corrected it at the beginning but missed so correct this before there are more usages of this macro (Bob Beckett). https://www.kernel.org/doc/Documentation/CodingStyle, Chapter 12, "Things to avoid when using macros", point 2): " 2) macros that depend on having a local variable with a magic name: #define FOO(val) bar(index, val) might look like a good thing, but it's confusing as hell when one reads the code and it's prone to breakage from seemingly innocent changes. " v2: Optimization to avoid multiple evaluation of 'index' in the macro. Since we invoke it multiple times, compiler, if it can, should be able to coalesce them into a single condition and remove multiple WARN_ON checks (Chris). Suggested-by:
Robert Beckett <robert.beckett@intel.com> Cc: Robert Beckett <robert.beckett@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Imre Deak <imre.deak@intel.com> Signed-off-by:
Arun Siluvery <arun.siluvery@linux.intel.com> Reviewed-by:
-
Sonika Jindal authored
Writing to PCH_PORT_HOTPLUG for each interrupt is not required. Handle it only if hpd has actually occurred like we handle other interrupts. v2: Make few variables local to if block (Ville) v3: Add check for ibx/cpt both (Ville). While at it, remove the redundant check for hotplug_trigger from pch_get_hpd_pins v4: Indentation (Ville) Reviewed-by:Sonika Jindal <sonika.jindal@intel.com> Signed-off-by:
-
Paulo Zanoni authored
So now all the calls are inside __intel_fbc_update(). Consistency! Signed-off-by:
-
Paulo Zanoni authored
I have two separate refactor ideas that require extracting this to a separate function. I'm not sure which idea I'll end choosing, but since both will require extracting this function, let's do this now. Notice that this is just code moving. Any possible problems with the current multiple pipes check should be fixed in later commits. Signed-off-by:
-
Paulo Zanoni authored
The poor in_dbg_master() check was the only one without a reason string. Give it a reason string so it won't feel excluded. Signed-off-by:
-
Paulo Zanoni authored
This is all internal i915.ko work, let's start using intel_crtc for everything. Signed-off-by:
-
Paulo Zanoni authored
Because the cool kids use dev_priv and FBC wants to be cool too. We've been historically using struct drm_device on the FBC function arguments, but we only really need it for intel_vgpu_active(): we can use dev_priv everywhere else. So let's fully switch to dev_priv since I'm getting tired of adding "struct drm_device *dev = dev_priv->dev" everywhere. If I get a NACK here I'll propose the opposite: convert all the functions that currently take dev_priv to take dev. Signed-off-by:
-
Paulo Zanoni authored
Because it makes more sense there, IMHO. Signed-off-by:
-
- 07 Jul, 2015 2 commits
-
-
Daniel Vetter authored
After the register save/restore code is gone there's just one user left and it just obfuscates that one. Remove it. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Suggested-by:
-
Daniel Vetter authored
Burning cpu cycles isn't awesome, so use sleeps instead. Signed-off-by:
-
