drm/i915/gvt: throw error basing on execlist submit result

throw error message in elsp emulation handler basing on execlist
submit result. guest will trigger tdr process for recovering, gvt
just follow guest's desire.

v2: populate error to top of mmio emulation logic, comments from
zhenyu

Signed-off-by: Bing Niu <bing.niu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: add full vGPU reset support

Full vGPU reset need to release all the shadow PPGGT pages to avoid
unnecessary write-protect and also should re-initialize pvinfo after
resetting vregs to keep pvinfo correct.

Signed-off-by: Ping Gao <ping.a.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: correct the reset logic

The current_vgpu will set to NULL after stopping the scheduler when
the reset is triggered by current vgpu, so here need change the
judgement condition for current vgpu detection.

Signed-off-by: Ping Gao <ping.a.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: add vreg write for GDRST handler

The emulation handler for MMIO GDRST miss vreg write in it, as result
the vreg cannot update correspondingly.

Signed-off-by: Ping Gao <ping.a.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix detect_host calling logic

Like other routines, intel_gvt_hypervisor_detect_host returns 0
for success.

Signed-off-by: Xiaoguang Chen <xiaoguang.chen@intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix an typo in skl_decode_mi_display_flip

Fix type to set correct pipe number.

Signed-off-by: Min He <min.he@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: get msi cap offset from pdev directly

Linux PCI driver saves the msi and msix capability offset
in pci_dev->msi_cap and pci_dev->msix_cap. We can use msi_cap
in pci_dev directly, no need hardcode.

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: use well wrapped set_mask_bits() instead of defining new one

Macro set_mask_bits() is ready for us, just invoke it and remove
our write_bits().

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix an error string format

It is better to use %p format for void pointers instead of casting them
because a void* is not necessary a 64 bits value.

Signed-off-by: Jérémy Lefaure <jeremy.lefaure@lse.epita.fr>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix bad 32 bit shift in gtt

Since ioread32 returns a 32-bit value, it is impossible to left-shift
this value by 32 bits (it produces a compilation error). Casting the
return value of ioread32 fix this issue.

Signed-off-by: Jérémy Lefaure <jeremy.lefaure@lse.epita.fr>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Fix broken mocs offset

Fix missed mocs offset reg setting for save/restore function.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Implement WaForceWakeRenderDuringMmioTLBInvalidate

When invalidating RCS TLB the device can enter RC6 state interrupting
the process, therefore the need for render forcewake for the whole
procedure.

This WA is needed for all production SKL SKUs.

v2: reworked putting and getting forcewake with help of Mika Kuoppala
v3: use I915_READ_FW and I915_WRITE_FW as we are handling forcewake on
    in the code path

References: HSD#2136899, HSD#1404391274
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915: GVT-g driver depends on 64BIT kernel

We currently don't support GVT-g driver on i386 kernel.
Add explicit dependence on 64bit kernel.

Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Fix failure when ACPI is not enabled

Directly use memremap instead of acpi function for opregion mapping.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: do not ignore return value of create_scratch_page

Function create_scratch_page() may fail in some cases.

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix spare warnings on odd constant _Bool cast

The function return values should has type int if it return
a integer value.

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: mark symbols static where possible

Mark all local functions & variables as static.

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: fix sparse warnings on different address spaces

Add proper __iomem annotation for pointers obtained via ioremap().

Signed-off-by: Du, Changbin <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: properly access enabled intel_engine_cs

Switch to use new for_each_engine() helper to properly access
enabled intel_engine_cs as i915 core has changed that to be
dynamic managed. At GVT-g init time would still depend on ring
mask to determine engine list as it's earlier.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Remove defunct vmap_batch()

This code was removed from i915_cmd_parser.c but still an obsolete
version wound up being duplicated into gvt/cmd_parser.c. Good riddance.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Use common mapping routines for shadow_bb object

We have the ability to map an object, so use it rather than opencode it
badly. Note that the object remains permanently pinned, this is poor
practise.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Use common mapping routines for indirect_ctx object

We have the ability to map an object, so use it rather than opencode it
badly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Stop waiting whilst holding struct_mutex

For whatever reason, the gvt scheduler runs synchronously. At the very
least, lets run synchronously without holding the struct_mutex.

v2: cut'n'paste mutex_lock instead of unlock.
Replace long hold of struct_mutex with a mutex to serialise the worker
threads.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Stop checking for impossible interrupts from a kthread

