drm/i915: Add a default case in gen7 hwsp switch-case

Gen7 won't get any new engines, and we already added VCS2 there to just
silence gcc's not handled in switch warnings.

Use a default case instead, otherwise we will need to keep adding extra
cases if changes happen in the future.

v2: Since reaching the default case is impossible, use GEM_BUG_ON (Chris).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170830180115.907-1-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Merge tag 'gvt-next-2017-09-08' of https://github.com/01org/gvt-linux into drm-intel-next-queued

gvt-next-2017-09-08

- PCI config sanitize series (Changbin)
- Workload submission error handling series (Fred)

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170908063155.l54lvpivxntjm7hq@zhen-hp.sh.intel.com

drm/i915/gvt: Add support for PCIe extended configuration space

IGD is PCIe device and has extended configuration space. Checking
the binary dump, we can see we have Caps located out of PCI compatible
Configuration Space range.

0x000: 86 80 12 19 17 04 10 00 06 00 00 03 00 00 00 00
0x010: 04 00 00 10 08 00 00 00 0c 00 00 00 08 00 00 00
0x020: 00 00 00 00 00 00 00 00 00 00 00 00 28 10 b9 06
0x030: 00 f8 ff ff 40 00 00 00 00 00 00 00 0b 01 00 00
0x040: 09 70 0c 01 71 26 01 62 c8 00 04 84 00 00 00 00
0x050: c1 00 00 00 39 00 00 00 00 00 00 00 01 00 00 a2
0x060: 00 00 01 00 00 00 00 00 00 00 00 00 00 00 00 00
0x070: 10 ac 92 00 00 80 00 10 00 00 00 00 00 00 00 00
0x080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x0a0: 00 00 00 00 00 00 00 00 00 00 00 00 05 d0 01 00
0x0b0: 18 00 e0 fe 00 00 00 00 00 00 00 00 00 00 00 00
0x0c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x0d0: 01 00 22 00 00 80 00 00 00 00 00 00 00 00 00 00
0x0e0: 00 00 00 00 00 00 00 00 00 80 00 00 00 00 00 00
0x0f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0x100: 1b 00 01 20 02 14 00 00 00 00 00 00 00 00 00 00
...

Currently, we only emulate the PCI compatible Configuration Space.
This is okay if we attach vGPU to PCI bus. But when we attach to
a PCI Express bus (when Qemu emulates a Intel Q35 chipset which has
PCIe slot), it will not work. Extended Configuration Space is required
for a PCIe device.

This patch extended the virtual configuration space from 256 bytes
to 4KB bytes. So we are to be a *real* PCIe device. And for the
Extended CapList we keep same to physical GPU.

Cc: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Fix incorrect PCI BARs reporting

Looking at our virtual PCI device, we can see surprising Region 4 and Region 5.
00:10.0 VGA compatible controller: Intel Corporation Sky Lake Integrated Graphics (rev 06) (prog-if 00 [VGA controller])
        ....
        Region 0: Memory at 140000000 (64-bit, non-prefetchable) [size=16M]
        Region 2: Memory at 180000000 (64-bit, prefetchable) [size=1G]
        Region 4: Memory at <ignored> (32-bit, non-prefetchable)
        Region 5: Memory at <ignored> (32-bit, non-prefetchable)
        Expansion ROM at febd6000 [disabled] [size=2K]

The fact is that we only implemented BAR0 and BAR2. Surprising Region 4 and
Region 5 are shown because we report their size as 0xffffffff. They should
report size 0 instead.

BTW, the physical GPU has a PIO BAR. GVTg hasn't implemented PIO access, so
we ignored this BAR for vGPU device.

v2: fix BAR size value calculation.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=1458032
Signed-off-by: Changbin Du <changbin.du@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/gvt: Refine error handling for perform_bb_shadow

fix the wrong return type and return error once the unknown
command is scanned.

v2:
- separate this error handle from healthy rating code. (Zhenyu)

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

drm/i915/gvt: Refine error handling in dispatch_workload

When an error occurs in dispatch_workload, this patch is to do the
proper cleanup and rollback to the original states before the workload
is abandoned.

v2:
- split the mixed several error paths for better review. (Zhenyu)

v3:
- original PTR_ERR(cs) is good and code cleanup. (Zhenyu)

v4:
- reuse the existing i915_add_request for error handling. (Zhenyu)

v5:
- remove the duplicate error handling release_shadow_wa_ctx and
  move the engine->context_unpin upper. (Zhenyu)

v6:
- keep the old label "out". (Zhenyu)

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

drm/i915/gvt: Refine error handling for intel_vgpu_pin_mm

When it is failed in shadow_mm, the pin_count should rollback
to the original states before return.

v2:
- split the mixed several error paths for better review. (Zhenyu)

v3:
  increase the pincount after shadow success. (Zhenyu)

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

drm/i915/gvt: Refine error handling for prepare_execlist_workload

refine the error handling for prepare_execlist_workload to restore to the
original states once error occurs.

only release the shadowed batch buffer and wa ctx when the workload is
completed successfully.

v2:
- split the mixed several error paths for better review. (Zhenyu)

v3:
- handle prepare batch buffer/wa ctx pin errors and
- emulate_schedule_in null issue. (Zhenyu)

v4:
- no need to handle emulate_schedule_in null issue. (Zhenyu)

v5:
- release the shadowed batch buffer and wa ctx only for the
  successful workload. (Zhenyu)

v6:
- polish the return style. (Zhenyu)

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

drm/i915/gvt: Add error handling for intel_gvt_scan_and_shadow_workload

When an error occurs after shadow_indirect_ctx, this patch is to do the
proper cleanup and rollback to the original states for shadowed indirect
context before the workload is abandoned.

v2:
- split the mixed several error paths for better review. (Zhenyu)

v3:
- no return check for clean up functions. (Changbin)

v4:
- expose and reuse the existing release_shadow_wa_ctx. (Zhenyu)

v5:
- move the release function to scheduler.c file. (Zhenyu)

v6:
- move error handling code of intel_gvt_scan_and_shadow_workload
  to here. (Zhenyu)

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

drm/i915/gvt: Separate cmd scan from request allocation

Currently i915 request structure and shadow ring buffer are allocated
before command scan, so it will have to restore to previous states once
any error happens afterwards in the long dispatch_workload path.

This patch is to introduce a reserved ring buffer created at the beginning
of vGPU initialization. Workload will be coped to this reserved buffer and
be scanned first, the i915 request and shadow ring buffer are only
allocated after the result of scan is successful.

To balance the memory usage and buffer alloc time, the coming bigger ring
buffer will be reallocated and kept until more bigger buffer is coming.

v2:
- use kmalloc for the smaller ring buffer, realloc if required. (Zhenyu)

v3:
- remove the dynamically allocated ring buffer. (Zhenyu)

v4:
- code style polish.
- kfree previous allocated buffer once kmalloc failed. (Zhenyu)

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

drm/i915/gvt: Add emulation for BAR2 (aperture) with normal file RW approach

For vfio-pci, if the region support MMAP then it should support both
mmap and normal file access. The user-space is free to choose which is
being used. For qemu, we just need add 'x-no-mmap=on' for vfio-pci
option.

