openwrt/staging/blogic.git
7 years agodrm/i915/gvt: Don't dump partial state in cmd parser
Changbin Du [Thu, 2 Nov 2017 05:33:23 +0000 (13:33 +0800)]
drm/i915/gvt: Don't dump partial state in cmd parser

I have seen the cmd parser dump partial odd info. Stop that and only dump
the full verbose info when debug enabled.

Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Reduce rcs mocs switch latency
Changbin Du [Mon, 30 Oct 2017 06:19:15 +0000 (14:19 +0800)]
drm/i915/gvt: Reduce rcs mocs switch latency

Use I915_WRITE_FW instead of I915_WRITE to reduce overhead.
The overall mmio switch latency lowers from ~600us to ~180us.

Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agoMAINTAINERS: Update gvt-linux.git new repo place
Zhenyu Wang [Thu, 26 Oct 2017 07:29:13 +0000 (15:29 +0800)]
MAINTAINERS: Update gvt-linux.git new repo place

gvt-linux.git repo is moved for its new place under
https://github.com/intel/gvt-linux.git

Old https://github.com/01org/gvt-linux.git is set
only for redirect now.

Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Add new debugfs tool mmio_diff
Changbin Du [Mon, 23 Oct 2017 03:46:44 +0000 (11:46 +0800)]
drm/i915/gvt: Add new debugfs tool mmio_diff

This new debugfs entry is used to figure out which registers of vGPU
is different to host. It is a useful tool for new platform enabling
and debugging. When read this entry, all the diff mmio are recognized
and sorted by mmio offset. Besides, the bit positions of different
value are listed in 'Diff' column. Here is a show:

$ sudo cat ./mmio_diff
Offset   HW       vGPU     Diff
00002030 000025f8 00000000 3-8,10,13
00002034 012025f8 00000000 3-8,10,13,21,24
00002038 027fb000 00000000 12-13,15-22,25
0000203c 00003000 00000000 12-13
00002054 0000000a 00000040 1,3,6
00002074 012025f8 00000000 3-8,10,13,21,24
00002080 fffe6000 00000000 13-14,17-31
000020a8 fffffeff ffffffff 8
000020d4 00000004 00000000 2
....
00145974 eb42718c 010c11b0 2-5,13-14,17-19,22,25,27,29-31
00145978 0000002f 0000002a 0,2
0014597c 0000002f 0000002a 0,2
00145980 0000002b 00000028 0-1
00145984 a5a87c9e b27d20c0 1-4,6,10-12,14,16,18,20,22-26,28
001459c0 88390000 883c0000 16,18
00146200 88350000 883a0000 16-19
Total: 72432, Diff: 901

v3: fix a typo.
v2: add mmio_hw_access_pre/post().

Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Add mmio iterator intel_gvt_for_each_tracked_mmio()
Changbin Du [Mon, 23 Oct 2017 03:46:43 +0000 (11:46 +0800)]
drm/i915/gvt: Add mmio iterator intel_gvt_for_each_tracked_mmio()

This patch add a function intel_gvt_for_each_tracked_mmio() to
iterate each tracked mmio. The caller don't be aware of how the
tracked mmios are presented internally.

v2: remove snapshot_hw_mmio_registers().

Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: opregion virtualization for win guest
Xiaolin Zhang [Fri, 20 Oct 2017 13:52:29 +0000 (21:52 +0800)]
drm/i915/gvt: opregion virtualization for win guest

this is an enhanced opregion emulation for win guest support
by initializing more data members including opregion header
size, version and child device propertity for display port.
for simplicity, redefined child_device_config structure.

Signed-off-by: Xiaolin Zhang <xiaolin.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: update CSB and CSB write pointer in virtual HWSP
Weinan Li [Fri, 20 Oct 2017 07:16:46 +0000 (15:16 +0800)]
drm/i915/gvt: update CSB and CSB write pointer in virtual HWSP

The engine provides a mirror of the CSB and CSB write pointer in the HWSP.
Read these status from virtual HWSP in VM can reduce CPU utilization while
applications have much more short GPU workloads. Here we update the
corresponding data in virtual HWSP as it in virtual MMIO.

Before read these status from HWSP in GVT-g VM, please ensure the host
support it by checking the BIT(3) of caps in PVINFO.

Virtual HWSP only support GEN8+ platform, since the HWSP MMIO may change
follow the platform update, please add the corresponding MMIO emulation
when enable new platforms in GVT-g.

v3 : Add address audit in HWSP address update.

v4 :
     Separate this patch with enalbe virtual HWSP in VM.
     Use intel_gvt_render_mmio_to_ring_id() to determine ring_id by offset.

v5 : Remove unnessary check about Gen8, GVT-g only support Gen8+.