The kthread will not be interrupted, don't even bother checking.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Hold a reference on the request

The workload took a pointer to the request, and even waited upon,
without holding a reference on the request. Take that reference
explicitly and fix up the error path following request allocation that
missed flushing the request.

v2: [zhenyuw]
- drop request put in error path for dispatch, as main thread
caller will handle it identically to a real request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Remove dangerous unpin of backing storage of bound GPU object

Unpinning the pages prior to the object being release from the GPU may
allow the GPU to read and write into system pages (i.e. use after free
by the hw).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Use the returned VMA to provide the virtual address

The purpose of returning the just-pinned VMA is so that we can use the
information within, like its address. Also it should be tracked and used
as the cookie to unpin...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: i915_gem_object_create() returns an error pointer

On failure from i915_gem_object_create(), we need to check for an error
pointer not NULL.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Add runtime pm around fences

Manipulating the fence_list requires the runtime wakelock, as does
writing to the fence registers. Acquire a wakelock for the former, and
assert that the device is awake for the latter.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

Documentation/gpu: Add section for Intel GVT-g host support

Update with brief overview and reference for more detailed
arch design documents.

Add new section for Intel GVT-g host support.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Fix warning on obsolete function usage

Don't use obsolete drm_gem_object_unreference() but switch to i915_gem_object_put().

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

MAINTAINERS: Add new Intel GVT-g driver maintainer

This adds new item for Intel GVT-g driver maintainer info.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: clean up intel_gvt.h as interface for i915 core

i915 core should only call functions and structures exposed through
intel_gvt.h. Remove internal gvt.h and i915_pvinfo.h.

Change for internal intel_gvt structure as private handler which
not requires to expose gvt internal structure for i915 core.

v2: Fix per Chris's comment
- carefully handle dev_priv->gvt assignment
- add necessary bracket for macro helper
- forward declartion struct intel_gvt
- keep free operation within same file handling alloc

v3: fix use after free and remove intel_gvt.initialized

v4: change to_gvt() to an inline

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Fix build failure after intel_engine_cs change

Change GVT-g code reference for intel_engine_cs from static array to
allocated pointer after commit 3b3f1650b1ca ("drm/i915: Allocate
intel_engine_cs structure only for the enabled engines").

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161018014007.29369-1-zhenyuw@linux.intel.com

drm/i915: Suppress underruns during DP link retraining

DP link retraining causes (spurious?) underruns. We can't really avoid
them, except perhaps by doing a full modeset (which has its own underrun
suppression anyway). So let's just hide them.

MST still has its own logic for retrainin, but a bigger hpd handling
cleanup/unification is needed there anyway, so let's leave that be for now.

References: https://bugs.freedesktop.org/show_bug.cgi?id=98251
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476464574-32230-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Extract intel_crtc_pch_transcoder()

Extract the code to determine which PCH transcoder we're using to a
small helper.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476464574-32230-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: GMBUS don't need no forcewake

GMBUS is part of the display engine, and thus has no need for
forcewake. Let's not bother trying to grab it then.

I don't recall if the display engine suffers from system hangs
due to multiple accesses to the same "cacheline" in mmio space.
I hope not since we're no longer protected by the uncore lock
since commit 4e6c2d58ba86 ("drm/i915: Take forcewake once for
the entire GMBUS transaction")

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476272687-15070-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

Merge tag 'gvt-next-2016-10-14' of https://github.com/01org/gvt-linux into drm-intel-next-queued

Zhenyu Wang writes:

This is first pull request to merge GVT-g device model in i915
which contains core GVT-g device model work to virtualize GPU
resources. This tries to add feature of Intel GVT-g technology
for full GPU virtualization. This version will support KVM based
virtualization solution named as KVMGT.

More background is on official project home: https://01.org/igvt-g

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915/gen9: Cleanup skl_pipe_wm_active_state

This function is a wreck, let's help it get its life back together and
cleanup all of the copy pasta here.

Signed-off-by: Lyude <cpaul@redhat.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/gen9: Make skl_wm_level per-plane

Having skl_wm_level contain all of the watermarks for each plane is
annoying since it prevents us from having any sort of object to
represent a single watermark level, something we take advantage of in
the next commit to cut down on all of the copy paste code in here.