Currently GVTg only support MMAP for BAR2. So GVTg will not work when
user turn on x-no-mmap option.

This patch added file style access for BAR2, aka the GPU aperture. We
map the entire aperture partition of active vGPU to kernel space when
guest driver try to enable PCI Memory Space. Then we redirect the file
RW operation from kvmgt to this mapped area.

Link: https://bugzilla.redhat.com/show_bug.cgi?id=1458032
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

drm/i915/kvmgt: Sanitize PCI bar emulation

For PCI, 64bit bar consumes two BAR registers, but this doesn't mean
both of two BAR are valid. Actually the second BAR is regarded as
reserved in this case. So we shouldn't emulate the second BAR.

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

drm/i915: Transform WaDisablePooledEuLoadBalancingFix into a simple register write

FF_SLICE_CS_CHICKEN2 does not belong to the context image.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-6-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Transform WaDisableDynamicCreditSharing into a simple register write

GAMT_CHKN_BIT_REG does not live in the context image.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-5-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Transform WaDisableGafsUnitClkGating into a simple reg write

GEN7_UCGCTL4 does not live in the context.

v2: Missing parenthesis

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-4-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: WaPushConstantDereferenceHoldDisable needs to modify a masked register

So do it correctly.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-3-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Transform WaDisableI2mCycleOnWRPort into a simple reg write

GAMT_CHKN_BIT_REG does not live in the context.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-2-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Transform WaInPlaceDecompressionHang into a simple reg write

Afaict, GEN9_GAMT_ECO_REG_RW_IA does not live in the context, so writing
it on every context creation is overkill (and wrong).

v2: Missing end parenthesis

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504798809-5653-1-git-send-email-oscar.mateo@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Apply the GTT write flush for all !llc machines

We also see the delayed GTT write issue on i915g/i915gm, so let's
presume that it is a universal problem for all !llc machines, and that we
just haven't yet noticed on g33, gen4 and gen5 machines.

v2: Use a register that exists on all platforms

Testcase: igt/gem_mmap_gtt/coherency # i915gm
References: https://bugs.freedesktop.org/show_bug.cgi?id=102577
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170907184520.5032-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

i915: Fix obj size vs. alignment for drm_pci_alloc()

drm_pci_alloc() refuses to cooperate if the passed alignment exceeds the
object size. So round up the obj size to the next power of two as well
to make this actually work.

Obviously things work just fine as long as the size was a power of two
to begin with. However kms_cursor_crc doesn't always use power of two
sizes so we hit a failure when we try to allocate the phys memory.

Testcase: igt/kms_cursor_crc
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170907143203.13055-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Disable mmio debugging during user access

If the user bypasses i915 and accesses mmio directly, that easily
confuses our automatic mmio debugging (any error we then detect is
likely to be as a result of the user). Since we expect userspace to open
debugfs/i915_forcewake_user if i915.ko is loaded and they want mmio
access, that makes the opportune time to disable our debugging for
duration of the bypass.

v2: Move the fiddling of uncore internals to uncore.c

References: https://bugs.freedesktop.org/show_bug.cgi?id=102543
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170907134441.12881-1-chris@chris-wilson.co.uk

drm/i915/skl+: debugfs entry to control IPC

This patch creates an entry in debugfs to check the status of IPC.
This can also be used to enable/disable IPC in supported platforms.

Changes since V1:
 - fix use of HAS_IPC
 - use kstrtobool_from_user (Maarten)
 - drm_info log, while enabling IPC (Maarten)

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-9-mahesh1.kumar@intel.com
[mlankhorst: enableddisabled -> yesno to match ipc write]

drm/i915/bxt+: Enable IPC support

This patch adds IPC support. This patch also enables IPC in all supported
platforms based on has_ipc flag.
IPC (Isochronous Priority Control) is the hardware feature, which
dynamically controls the memory read priority of Display.

When IPC is enabled, plane read requests are sent at high priority until
filling above the transition watermark, then the requests are sent at
lower priority until dropping below the level 0 watermark.
The lower priority requests allow other memory clients to have better
memory access. When IPC is disabled, all plane read requests are sent at
high priority.

Changes since V1:
 - Remove commandline parameter to disable ipc
 - Address Paulo's comments
Changes since V2:
 - Address review comments
 - Set ipc_enabled flag
Changes since V3:
 - move ipc_enabled flag assignment inside intel_ipc_enable function
Changes since V4:
 - Re-enable IPC after suspend/resume
Changes since V5:
 - Enable IPC for all gen >=9 except SKL
Changes since V6:
 - fix commit msg
 - after resume program IPC based on SW state.
Changes since V7:
 - Modify IPC support check based on HAS_IPC macro (suggested by Chris)

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-8-mahesh1.kumar@intel.com

drm/i915/gen9+: Add has_ipc flag in device info structure

New Isochronous Priority Control (IPC) capability is introduced in newer
GEN platforms. This patch adds a device info flag to indicate if platform
supports IPC. Patch also sets this flag in supported platforms.

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-7-mahesh1.kumar@intel.com

drm/i915/cnl: Extend WM workaround with IPC for CNL

CNL:A & CNL:B have same workaround as KBL to increase wm level latency
by 4us if IPC is enabled.

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-6-mahesh1.kumar@intel.com

drm/i915/glk: IPC linetime watermark workaround for GLK

IF IPC is enabled LINETIME_WM value should be half of calculated value
 line time = ROUNDDOWN(1/2 * Calculated Line Time)

Earlier code was rounding-up the value, But updated Bspec says we should
take the ROUNDDOWN. This patch corrects that as well.

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-5-mahesh1.kumar@intel.com

drm/i915/gen10: Calculate and enable transition WM

GEN > 9 require transition WM to be programmed if IPC is enabled.
This patch calculates & enable transition WM for supported platforms.
If transition WM is enabled, Plane read requests are sent at high
priority until filling above the transition watermark, then the
requests are sent at lower priority until dropping below the level-0 WM.
The lower priority requests allow other memory clients to have better
memory access.

transition minimum is the minimum amount needed for trans_wm to work to
ensure  the demote does not happen before enough data has been read to
meet the level 0 watermark requirements.

transition amount is configurable value. Higher values will
tend to cause longer periods of high priority reads followed by longer
periods of lower priority reads. Tuning to lower values will tend to
cause shorter periods of high and lower priority reads.

Keeping transition amount to 10 in this patch, as suggested by HW team.