Changes since v1:
- Style nitpicks
- Fix accidental usage of i vs. PLANE_CURSOR
- Split out skl_pipe_wm_active_state simplification into separate patch

Signed-off-by: Lyude <cpaul@redhat.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/skl: Remove linetime from skl_wm_values

Next part of cleaning up the watermark code for skl. This is easy, since
it seems that we never actually needed to keep track of the linetime in
the skl_wm_values struct anyway.

Signed-off-by: Lyude <cpaul@redhat.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/skl: Move per-pipe ddb allocations into crtc states

First part of cleaning up all of the skl watermark code. This moves the
structures for storing the ddb allocations of each pipe into
intel_crtc_state, along with moving the structures for storing the
current ddb allocations active on hardware into intel_crtc.

Changes since v1:
- Don't replace alloc->start = alloc->end = 0;

Signed-off-by: Lyude <cpaul@redhat.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Fix cxsr_latency_table reorg

I have re-ordered some struct members in patch:

  commit 44a655cae3043453f9dd8076538712d52e2e0ce4
  Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
  Date:   Thu Oct 13 11:09:23 2016 +0100

      drm/i915: Shrink cxsr_latency_table

but that particular one is not initialized with named
initializers which broke it.

Move the bitfields back at the beginning. Space saving
is still there.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 44a655cae304 ("drm/i915: Shrink cxsr_latency_table")
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476453302-7580-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915/gtt: Free unused lower-level page tables

Since "Dynamic page table allocations" were introduced, our page tables
can grow (being dynamically allocated) with address space range usage.
Unfortunately, their lifetime is bound to vm. This is not a huge problem
when we're not using softpin - drm_mm is creating an upper bound on used
range by causing addresses for our VMAs to eventually be reused.

With softpin, long lived contexts can drain the system out of memory
even with a single "small" object. For example:

bo = bo_alloc(size);
while(true)
    offset += size;
    exec(bo, offset);

Will cause us to create new allocations until all memory in the system
is used for tracking GPU pages (even though almost all PTEs in this vm
are pointing to scratch).

Let's free unused page tables in clear_range to prevent this - if no
entries are used, we can safely free it and return this information to
the caller (so that higher-level entry is pointing to scratch).

v2: Document return value and free semantics (Joonas)
v3: No newlines in vars block (Joonas)
v4: Drop redundant local 'reduce' variable
v5: Handle CI fail with enable_ppgtt=2

Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1476360162-24062-3-git-send-email-michal.winiarski@intel.com

drm/i915/gtt: Split gen8_ppgtt_clear_pte_range

Let's use more top-down approach, where each gen8_ppgtt_clear_* function
is responsible for clearing the struct passed as an argument and calling
relevant clear_range functions on lower-level tables.
Doing this rather than operating on PTE ranges makes the implementation
of shrinking page tables quite simple.

v2: Drop min when calculating num_entries, no negation in 48b ppgtt
check, no newlines in vars block (Joonas)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1476360162-24062-2-git-send-email-michal.winiarski@intel.com

drm/i915: Remove unused "valid" parameter from pte_encode

We never used any invalid ptes, those were put in place for
a possibility of doing gpu faults. However our batchbuffers are not
restricted in length, so everything needs to be pointing to something
and thus out-of-bounds is pointing to scratch.

Remove the valid flag as it is always true.

v2: Expand commit msg, patch reorder (Mika)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1476360162-24062-1-git-send-email-michal.winiarski@intel.com

drm/i915: Fix mismatched INIT power domain disabling during suspend

Currently the display INIT power domain disabling/enabling happens in a
mismatched way in the suspend/resume_early hooks respectively. This can
leave display power wells incorrectly disabled in the resume hook if the
suspend sequence is aborted for some reason resulting in the
suspend/resume hooks getting called but the suspend_late/resume_early
hooks being skipped. In particular this change fixes "Unclaimed read
from register 0x1e1204" on BYT/BSW triggered from i915_drm_resume()->
intel_pps_unlock_regs_wa() when suspending with /sys/power/pm_test set
to devices.