Changes since V1:
 - Address review comments from Maarten

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-4-mahesh1.kumar@intel.com

drm/i915/skl+: Optimize WM calculation

Plane configuration parameters doesn't change for each WM-level
calculation. Currently we compute same parameters 8 times for each
wm-level.
This patch optimizes it by calculating these parameters in beginning
& reuse during each level-wm calculation.

Changes since V1:
 - rebase on top of Rodrigo's series for CNL

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-3-mahesh1.kumar@intel.com

drm/i915: Fixed point fixed16 wrapper cleanup

As per suggestion from Jani, cleanup the code. Cleanup includes
 - Instead of left shifting & check, compare with U32/16_MAX
 - Use typecast instead of clamp_t

Signed-off-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170817134529.2839-2-mahesh1.kumar@intel.com

tools/testing/scatterlist: Test new __sg_alloc_table_from_pages

Exercise the new __sg_alloc_table_from_pages API (and through
it also the old sg_alloc_table_from_pages), checking that the
created table has the expected number of segments depending on
the sequence of input pages and other conditions.

v2: Move to data driven for readability.
v3: Add some more testcases and -fsanitize=undefined. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906145506.14952-1-tvrtko.ursulin@linux.intel.com
[tursulin: whitespace fixup]

drm/i915: Use __sg_alloc_table_from_pages for userptr allocations

With the addition of __sg_alloc_table_from_pages we can control
the maximum coalescing size and eliminate a separate path for
allocating backing store here.

Similar to 871dfbd67d4e ("drm/i915: Allow compaction upto
SWIOTLB max segment size") this enables more compact sg lists to
be created and so has a beneficial effect on workloads with many
and/or large objects of this class.

v2:
 * Rename helper to i915_sg_segment_size and fix swiotlb override.
 * Commit message update.

v3:
 * Actually include the swiotlb override fix.

v4:
 * Regroup parameters a bit. (Chris Wilson)

v5:
 * Rebase for swiotlb_max_segment.
 * Add DMA map failure handling as in abb0deacb5a6
   ("drm/i915: Fallback to single PAGE_SIZE segments for DMA remapping").

v6: Handle swiotlb_max_segment() returning 1. (Joonas Lahtinen)

v7: Rebase.
v8: Commit spelling fix.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170803091417.23677-1-tvrtko.ursulin@linux.intel.com

lib/scatterlist: Introduce and export __sg_alloc_table_from_pages

Drivers like i915 benefit from being able to control the maxium
size of the sg coalesced segment while building the scatter-
gather list.

Introduce and export the __sg_alloc_table_from_pages function
which will allow it that control.

v2: Reorder parameters. (Chris Wilson)
v3: Fix incomplete reordering in v2.
v4: max_segment needs to be page aligned.
v5: Rebase.
v6: Rebase.
v7: Fix spelling in commit and mention max segment size in
    __sg_alloc_table_from_pages kerneldoc. (Andrew Morton)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: linux-kernel@vger.kernel.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20170803091351.23594-1-tvrtko.ursulin@linux.intel.com

lib/scatterlist: Avoid potential scatterlist entry overflow

Since the scatterlist length field is an unsigned int, make
sure that sg_alloc_table_from_pages does not overflow it while
coalescing pages to a single entry.

v2: Drop reference to future use. Use UINT_MAX.
v3: max_segment must be page aligned.
v4: Do not rely on compiler to optimise out the rounddown.
    (Joonas Lahtinen)
v5: Simplified loops and use post-increments rather than
    pre-increments. Use PAGE_MASK and fix comment typo.
    (Andy Shevchenko)
v6: Commit spelling fix.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: linux-kernel@vger.kernel.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170803091312.22875-1-tvrtko.ursulin@linux.intel.com

lib/scatterlist: Fix offset type in sg_alloc_table_from_pages

Scatterlist entries have an unsigned int for the offset so
correct the sg_alloc_table_from_pages function accordingly.

Since these are offsets withing a page, unsigned int is
wide enough.

Also converts callers which were using unsigned long locally
with the lower_32_bits annotation to make it explicitly
clear what is happening.

v2: Use offset_in_page. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Pawel Osciak <pawel@osciak.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Tomasz Stanislawski <t.stanislaws@samsung.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: linux-media@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> (v1)
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170731185512.20010-1-tvrtko.ursulin@linux.intel.com

drm/i915: Disable snooping (userptr, set-cache-level) on gen4

The original gen4 has an issue where writes (both render and blt) into
snoopable pages are lost. We've previously worked around this in
userspace (ddx, igt) by simply not requesting snoopable buffers, but upon
rediscovering this problem for a third time, make the kernel reject such
requests with -ENODEV.

This disables snooping on userspace buffers for i965g and i965gm (original
gen4) machines.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906192424.26970-1-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20170907

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/cnl: WaThrottleEUPerfToAvoidTDBackPressure:cnl(pre-prod)

Wa for B-stepping only.

A for a hang issue that requires throttling EU performace
to 12.5% to avoid back pressure to thread dispatch

v2: Rebased. No change from v1.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Oscar Mateo <oscar.mateo@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906220325.24524-1-rodrigo.vivi@intel.com

drm/i915: Lift has-pinned-pages assert to caller of ____i915_gem_object_get_pages

i915_gem_object_attach_phys() is trying to swap out its shmemfs pages
for a new set of physically contiguous pages, but unfortunately triggers
an assert inside get-pages.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102561
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906135220.13508-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Display WA #1133 WaFbcSkipSegments:cnl, glk

Skip compressing 1 segment at the end of the frame,
avoid a pixel count mismatch nuke event when last active
pixel and dummy pixel has same color for Odd Plane
Width / Height.

For both platforms Gemini Lake and Cannon Lake.

v2: Use function-like macro and also use mask to clean
    to make sure bit 11 is 0. (Suggested by Paulo).
v3: Add Display WA notation and also apply for GLK.
    Both Forgotten on v2.
    Using "GLK_" prefix since GLK came before CNL.
v4: Forgot to "|=" when moving directly macro to masked
    val. (Noticed by Paulo.)
v5: Rebased on top of 0a46ddd57c9e ("drm/i915/cnp: Wa 1181:
    Fix Backlight issue")

Cc: Imre Deak <imre.deak@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170905193013.31710-1-rodrigo.vivi@intel.com

drm/i915/cnl: Allow the reg_read ioctl to read the RCS TIMESTAMP register

This enables the Mesa driver to advertise support for ARB_timer_query, and
thus an OpenGL version higher than 3.2.