Fixes: 85e90679335f ("drm/i915: disable power wells on suspend")
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: David Weinehall <david.weinehall@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476358446-11621-1-git-send-email-imre.deak@intel.com

drm/i915: Make IS_GEN macros only take dev_priv

Saves 1416 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476352990-2504-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Make INTEL_GEN only take dev_priv

Saves 968 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)
v3: Rebase.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_VALLEYVIEW only take dev_priv

Saves 944 bytes of .rodata strings and 128 bytes of .text.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_CHERRYVIEW only take dev_priv

Saves 864 bytes of .rodata strings and ~100 of .text.

v2: Add parantheses around dev_priv. (Ville Syrjala)
v3: Rebase.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_G4X only take dev_priv

Saves 472 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make HAS_L3_DPF only take dev_priv

Saves 472 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_BROXTON only take dev_priv

Saves 1392 bytes of .rodata strings.

Also change a few function/macro prototypes in i915_gem_gtt.c
from dev to dev_priv where it made more sense to do so.

v2: Add parantheses around dev_priv. (Ville Syrjala)
v3: Mention function prototype changes. (David Weinehall)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>

drm/i915: Make IS_SKYLAKE only take dev_priv

Saves 1016 bytes of .rodata strings and couple hundred of .text.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_KABYLAKE only take dev_priv

Saves 1320 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_HASWELL only take dev_priv

Saves 2432 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_BROADWELL only take dev_priv

Saves 1808 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_IVYBRIDGE only take dev_priv

Saves 848 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)
v3: Rebase.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make INTEL_DEVID only take dev_priv

Saves 4472 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make IS_GEN-range macro only take dev_priv

Saves 944 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Do not use INTEL_INFO(dev_priv)->ring_mask inside WARNs

Saves 1520 bytes of .rodata strings.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make HAS_RUNTIME_PM only take dev_priv

Saves 960 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make HAS_GMCH_DISPLAY only take dev_priv

More .rodata string saving by avoid __I915__ magic inside WARNs.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make INTEL_PCH_TYPE & co only take dev_priv

This saves 1872 bytes of .rodata strings.

v2:
 * Rebase.
 * Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Make HAS_DDI and HAS_PCH_LPT_LP only take dev_priv

This saves 3248 bytes of .rodata strings.

v2: Add parantheses around dev_priv. (Ville Syrjala)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915/gvt: Support GVT-g on Skylake

GVT-g supports Intel Skylake platform. Enable it.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU command scanner

This patch introduces a command scanner to scan guest command buffers.

Signed-off-by: Yulei Zhang <yulei.zhang@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU context switch

As different VM may configure different render MMIOs when executing
workload, to schedule workloads between different VM, the render MMIOs
have to be switched.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU schedule policy framework

This patch introduces a vGPU schedule policy framework, with a timer based
schedule policy module for now

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU workload scheduler

This patch introduces the vGPU workload scheduler routines.

GVT workload scheduler is responsible for picking and executing GVT workload
from current scheduled vGPU. Before the workload is submitted to host i915,
the guest execlist context will be shadowed in the host GVT shadow context.
the instructions in guest ring buffer will be copied into GVT shadow ring
buffer. Then GVT-g workload scheduler will scan the instructions in guest
ring buffer and submit it to host i915.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU workload submission

This patch introduces the vGPU workload submission logics.

Under virtualization environment, guest will submit workload through
virtual execlist submit port. The submitted workload load will be wrapped
into an gvt workload which will be picked by GVT workload scheduler and
executed on host i915 later.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU execlist virtualization

This patch introduces the vGPU execlist virtualization.

Under virtulization environment, HW execlist interface are fully emulated
including virtual CSB emulation, virtual execlist emulation. The framework
will emulate the virtual CSB according to the guest workload running status

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU display virtualization

This patch introduces the GVT-g display virtualization.

It consists a collection of display MMIO handlers, like power well register
handler, pipe register handler, plane register handler, which will emulate
all display MMIOs behavior to support virtual mode setting sequence for
guest.

Signed-off-by: Bing Niu <bing.niu@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU MMIO virtualization

This patch introduces the generic vGPU MMIO emulation intercept
framework.  The MPT modules will request GVT-g core logic to
emulate MMIO read/write through IO emulation operations
callback when hypervisor trapped a guest GTTMMIO read/write.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU PCI configuration space virtualization

This patch introduces vGPU PCI configuration space virtualization.

- Adjust the trapped GPFN(Guest Page Frame Number) window of virtual GEN
PCI BAR 0 when guest initializes PCI BAR 0 address.

- Emulate OpRegion when guest touches OpRegion.

- Pass-through a part of aperture to guest when guest initializes
aperture BAR.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU graphics memory virtualization

The vGPU graphics memory emulation framework is responsible for graphics
memory table virtualization. Under virtualization environment, a VM will
populate the page table entry with guest page frame number(GPFN/GFN), while
HW needs a page table filled with MFN(Machine frame number). The
relationship between GFN and MFN(Machine frame number) is managed by
hypervisor, while GEN HW doesn't have such knowledge to translate a GFN.

To solve this gap, shadow GGTT/PPGTT page table is introdcued.

For GGTT, the GFN inside the guest GGTT page table entry will be translated
into MFN and written into physical GTT MMIO registers when guest write
virtual GTT MMIO registers.

For PPGTT, a shadow PPGTT page table will be created and write-protected
translated from guest PPGTT page table.  And the shadow page table root
pointers will be written into the shadow context after a guest workload
is shadowed.

vGPU graphics memory emulation framework consists:

- Per-GEN HW platform page table entry bits extract/de-extract routines.
- GTT MMIO register emulation handlers, which will call hypercall to do
GFN->MFN translation when guest write GTT MMIO register
- PPGTT shadow page table routines, e.g. shadow create/destroy/out-of-sync

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU interrupt virtualization.

This patch introduces vGPU interrupt emulation framework.

The vGPU intrerrupt emulation framework is an event-based interrupt
emulation framework. It's responsible for emulating GEN hardware interrupts
during emulating other HW behaviour.

It consists several components:

- Descriptions of interrupt register bit
- Upper level <-> lower level interrupt mapping
- GEN HW IER/IMR/IIR register emulation routines
- Event-based interrupt propagation interface

When a GVT-g component wants to inject an interrupt to a VM during a
emulation, first it should specify the event needs to be emulated and the
framework will deal with the rest of emulation:

- Generating related virtual IIR bit according to virtual IER and IMRs,
- Generate related virtual upper level virtual IIR bit accodring to the
per-platform interrupt mapping
- Injecting a MSI to VM

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: trace stub

v2:
- Make checkpatch.pl happy(Joonas)

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Introduce basic vGPU life cycle management

A vGPU represents a virtual Intel GEN hardware, which consists following
virtual resources:

- Configuration space (virtualized)
- HW registers (virtualized)
- GGTT memory space (partitioned)
- GPU page table (shadowed)
- Fence registers (partitioned)

* virtualized: fully emulated by GVT-g.
* partitioned: Only a part of the HW resource is allowed to be accessed
by VM.
* shadowed: Resource needs to be translated and shadowed before getting
applied into HW.

This patch introduces vGPU life cycle management framework, which is
responsible for creating/destroying a vGPU and preparing/free resources
related to a vGPU.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: golden virtual HW state management

Each vGPU expects a golden virtual HW state, which is just the state after
system is freshly powered on. GVT-g will try to load the golden virtual HW
state via kernel firmware interface.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Introduce a framework for tracking HW registers.

This patch introduces a framework for tracking HW registers on different
GEN platforms.

Accesses to GEN HW registers from VMs will be trapped by hypervisor. It
will forward these emulation requests to GVT-g device model, which
requires this framework to search for related register descriptions.

Each MMIO entry in this framework describes a GEN HW registers, e.g.
offset, length, whether it contains RO bits, whether it can be accessed by
LRIs...and also emulation handlers for emulating register reading and
writing.

- Use i915 MMIO register definition & statement.(Joonas)

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: vGPU HW resource management

This patch introduces the GVT-g vGPU HW resource management. Under
GVT-g virtualizaion environment, each vGPU requires portions HW
resources, including aperture, hidden GM space, and fence registers.

When creating a vGPU, GVT-g will request these HW resources from host,
and return them to host after a vGPU is destroyed.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915: Allocate intel_engine_cs structure only for the enabled engines

With the possibility of addition of many more number of rings in future,
the drm_i915_private structure could bloat as an array, of type
intel_engine_cs, is embedded inside it.
struct intel_engine_cs engine[I915_NUM_ENGINES];
Though this is still fine as generally there is only a single instance of
drm_i915_private structure used, but not all of the possible rings would be
enabled or active on most of the platforms. Some memory can be saved by
allocating intel_engine_cs structure only for the enabled/active engines.
Currently the engine/ring ID is kept static and dev_priv->engine[] is simply
indexed using the enums defined in intel_engine_id.
To save memory and continue using the static engine/ring IDs, 'engine' is
defined as an array of pointers.
struct intel_engine_cs *engine[I915_NUM_ENGINES];
dev_priv->engine[engine_ID] will be NULL for disabled engine instances.