Suggested-by: Ben Widawsky <ben@bwidawsk.net>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Nanley Chery <nanley.g.chery@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170905184507.30046-1-nanley.g.chery@intel.com

drm/i915: Move device_info.has_snoop into the static tables

Currently we define any !llc machine as using snoop instead. However,
some platforms run into trouble using snoop that we would like to
disable, and to do so easily we want to be able to use the static
device_info tables.

v2: Leave the old snoop = !llc as a warning for the time being to check
that all stanzas are filled as either llc or snoop.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906105653.3665-1-chris@chris-wilson.co.uk

drm/i915: Disable MI_STORE_DATA_IMM for i915g/i915gm

The early gen3 machines (i915g/Grantsdale and i915gm/Alviso) share a lot
of characteristics in their MI/GTT blocks with gen2, and in particular
can only use physical addresses in MI_STORE_DATA_IMM. This makes it
incompatible with our usage, so include those two machines in the
blacklist to prevent usage.

v2: Make it easy for gcc and rewrite it as a switch to save some space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20170906152859.5304-1-chris@chris-wilson.co.uk

drm/i915: Re-enable GTT following a device reset

Ville Syrjälä spotted that PGETBL_CTL was losing its enable bit upon a
reset. That was causing the display to show garbage on his 945gm. On my
i915gm the effect was far more severe; re-enabling the display following
the reset without PGETBL_CTL being enabled lead to an immediate hard
hang.

We do have a routine to re-enable PGETBL_CTL which is applicable to
gen2-4, although on gen4 it is documented that a graphics reset doesn't
alter the register (no such wording is given for gen3) and should be safe
to call to punch back in the enable bit. However, that leaves the question
of whether we need to completely re-initialise the register and the
rest of the GSM. For g33/pnv/gen4+, where we do have a configurable
page table, its contents do seem to be kept, and so we should be able to
recover without having to reinitialise the GTT from scratch (as prior to
g33, that register is configured by the BIOS and we leave alone except
for the enable bit).

This appears to have been broken by commit 5fbd0418eef2 ("drm/i915:
Re-enable GGTT earlier during resume on pre-gen6 platforms"), which
moved the intel_enable_gtt() from i915_gem_init_hw() (also used by
reset) to add it earlier during hw init and resume, missing the reset
path.

v2: Find the culprit, rearrange ggtt_enable to be before gem_init_hw to
match init/resume

Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: 5fbd0418eef2 ("drm/i915: Re-enable GGTT earlier during resume on pre-gen6 platforms")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101852
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Reviewed-by: Daniel Vetter <daniel@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20170906111405.27110-1-chris@chris-wilson.co.uk
Tested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/cnp: Wa 1181: Fix Backlight issue

This workaround fixes a CNL PCH bug when changing
backlight from a lower frequency to a higher frequency.

During random reboot cycles, display backlight seems to
be off/ dim for 2-3 mins.

The only functional change on this patch is to
set bit 13 of 0xC2020 for CNL PCH.

The rest of patch is organizing identation around
those bits definitions and re-organizing CFL workarounds.

v2: Only add the bit that matters without touching others
    around (Jani).
    Rebase on top of clock gating functions rename.

Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Arthur J Runyan <arthur.j.runyan@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170831045223.3960-1-rodrigo.vivi@intel.com

drm/i915: Annotate user relocs with __user

Add the missing __user to the urelocs cast to fix the following sparse
warning:
i915_gem_execbuffer.c:1541:47: warning: cast removes address space of expression
i915_gem_execbuffer.c:1541:62: warning: incorrect type in argument 2 (different address spaces)
i915_gem_execbuffer.c:1541:62:    expected void const [noderef] <asn:1>*from
i915_gem_execbuffer.c:1541:62:    got char *

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: 2889caa92321 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901165434.24636-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #irc

drm/i915: Constify load detect mode

Make the mode used for load detection const, and adjust all relevant
functions to accept a const mode.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170518193837.393-2-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/perf: Remove __user from u64 in drm_i915_perf_oa_config

Sparse complains that these integers from which we form void __user *,
and so we don't need the annotation itself inside the uABI.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901145729.21363-2-chris@chris-wilson.co.uk
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

drm/i915: Silence sparse by using gfp_t

Sparse enforces that GFP flags are only manipulated inside gfp_t locals.

Fixes: 4d470f7359c4 ("drm/i915: Avoid undefined behaviour of "u32 >> 32"")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901145729.21363-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Merge drm-upstream/drm-next into drm-intel-next-queued

Catch up with upstream while it's easy.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: io unmap functions want __iomem

Don't cast away the __iomem from the io_mapping functions so that
sparse won't be so unhappy when we pass the pointer to the unmap
functions. Instead let's move the cast to where we actually use the
pointer.

Fixes the following sparse warnings:
i915_gem.c:1022:33: warning: incorrect type in argument 1 (different address spaces)
i915_gem.c:1022:33:    expected void [noderef] <asn:2>*vaddr
i915_gem.c:1022:33:    got void *[assigned] vaddr
i915_gem.c:1027:34: warning: incorrect type in argument 1 (different address spaces)
i915_gem.c:1027:34:    expected void [noderef] <asn:2>*vaddr
i915_gem.c:1027:34:    got void *[assigned] vaddr
i915_gem.c:1199:33: warning: incorrect type in argument 1 (different address spaces)
i915_gem.c:1199:33:    expected void [noderef] <asn:2>*vaddr
i915_gem.c:1199:33:    got void *[assigned] vaddr
i915_gem.c:1204:34: warning: incorrect type in argument 1 (different address spaces)
i915_gem.c:1204:34:    expected void [noderef] <asn:2>*vaddr
i915_gem.c:1204:34:    got void *[assigned] vaddr

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901171252.31025-2-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Add __rcu to radix tree slot pointer

radix_tree_for_each_slot() wants an __rcu annotated pointer for the
slot. So let's add the annotation.

Fixes the following sparse warnings:
i915_gem.c:2217:9: warning: incorrect type in assignment (different address spaces)
i915_gem.c:2217:9:    expected void **slot
i915_gem.c:2217:9:    got void [noderef] <asn:4>**
i915_gem.c:2217:9: warning: incorrect type in assignment (different address spaces)
i915_gem.c:2217:9:    expected void **slot
i915_gem.c:2217:9:    got void [noderef] <asn:4>**
i915_gem.c:2217:9: warning: incorrect type in argument 1 (different address spaces)
i915_gem.c:2217:9:    expected void [noderef] <asn:4>**slot
i915_gem.c:2217:9:    got void **slot
i915_gem.c:2217:9: warning: incorrect type in assignment (different address spaces)
i915_gem.c:2217:9:    expected void **slot
i915_gem.c:2217:9:    got void [noderef] <asn:4>**

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: 96d776345277 ("drm/i915: Use a radixtree for random access to the object's backing storage")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901171252.31025-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Wake up the device for the fbdev setup

Our fbdev setup requires the device to be awake for access
through the GTT. If one boots without connected displays and
later plugs one in, we won't have any runtime PM references when
the fbdev setup runs. Explicitly grab a runtime PM reference during
the fbdev setup to avoid the following spew:

[   62.518435] RPM wakelock ref not held during HW access
[   62.518459] ------------[ cut here ]------------
[   62.518546] WARNING: CPU: 3 PID: 37 at ../drivers/gpu/drm/i915/intel_drv.h:1800 i915_vma_pin_iomap+0x144/0x150 [i915]
[   62.518585] Modules linked in: i915 i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops drm intel_gtt agpgart netconsole nls_iso8859_1 nls_cp437 vfat fat efi_pstore coretemp hwmon intel_rapl x86_pkg_temp_thermal e1000e efivars ptp pps_core video evdev ip_tables x_tables ipv6 autofs4
[   62.518741] CPU: 3 PID: 37 Comm: kworker/3:1 Not tainted 4.13.0-rc7-skl+ #1077
[   62.518770] Hardware name:                  /NUC7i5BNB, BIOS BNKBL357.86A.0048.2017.0704.1415 07/04/2017
[   62.518827] Workqueue: events i915_hotplug_work_func [i915]
[   62.518853] task: ffff88046c00dc00 task.stack: ffffc90000184000
[   62.518896] RIP: 0010:i915_vma_pin_iomap+0x144/0x150 [i915]
[   62.518919] RSP: 0018:ffffc90000187cc8 EFLAGS: 00010292
[   62.518942] RAX: 000000000000002a RBX: ffff880460044000 RCX: 0000000000000006
[   62.518969] RDX: 0000000000000006 RSI: ffffffff819c3e6f RDI: ffffffff819f1c0e
[   62.518996] RBP: ffffc90000187cd8 R08: ffff88046c00e4f0 R09: 0000000000000000
[   62.519022] R10: ffff8804669ca800 R11: 0000000000000000 R12: ffff880461d20000
[   62.519049] R13: ffffc90000187d48 R14: ffff880461d20000 R15: ffff880460044000
[   62.519076] FS:  0000000000000000(0000) GS:ffff88047ed80000(0000) knlGS:0000000000000000
[   62.519107] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   62.519130] CR2: 000056478ae213f0 CR3: 0000000002c0f000 CR4: 00000000003406e0
[   62.519156] Call Trace:
[   62.519190]  intelfb_create+0x176/0x360 [i915]
[   62.519216]  __drm_fb_helper_initial_config_and_unlock+0x1c7/0x3c0 [drm_kms_helper]
[   62.519251]  drm_fb_helper_hotplug_event.part.18+0xac/0xc0 [drm_kms_helper]
[   62.519282]  drm_fb_helper_hotplug_event+0x1a/0x20 [drm_kms_helper]
[   62.519324]  intel_fbdev_output_poll_changed+0x1a/0x20 [i915]
[   62.519352]  drm_kms_helper_hotplug_event+0x27/0x30 [drm_kms_helper]
[   62.519395]  i915_hotplug_work_func+0x24e/0x2b0 [i915]
[   62.519420]  process_one_work+0x1d3/0x6d0
[   62.519440]  worker_thread+0x4b/0x400
[   62.519458]  ? schedule+0x4a/0x90
[   62.519475]  ? preempt_count_sub+0x97/0xf0
[   62.519495]  kthread+0x114/0x150
[   62.519511]  ? process_one_work+0x6d0/0x6d0
[   62.519530]  ? kthread_create_on_node+0x40/0x40
[   62.519551]  ret_from_fork+0x27/0x40
[   62.519569] Code: c4 78 e6 e0 0f ff e9 08 ff ff ff 80 3d d5 bc 0c 00 00 0f 85 0b ff ff ff 48 c7 c7 d8 50 32 a0 c6 05 c1 bc 0c 00 01 e8 9d 78 e6 e0 <0f> ff e9 f1 fe ff ff 0f 1f 44 00 00 0f 1f 44 00 00 0f b6 87 98
[   62.519771] ---[ end trace 5fbe271f991a58ae ]---

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901195456.6386-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Add interface to reserve fence registers for vGPU

In the past, vGPU alloc fence registers by walking through mm.fence_list
to find fence which pin_count = 0 and vma is empty. vGPU may not find
enough fence registers this way. Because a fence can be bind to vma even
though it is not in using. We have found such failure many times these
days.

An option to resolve this issue is that we can force-remove fence from
vma in this case.

This patch added two new api to the fence management code:
 - i915_reserve_fence() will try to find a free fence from fence_list
   and force-remove vma if need.
 - i915_unreserve_fence() reclaim a reserved fence after vGPU has
   finished.

With this change, the fence management is more clear to work with vGPU.
GVTg do not need remove fence from fence_list in private.

v3: (Chris)
  - Add struct_mutex lock assertion.
  - Only count for unpinned fence.

v2: (Chris)
  - Rename the new api for symmetry.
  - Add safeguard to ensure at least 1 fence remained for host display.

Signed-off-by: Changbin Du <changbin.du@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1504512061-5892-1-git-send-email-changbin.du@intel.com
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Use correct path to trace include

The header comment in include/trace/define_trace.h specifies that the
TRACE_INCLUDE_PATH needs to be relative to the define_trace.h header
rather than the trace file including it. Most instances get that wrong
and work around it by adding the $(src) directory to the include path.

While this works, it is preferable to refer to the correct path to the
trace file in the first place and avoid any workaround.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901144954.19620-4-thierry.reding@gmail.com

drm/i915: Fix the missing PPAT cache attributes on CNL

Add back the GEN8_PPAT_WB cache attributes in cnl_setup_private_ppat(),
which are missed on CNL.

Fixes: 4e34935fcf69 ("drm/i915/cnl: Setup PAT Index.")
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Suggested-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1504208177-27784-1-git-send-email-zhi.a.wang@intel.com

drm/i915: Fix enum pipe vs. enum transcoder for the PCH transcoder

Use enum pipe for PCH transcoders also in the FIFO underrun code.

Fixes the following new sparse warnings:
intel_fifo_underrun.c:340:49: warning: mixing different enum types
intel_fifo_underrun.c:340:49:     int enum pipe  versus
intel_fifo_underrun.c:340:49:     int enum transcoder
intel_fifo_underrun.c:344:49: warning: mixing different enum types
intel_fifo_underrun.c:344:49:     int enum pipe  versus
intel_fifo_underrun.c:344:49:     int enum transcoder
intel_fifo_underrun.c:397:57: warning: mixing different enum types
intel_fifo_underrun.c:397:57:     int enum pipe  versus
intel_fifo_underrun.c:397:57:     int enum transcoder
intel_fifo_underrun.c:398:17: warning: mixing different enum types
intel_fifo_underrun.c:398:17:     int enum pipe  versus
intel_fifo_underrun.c:398:17:     int enum transcoder

Cc: Matthias Kaehlcke <mka@chromium.org>
Fixes: a21960339c8c ("drm/i915: Consistently use enum pipe for PCH transcoders")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901143123.7590-3-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Make i2c lock ops static

Make gmbus_lock_ops and proxy_lock_ops static to appease sparse
intel_i2c.c:652:34: warning: symbol 'gmbus_lock_ops' was not declared. Should it be static?
intel_sdvo.c:2981:34: warning: symbol 'proxy_lock_ops' was not declared. Should it be static?