There is a text size reduction of 928 bytes, from 1028200 to 1027272, for
i915.o file (but for i915.ko file text size remain same as 1193131 bytes).

v2:
- Remove the engine iterator field added in drm_i915_private structure,
  instead pass a local iterator variable to the for_each_engine**
  macros. (Chris)
- Do away with intel_engine_initialized() and instead directly use the
  NULL pointer check on engine pointer. (Chris)

v3:
- Remove for_each_engine_id() macro, as the updated macro for_each_engine()
  can be used in place of it. (Chris)
- Protect the access to Render engine Fault register with a NULL check, as
  engine specific init is done later in Driver load sequence.

v4:
- Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris)
- Kill the superfluous init_engine_lists().

v5:
- Cleanup the intel_engines_init() & intel_engines_setup(), with respect to
  allocation of intel_engine_cs structure. (Chris)

v6:
- Rebase.

v7:
- Optimize the for_each_engine_masked() macro. (Chris)
- Change the type of 'iter' local variable to enum intel_engine_id. (Chris)
- Rebase.

v8: Rebase.

v9: Rebase.

v10:
- For index calculation use engine ID instead of pointer based arithmetic in
  intel_engine_sync_index() as engine pointers are not contiguous now (Chris)
- For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas)
- Use for_each_engine macro for cleanup in intel_engines_init() and remove
  check for NULL engine pointer in cleanup() routines. (Joonas)

v11: Rebase.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com

drm/i915: Merge duplicate gen4 and vlv/chv enable vblank callbacks

gen4/vlv/chv all use the same bits in pipestat to enable the vblank
interrupt, so they can share the same callbacks to enable/disable.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161007194953.15616-1-chris@chris-wilson.co.uk
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: fix a read size argument

We want to read 3 bytes here, but because the parenthesis are in the
wrong place we instead read:

sizeof(intel_dp->edp_dpcd) == sizeof(intel_dp->edp_dpcd)

which is one byte.

Fixes: fe5a66f91c88 ("drm/i915: Read PSR caps/intermediate freqs/etc. only once on eDP")
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: <drm-intel-fixes@lists.freedesktop.org>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161013085508.GJ16198@mwanda

drm/i915: Skip unbinding large unmappable global buffers

If the user requests a mappable binding to the global GTT, we will first
unbind an existing mapping if it doesn't match. We will unbind even if
there is no possibility that the object can fit in the mappable
aperture. This may lead to a ping-pong migration of the object, for
example igt/gem_exec_big.

v2: Comment upon the reasoning, or lack thereof!, behind the choice of
magic numbers.

Testcase: igt/gem_exec_big
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20161013085504.30705-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com

drm/i915: Fix misplaced '\n' in printing the GPU error's RING_HEAD

'\n' is supposed to be at the end of the line, not in the middle.

Fixes: cdb324bde570 ("drm/i915: Show bounds of active request in the ring...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161013101815.26978-2-chris@chris-wilson.co.uk

drm/i915: Record the current requests queue for execlists upon hang

Mika wanted to know what requests were pending at the time of a hang as
we now track which requests we have submitted to the hardware.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161013101815.26978-1-chris@chris-wilson.co.uk

drm/i915: Shrink TV modes const data

Make struct video_levels and struct tv_mode use data types
of sufficient width to save approximately one kilobyte in
the .rodata section.

v2: Do not align struct members. (Jani Nikula, Joonas Lahtinen)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476353366-13931-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Shrink per-platform watermark configuration

Use types of more appropriate size in struct
intel_watermark_params to save 512 bytes of .rodata.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Shrink sdvo_cmd_names

Pack the struct _sdvo_cmd_name to save 736 bytes of .rodata.

This is fine since the name pointers are used only for debug.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Shrink cxsr_latency_table

unsigned long is too wide - use smaller types in
struct cxsr_latency to save 800-something bytes of .rodata.

v2: All data even fits in u16 for even more saving. (Ville Syrjala)
v3: Move bitfields to the end of the struct. (Joonas Lahtinen)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/hsw: Fix GPU hang during resume from S3-devices state

Currently resuming on HSW from S3 pm_test/devices state leads to an
unrecoverable GPU hang. Resetting the GPU during suspend fixes this. For
a full S3 cycle this change only means the reset happens earlier (before
reaching S3). For S4 the reset will happen now both during the freeze
and quiesce phases, which is a benefit since it will guarantee that the
GPU is idle before creating and loading the hibernation image.