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Fixes: a85066840d29 ("drm/i915: Rework sdvo proxy i2c locking")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901143123.7590-2-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Make i9xx_load_ycbcr_conversion_matrix() static

Make i9xx_load_ycbcr_conversion_matrix() static to appease sparse:
intel_color.c:110:6: warning: symbol 'i9xx_load_ycbcr_conversion_matrix' was not declared. Should it be static?

Cc: Shashank Sharma <shashank.sharma@intel.com>
Fixes: 25edf91501b8 ("drm/i915: prepare csc unit for YCBCR420 output")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901143123.7590-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Eliminate crtc->state usage from intel_atomic_commit_tail and .crtc_update()

We already have the correct new crtc state so just use that instead of
crtc->state.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-7-ville.syrjala@linux.intel.com

drm/i915: Eliminate crtc->state usage from intel_update_pipe_config()

Pass the correct new crtc state to intel_update_pipe_config() instead
of using crtc->state.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-6-ville.syrjala@linux.intel.com

drm/i915: Eliminate obj->state usage from pre/post plane update

Dig up the appropriate new crtc and plane states from the top level
atomic state in intel_pre_plane_update() and intel_post_plane_update().

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-5-ville.syrjala@linux.intel.com

drm/i915: Pass proper old/new states to intel_plane_atomic_check_with_state()

Eliminate plane->state and crtc->state usage from
intel_plane_atomic_check_with_state() and its callers. Instead pass the
proper states in or dig them up from the top level atomic state.

Note that intel_plane_atomic_check_with_state() itself isn't allowed to
use the top level atomic state as there is none when it gets called from
the legacy cursor short circuit path.

v2: Rename some variables for easier comprehension (Maarten)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-4-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: rework IS_*_GT* macros

We can now make use of the intel_device_info.gt field.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170830161208.29221-4-lionel.g.landwerlin@intel.com

drm/i915: mark all device info struct with __initconst

As recommended by Chris.

v2: Switch from __initdata to __initconst.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170830161208.29221-3-lionel.g.landwerlin@intel.com

drm/i915: add GT number to intel_device_info

Up to Coffeelake we could deduce this GT number from the device ID.
This doesn't seem to be the case anymore. This change reorders pciids
per GT and adds a gt field to intel_device_info. We set this field on
the following platforms :

   - SNB/IVB/HSW/BDW/SKL/KBL/CFL/CNL

Before & After :

$ modinfo drivers/gpu/drm/i915/i915.ko | grep ^alias | wc -l
209

v2: Add SNB & IVB (Chris)

v3: Fix compilation error in early-quirks (Lionel)

v4: Fix inconsistency between FEATURE/PLATFORM macros (Ville)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170830161208.29221-2-lionel.g.landwerlin@intel.com

drm/i915/edp: Increase T12 panel delay to 900 ms to fix DP AUX CH timeouts

This patch fixes the DP AUX CH timeouts observed during CI runs causing
CI Failures on a specific PCI device. This issue was fixed previously
by adding a quirk but looks like we need to increase this delay even more
in order to get rid all the DP AUX CH timeouts.

Fixes: c99a259b4b5192ba ("drm/i915/edp: Add a T12 panel delay quirk to fix
DP AUX CH timeouts")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101144
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Cc: Clinton Taylor <clinton.a.taylor@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Tomi Sarvela <tomi.p.sarvela@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1502823591-25310-1-git-send-email-manasi.d.navare@intel.com

drm/i915: Eliminate obj->state usage in g4x/vlv/chv wm computation

Use explicit old/new states instead of relying on obj->state.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-4-ville.syrjala@linux.intel.com

drm/i915: Pass the crtc state explicitly to intel_pipe_update_start/end()

Pass the appropriate new crtc state explicitly to
intel_pipe_update_start/end() instead of of mucking around with
crtc->state.

v2: The mmio flip stuff is gone

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #v1
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-2-ville.syrjala@linux.intel.com

drm/i915: Pass the new crtc state to color management code

In an effort to eliminate the obj->state usage let's pass on the
new crtc state pointer (which we already have!) to the color management
code.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170823152226.22938-1-ville.syrjala@linux.intel.com

drm/i915: Consolidate max_cdclk_freq check in intel_crtc_compute_min_cdclk()

Currently the .modeset_calc_cdclk() hooks check the final cdclk value
against the max allowed. That's not really sufficient since the low
level calc_cdclk() functions effectively clamp the minimum required
cdclk to the max supported by the platform. Hence if the minimum
required exceeds the platforms capabilities we'd keep going anyway
using the max cdclk frequency.

To fix that let's move the check earlier into
intel_crtc_compute_min_cdclk() and we'll check the minimum required
cdclk of the pipe against the maximum supported by the platform.

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170710193347.8734-2-ville.syrjala@linux.intel.com
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>

drm/i915: Track minimum acceptable cdclk instead of "minimum dotclock"

Make the min_pixclk thing less confusing by changing it to track
the minimum acceptable cdclk frequency instead. This means moving
the application of the guardbands to a slightly higher level from
the low level platform specific calc_cdclk() functions.

The immediate benefit is elimination of the confusing 2x factors
on GLK/CNL+ in the audio workarounds (which stems from the fact
that the pipes produce two pixels per clock).

v2: Keep cdclk higher on CNL to workaround missing DDI clock voltage handling
v3: Squash with the CNL cdclk limits patch (DK)
v4: s/intel_min_cdclk/intel_pixel_rate_to_cdclk/ (DK)

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170830185703.8189-1-ville.syrjala@linux.intel.com

drm/i915/cnl: Fix DP max voltage

On clock recovery this function is called to find out
the max voltage swing level that we could go.

However gen 9 functions use the old buffer translation tables
to figure that out. That table is not valid for CNL
causing an invalid number of entries and an invalid selection
on the max voltage swing level.

v2: Let's use same approach that previous platforms.
v3: Actually use n_entries and avoid duplicated -1.
v4: Avoid cnl_max_level and use current style.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Clint Taylor <clinton.a.taylor@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170831145356.15932-1-rodrigo.vivi@intel.com

drm/i915/cnl: Fix DDI hdmi level selection.

Let's get a proper HDMI DDI entry level for vswing programming
sequences on CNL.

Spec doesn't specify any default for HDMI tables,
so let's pick the last entry as the default for now.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-7-rodrigo.vivi@intel.com

drm/i915/cnl: Move ddi buf trans related functions up.

No functional changes. But those functions will be needed
to get max level for HDMI and DP, so let's move those
up closer to other similar functions existent for previous
platforms.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-6-rodrigo.vivi@intel.com

drm/i915/cnl: Move voltage check into ddi buf trans functions.

Let's start converging CNL buf translations to same style
used on previous platforms. So first thing is to use the
standard signature so we don't need to propagate the voltage
check into other parts of the code, but only on the parts
that it is really useful.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-5-rodrigo.vivi@intel.com

drm/i915: Enable voltage swing before enabling DDI_BUF_CTL.

Sequences for DisplayPort asks us to
" Configure voltage swing and related IO settings.
Refer to DDI Buffer section."

before "Configure and enable DDI_BUF_CTL"

On BXT and CNL this means to execute the ddi vswing sequences.

At this point these sequences calls are getting duplicated for DP
because they are all called from DP link trainning sequences.

However this patch is not yet removing it before a futher discussion
since spec also allows that during link training without disabling
anything:

"
Notes
Changing voltage swing during link training:
Change the swing setting following the DDI Buffer section.
The port does not need to be disabled.
"

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-4-rodrigo.vivi@intel.com

drm/i915: Align vswing sequences with old ddi buffer registers.

Vswing sequences on BXT and CNL are equivalent
to the ddi buffer registers setting on other platforms.

For some reason it got aligned with skl_ddi_set_iboost what
is semantically incorrect. This forced us to keep skipping
ddi buffer translation tables on the platforms that has
the vswing sequences.

v2: Don't mess with DP signal levels on this patch.

Cc: Vandana Kannan <vandana.kannan@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-3-rodrigo.vivi@intel.com

drm/i915: decouple gen9 and gen10 dp signal levels.

Let's decouple bxt, glk and cnl dp signal levels
from other DDIs to avoid confusion.

No functional change. Only a reorg to avoid messing
with currently working DP signal levels when
moving voltage swing sequences around to match spec.

v2: ddi_signal_levels is also called from other ddi
    platforms, so don't remove IS_GEN9_BC check from
    skl_ddi_set_iboos. (Ville).

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-2-rodrigo.vivi@intel.com

drm/i915: Introduce intel_ddi_dp_level.

No functional changes. This only moves the DP level
selection to a separated function that will be later
used to organize better the vswing sequences.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829232230.23051-1-rodrigo.vivi@intel.com

drm/i915/cnl: Avoid ioremap_wc on Cannonlake as well.

Driver’s CPU access to GTT is via the GTTMMADR BAR.

The current HW implementation of that BAR is to only
support <= DW (and maybe QW) writes—not 16/32/64B writes
that could occur with WC and/or SSE/AVX moves.

GTTMMADR must be marked uncacheable (UC).
Accesses to GTTMMADR(GTT), must be 64 bits or less (ie. 1 GTT entry).

v2: Get clarification on the reasons and spec is getting
    updated to reflect it now.

Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Suggested-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829230907.21363-1-rodrigo.vivi@intel.com

drm/i915/cnl: WaDisableI2mCycleOnWRPort

On CNL B0 stepping GAM is not able to detect some deadlock
condition and then rise the rise the gam_coh_flush.

WA database and spec both mentions to set 4AB8[24]=1 as
workaround. Although register offset 0x4AB8 is not
documented for any platform.

References: HSD#1945815, BSID#1112

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829230751.21047-1-rodrigo.vivi@intel.com

drm/i915/cnl: WA FtrEnableFastAnisoL1BankingFix

WA to enable HW L1 Banking fix that allows aniso to operate
at full sample rate.

References: HSD#1937670

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Anuj Phogat <anuj.phogat@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829230723.20898-1-rodrigo.vivi@intel.com

drm/i915: Stop using long platform names on clock gating functions.

No functional changes.

Our code was only a bit messy with mixed style there so
let's clean up a bit using the short codenames for the platforms.

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829052026.15038-1-rodrigo.vivi@intel.com

drm/i915: Skip fence alignemnt check for the CCS plane

The CCS won't have the same stride as the main surface anyway so trying
to guard against the fence stride not matching the CCS stride is
not sensible. Just skip the fence vs. fb alignment check for the aux
plane.

Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Cc: Daniel Stone <daniels@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170824191100.10949-3-ville.syrjala@linux.intel.com
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Fixes: 2e2adb05736c ("drm/i915: Add render decompression support")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Treat fb->offsets[] as a raw byte offset instead of a linear offset

Userspace wants to treat fb->offsets[] as raw byte offsets into the gem
bo. Adjust the kernel code to match.

Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Cc: Daniel Stone <daniels@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170824191100.10949-2-ville.syrjala@linux.intel.com
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Fixes: 2e2adb05736c ("drm/i915: Add render decompression support")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Always wake the device to flush the GTT

Since we hold the device wakeref when writing through the GTT (otherwise
the writes would fail), we presumed that before the device sleeps those
writes would naturally be flushed and that we wouldn't need our mmio
read trick. However, that presumption seems false and a sleepy bxt seems
to require us to always manually flush the GTT writes prior to direct
access.

Fixes: e2a2aa36a509 ("drm/i915: Check we have an wake device before flushing GTT writes")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170829192546.1087-1-chris@chris-wilson.co.uk

drm/i915: Discard the request queue if we fail to sleep before suspend

If we fail to clear the outstanding request queue before suspending,
mark those requests as lost.

References: https://bugs.freedesktop.org/show_bug.cgi?id=102037
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170826110935.10237-3-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Clear wedged status upon resume

When we wake up from suspend, the device has been powered down and
should come back afresh. We should be able to safely remove the wedged
status from the previous session and start afresh.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170826110935.10237-2-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Always sanity check engine state upon idling

When we do a locked idle we know that afterwards all requests have been
completed and the engines have been cleared of tasks. For whatever
reason, this doesn't always happen and we may go into a suspend with
ELSP still full, and this causes an issue upon resume as we get very,
very confused.

If the engines refuse to idle, mark the device as wedged. In the process
we get rid of the maybe unused open-coded version of wait_for_engines
reported by Nick Desaulniers and Matthias Kaehlcke.

v2: Suppress the -EIO before suspend, but keep it for seqno wrap.

References: https://bugs.freedesktop.org/show_bug.cgi?id=101891
References: https://bugs.freedesktop.org/show_bug.cgi?id=102456
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthias Kaehlcke <mka@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20170826110935.10237-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

drm/i915: Don't use GPU relocations prior to cmdparser stalls

If we are using the cmdparser, we will have to copy the batch and so
stall for the relocations. Rather than prolong that stall by adding more
relocation requests, just use CPU relocations and do the stall upfront.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170826135620.25949-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Recreate vmapping even when the object is pinned

Sometimes we know we are the only user of the bo, but since we take a
protective pin_pages early on, an attempt to change the vmap on the
object is denied because it is busy. i915_gem_object_pin_map() cannot
tell from our single pin_count if the operation is safe. Instead we must
pass that information down from the caller in the manner of
I915_MAP_OVERRIDE.

This issue has existed from the introduction of the mapping, but was
never noticed as the only place where this conflict might happen is for
cached kernel buffers (such as allocated by i915_gem_batch_pool_get()).
Until recently there was only a single user (the cmdparser) so no
conflicts ever occurred. However, we now use it to allocate batches for
different operations (using MAP_WC on !llc for writes) in addition to the
existing shadow batch (using MAP_WB for reads).