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1476283597-580-1-git-send-email-imre.deak@intel.com

drm/i915: Treat a framebuffer reference as an active reference whilst shrinking

Treat a framebuffer reference with the same priority as an active
reference whilst shrinking. Framebuffers are likely to be reused and
typically cost more to migrate to and from GPU memory (on LLC
architectures we need to clflush), so defer the temptation to purge them
during a kswapd run until we have run out of cheap buffers.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: John Harrison <john.c.harrison@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012124824.23521-1-chris@chris-wilson.co.uk

drm/i915: Update debugfs describe_obj() to show fault-mappable

The current meaning of whether an object has a GGTT vma is very
ill-defined (and note we don't check for any partials either), it just
means that at some point it was in the GGTT but it may not be now. The
information we really care about here is whether it is taking up
precious mappable aperture space. This is the obj->fault_mappable flag.
We have a redundant long form reprinting of this information, so remove
that in favour of the compact flag.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012114827.17031-2-chris@chris-wilson.co.uk

drm/i915: Use fence_write() from rpm resume

During rpm resume we restore the fences, but we do not have the
protection of struct_mutex. This rules out updating the activity
tracking on the fences, and requires us to rely on the rpm as the
serialisation barrier instead.

[  350.298052] [drm:intel_runtime_resume [i915]] Resuming device
[  350.308606]
[  350.310520] ===============================
[  350.315560] [ INFO: suspicious RCU usage. ]
[  350.320554] 4.8.0-rc8-bsw-rapl+ #3133 Tainted: G     U  W
[  350.327208] -------------------------------
[  350.331977] ../drivers/gpu/drm/i915/i915_gem_request.h:371 suspicious rcu_dereference_protected() usage!
[  350.342619]
[  350.342619] other info that might help us debug this:
[  350.342619]
[  350.351593]
[  350.351593] rcu_scheduler_active = 1, debug_locks = 0
[  350.358952] 3 locks held by Xorg/320:
[  350.363077]  #0:  (&dev->mode_config.mutex){+.+.+.}, at: [<ffffffffa030589c>] drm_modeset_lock_all+0x3c/0xd0 [drm]
[  350.375162]  #1:  (crtc_ww_class_acquire){+.+.+.}, at: [<ffffffffa03058a6>] drm_modeset_lock_all+0x46/0xd0 [drm]
[  350.387022]  #2:  (crtc_ww_class_mutex){+.+.+.}, at: [<ffffffffa0305056>] drm_modeset_lock+0x36/0x110 [drm]
[  350.398236]
[  350.398236] stack backtrace:
[  350.403196] CPU: 1 PID: 320 Comm: Xorg Tainted: G     U  W       4.8.0-rc8-bsw-rapl+ #3133
[  350.412457] Hardware name: Intel Corporation CHERRYVIEW C0 PLATFORM/Braswell CRB, BIOS BRAS.X64.X088.R00.1510270350 10/27/2015
[  350.425212]  0000000000000000 ffff8801680a78c8 ffffffff81332187 ffff88016c5c5000
[  350.433611]  0000000000000001 ffff8801680a78f8 ffffffff810ca6da ffff88016cc8b0f0
[  350.442012]  ffff88016cc80000 ffff88016cc80000 ffff880177ad0000 ffff8801680a7948
[  350.450409] Call Trace:
[  350.453165]  [<ffffffff81332187>] dump_stack+0x67/0x90
[  350.458931]  [<ffffffff810ca6da>] lockdep_rcu_suspicious+0xea/0x120
[  350.466002]  [<ffffffffa039e8dd>] fence_update+0xbd/0x670 [i915]
[  350.472766]  [<ffffffffa039efe2>] i915_gem_restore_fences+0x52/0x70 [i915]
[  350.480496]  [<ffffffffa0368f42>] vlv_resume_prepare+0x72/0x570 [i915]
[  350.487839]  [<ffffffffa0369802>] intel_runtime_resume+0x102/0x210 [i915]
[  350.495442]  [<ffffffff8137f26f>] pci_pm_runtime_resume+0x7f/0xb0
[  350.502274]  [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40
[  350.509883]  [<ffffffff814401c5>] __rpm_callback+0x35/0x70
[  350.516037]  [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40
[  350.523646]  [<ffffffff81440224>] rpm_callback+0x24/0x80
[  350.529604]  [<ffffffff8137f1f0>] ? pci_restore_standard_config+0x40/0x40
[  350.537212]  [<ffffffff814417bd>] rpm_resume+0x4ad/0x740
[  350.543161]  [<ffffffff81441aa1>] __pm_runtime_resume+0x51/0x80
[  350.549824]  [<ffffffffa03889c8>] intel_runtime_pm_get+0x28/0x90 [i915]
[  350.557265]  [<ffffffffa0388a53>] intel_display_power_get+0x23/0x50 [i915]
[  350.565001]  [<ffffffffa03ef23d>] intel_atomic_commit_tail+0xdfd/0x10b0 [i915]
[  350.573106]  [<ffffffffa034b2e9>] ? drm_atomic_helper_swap_state+0x159/0x300 [drm_kms_helper]
[  350.582659]  [<ffffffff81615091>] ? _raw_spin_unlock+0x31/0x50
[  350.589205]  [<ffffffffa034b2e9>] ? drm_atomic_helper_swap_state+0x159/0x300 [drm_kms_helper]
[  350.598787]  [<ffffffffa03ef8a5>] intel_atomic_commit+0x3b5/0x500 [i915]
[  350.606319]  [<ffffffffa03061dc>] ? drm_atomic_set_crtc_for_connector+0xcc/0x100 [drm]
[  350.615209]  [<ffffffffa0306b49>] drm_atomic_commit+0x49/0x50 [drm]
[  350.622242]  [<ffffffffa034dee8>] drm_atomic_helper_set_config+0x88/0xc0 [drm_kms_helper]
[  350.631419]  [<ffffffffa02f94ac>] drm_mode_set_config_internal+0x6c/0x120 [drm]
[  350.639623]  [<ffffffffa02fa94c>] drm_mode_setcrtc+0x22c/0x4d0 [drm]
[  350.646760]  [<ffffffffa02f0f19>] drm_ioctl+0x209/0x460 [drm]
[  350.653217]  [<ffffffffa02fa720>] ? drm_mode_getcrtc+0x150/0x150 [drm]
[  350.660536]  [<ffffffff810c984a>] ? __lock_is_held+0x4a/0x70
[  350.666885]  [<ffffffff81202303>] do_vfs_ioctl+0x93/0x6b0
[  350.672939]  [<ffffffff8120f843>] ? __fget+0x113/0x200
[  350.678797]  [<ffffffff8120f735>] ? __fget+0x5/0x200
[  350.684361]  [<ffffffff81202964>] SyS_ioctl+0x44/0x80
[  350.690030]  [<ffffffff81001deb>] do_syscall_64+0x5b/0x120
[  350.696184]  [<ffffffff81615ada>] entry_SYSCALL64_slow_path+0x25/0x25

Note we also have to remember the lesson from commit 4fc788f5ee3d
("drm/i915: Flush delayed fence releases after reset") where we have to
flush any changes to the fence on restore.

v2: Replace call to release user mmaps with an assertion that they have
already been zapped.

Fixes: 49ef5294cda2 ("drm/i915: Move fence tracking from object to vma")
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012114827.17031-1-chris@chris-wilson.co.uk

drm/i915: Compress GPU objects in error state

Our error states are quickly growing, pinning kernel memory with them.
The majority of the space is taken up by the error objects. These
compress well using zlib and without decode are mostly meaningless, so
encoding them does not hinder quickly parsing the error state for
familiarity.

v2: Make the zlib dependency optional

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012090522.367-6-chris@chris-wilson.co.uk

drm/i915: Consolidate error object printing

Leave all the pretty printing to userspace and simplify the error
capture to only have a single common object printer. It makes the kernel
code more compact, and the refactoring allows us to apply more complex
transformations like compressing the output.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012090522.367-5-chris@chris-wilson.co.uk

drm/i915: Always use the GTT for error capture

Since the GTT provides universal access to any GPU page, we can use it
to reduce our plethora of read methods to just one. It also has the
important characteristic of being exactly what the GPU sees - if there
are incoherency problems, seeing the batch as executed (rather than as
trapped inside the cpu cache) is important.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161012090522.367-4-chris@chris-wilson.co.uk