We could either keep both mappings cached, or use a different write
mechanism if we detect a MAP_WB already exists (i.e. clflush
afterwards), but as we haven't seen this issue in the wild (it requires
hitting the GPU reloc path in addition to the cmdparser) for simplicity
just allow the mappings to be recreated.

v2: Include the i915_MAP_OVERRIDE bit in the enum so the compiler knows
about all the valid values.

Fixes: 7dd4f6729f92 ("drm/i915: Async GPU relocation processing")
Testcase: igt/gem_lut_handle # byt, completely by accident
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170828104631.8606-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Remove excess indent in intel_finish_reset() caught by sparse

  CHECK   drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_display.c:3753 intel_finish_reset() warn: inconsistent indenting

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170828104604.8552-1-chris@chris-wilson.co.uk
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Merge branch 'drm-vmwgfx-next' of git://people.freedesktop.org/~syeh/repos_linux into drm-next

vmwgfx add fence fd support.

* 'drm-vmwgfx-next' of git://people.freedesktop.org/~syeh/repos_linux:
  drm/vmwgfx: Bump the version for fence FD support
  drm/vmwgfx: Add export fence to file descriptor support
  drm/vmwgfx: Add support for imported Fence File Descriptor
  drm/vmwgfx: Prepare to support fence fd
  drm/vmwgfx: Fix incorrect command header offset at restart
  drm/vmwgfx: Support the NOP_ERROR command
  drm/vmwgfx: Restart command buffers after errors
  drm/vmwgfx: Move irq bottom half processing to threads
  drm/vmwgfx: Don't use drm_irq_[un]install

Merge tag 'exynos-drm-next-for-v4.14' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

Summary:
- Provide NV12MT pixel format support of Mixer driver in generic way.
- Refactor Exynos KMS drivers
  . Refactoring to panel detection way
  . Refactoring to setting up possible_crtcs
  . Refactoring to video and command mode support
- Some cleanups

* tag 'exynos-drm-next-for-v4.14' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: simplify set_pixfmt() in DECON and FIMD drivers
  drm/exynos: consistent use of cpp
  drm/exynos: mixer: remove src offset from mixer_graph_buffer()
  drm/exynos: mixer: simplify mixer_graph_buffer()
  drm/exynos: mixer: simplify vp_video_buffer()
  drm/exynos: mixer: enable NV12MT support for the video plane
  drm/exynos: mixer: fix chroma comment in vp_video_buffer()
  arm64: dts: exynos: remove i80-if-timings nodes
  dt-bindings: exynos5433-decon: remove i80-if-timings property
  drm/exynos/decon5433: use mode info stored in CRTC to detect i80 mode
  drm/exynos: add mode_valid callback to exynos_drm
  drm/exynos/decon5433: refactor irq requesting code
  drm/exynos/mic: use mode info stored in CRTC to detect i80 mode
  drm/exynos/dsi: propagate info about command mode from panel
  drm/exynos/dsi: refactor panel detection logic
  drm/exynos: use helper to set possible crtcs
  drm/exynos/decon5433: use readl_poll_timeout helpers

Merge tag 'drm-misc-next-fixes-2017-08-28' of git://anongit.freedesktop.org/git/drm-misc into drm-next

UAPI Changes:
- Rename u32 to __u32 in struct drm_format_modifier_blob (Lionel)

Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
* tag 'drm-misc-next-fixes-2017-08-28' of git://anongit.freedesktop.org/git/drm-misc:
  drm: rename u32 in __u32 in uapi

drm/syncobj: Add a signal ioctl (v3)

This IOCTL provides a mechanism for userspace to trigger a sync object
directly.  There are other ways that userspace can trigger a syncobj
such as submitting a dummy batch somewhere or hanging on to a triggered
sync_file and doing an import.  This just provides an easy way to
manually trigger the sync object without weird hacks.

The motivation for this IOCTL is Vulkan fences.  Vulkan lets you create
a fence already in the signaled state so that you can wait on it
immediatly without stalling.  We could also handle this with a new
create flag to ask the driver to create a syncobj that is already
signaled but the IOCTL seemed a bit cleaner and more generic.

v2:
 - Take an array of sync objects (Dave Airlie)
v3:
 - Throw -EINVAL if pad != 0

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/syncobj: Add a reset ioctl (v3)

This just resets the dma_fence to NULL so it looks like it's never been
signaled.  This will be useful once we add the new wait API for allowing
wait on "submit and signal" behavior.

v2:
 - Take an array of sync objects (Dave Airlie)
v3:
 - Throw -EINVAL if pad != 0

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Christian König <christian.koenig@amd.com> (v1)
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/syncobj: Add a syncobj_array_find helper

The wait ioctl has a bunch of code to read an syncobj handle array from
userspace and turn it into an array of syncobj pointers.  We're about to
add two new IOCTLs which will need to work with arrays of syncobj
handles so let's make some helpers.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/syncobj: Allow wait for submit and signal behavior (v5)

Vulkan VkFence semantics require that the application be able to perform
a CPU wait on work which may not yet have been submitted.  This is
perfectly safe because the CPU wait has a timeout which will get
triggered eventually if no work is ever submitted.  This behavior is
advantageous for multi-threaded workloads because, so long as all of the
threads agree on what fences to use up-front, you don't have the extra
cross-thread synchronization cost of thread A telling thread B that it
has submitted its dependent work and thread B is now free to wait.

Within a single process, this can be implemented in the userspace driver
by doing exactly the same kind of tracking the app would have to do
using posix condition variables or similar.  However, in order for this
to work cross-process (as is required by VK_KHR_external_fence), we need
to handle this in the kernel.

This commit adds a WAIT_FOR_SUBMIT flag to DRM_IOCTL_SYNCOBJ_WAIT which
instructs the IOCTL to wait for the syncobj to have a non-null fence and
then wait on the fence.  Combined with DRM_IOCTL_SYNCOBJ_RESET, you can
easily get the Vulkan behavior.

v2:
 - Fix a bug in the invalid syncobj error path
 - Unify the wait-all and wait-any cases
v3:
 - Unify the timeout == 0 case a bit with the timeout > 0 case
 - Use wait_event_interruptible_timeout
v4:
 - Use proxy fence
v5:
 - Revert to a combination of v2 and v3
 - Don't use proxy fences
 - Don't use wait_event_interruptible_timeout because it just adds an
   extra layer of callbacks

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/syncobj: Add a CREATE_SIGNALED flag

This requests that the driver create the sync object such that it
already has a signaled dma_fence attached.  Because we don't need
anything in particular (just something signaled), we use a dummy null
fence.  This is useful for Vulkan which has a similar flag that can be
passed to vkCreateFence.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>