Signed-off-by: Weinan Li <weinan.z.li@intel.com>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Refine broken PPGTT scratch
Zhi Wang [Mon, 16 Oct 2017 18:00:27 +0000 (02:00 +0800)]
drm/i915/gvt: Refine broken PPGTT scratch

Refine previously broken PPGTT scratch. Scratch PTE was no correctly
handled and also the handling of scratch entries in page table walk was
not well organized, which brings gaps of introducing lazy shadow.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Introduce ops->set_present()
Zhi Wang [Tue, 10 Oct 2017 09:24:26 +0000 (17:24 +0800)]
drm/i915/gvt: Introduce ops->set_present()

We need ops->set_present() during generating a new scratch page table
entry.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Introduce page table type of current level in GTT type enumerations
Zhi Wang [Tue, 10 Oct 2017 09:19:30 +0000 (17:19 +0800)]
drm/i915/gvt: Introduce page table type of current level in GTT type enumerations

Need to figure out page table type of current level by GTT entry type
during getting a scratch page table entry.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Fix a bug of unexpectedly clear scratch page table
Zhi Wang [Tue, 10 Oct 2017 09:14:13 +0000 (17:14 +0800)]
drm/i915/gvt: Fix a bug of unexpectedly clear scratch page table

During a vGPU reset, the scratch page table shouldn't be cleared, what
needs to be cleared should be the scratch page.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Let the caller choose if a shadow page should be put into hash table
Zhi Wang [Tue, 10 Oct 2017 06:34:11 +0000 (14:34 +0800)]
drm/i915/gvt: Let the caller choose if a shadow page should be put into hash table

As we want to re-use intel_vgpu_shadow_page in buidling scrach page table
and we don't want to put scrach page table page into hash table, a new
param is introduced to give the caller a choice to decide if a shadow page
should be put into hash table.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Use I915_GTT_PAGE_SIZE
Zhi Wang [Tue, 10 Oct 2017 05:51:32 +0000 (13:51 +0800)]
drm/i915/gvt: Use I915_GTT_PAGE_SIZE

As there is already an I915_GTT_PAGE_SIZE marco in i915, let GVT-g use it
as well. Also this patch re-names some GTT marcos with additional prefix.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Export intel_gvt_render_mmio_to_ring_id()
Zhi Wang [Sat, 30 Sep 2017 09:42:20 +0000 (17:42 +0800)]
drm/i915/gvt: Export intel_gvt_render_mmio_to_ring_id()

Since many emulation logic needs to convert the offset of ring registers
into ring id, we export it for other caller which might need it.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Factor intel_vgpu_page_track
Zhi Wang [Thu, 28 Sep 2017 18:47:55 +0000 (02:47 +0800)]
drm/i915/gvt: Factor intel_vgpu_page_track

As the data structure of "intel_vgpu_guest_page" will become much heavier
in future, it's better to factor out the guest memory page track mechnisim
as early as possible.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Refine shadow batch buffer
Zhi Wang [Sun, 24 Sep 2017 13:53:03 +0000 (21:53 +0800)]
drm/i915/gvt: Refine shadow batch buffer

1) Use standard i915 GEM object sequence to access the shadow batch buffer.
2) Manage i915 vma life cycle to solve one FIXME.

v2:
- Refine code structure.
- Refine the usage of GEM APIs.
- Add the missing lock/unlock in release_shadow_batch_buffer.

Test on my SKL NuC.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Refine find_bb_size()
Zhi Wang [Fri, 22 Sep 2017 13:12:03 +0000 (21:12 +0800)]
drm/i915/gvt: Refine find_bb_size()

Returns the error code if something is wrong and the size of batch buffer
is passed through the pointer.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Use BIT() to make klockwork happy
Zhi Wang [Tue, 26 Sep 2017 07:02:21 +0000 (15:02 +0800)]
drm/i915/gvt: Use BIT() to make klockwork happy

Replace the plain bit usage with BIT() to make klockwork happy.

Cc: Deng Hongyi <hongyi.deng@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Add basic debugfs infrastructure
Changbin Du [Tue, 26 Sep 2017 08:19:13 +0000 (16:19 +0800)]
drm/i915/gvt: Add basic debugfs infrastructure

We need debugfs entry to expose some debug information of gvt and vGPUs.
The first tool will be added is mmio-diff, which help to find the
difference values of host and vGPU mmio. It's useful for platform
enabling.

This patch just add a basic debugfs infrastructure, each vGPU has its own
sub-folder. Two simple attributes are created as a template.
.
├── num_tracked_mmio
├── vgpu1
|   └── active
└── vgpu2
    └── active

Signed-off-by: Changbin Du <changbin.du@intel.com>
Reviewed-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Refactor vGPU type code in kvmgt part
fred gao [Thu, 28 Sep 2017 03:03:03 +0000 (11:03 +0800)]
drm/i915/gvt: Refactor vGPU type code in kvmgt part

all the vGPU type related code in kvmgt will be moved into
gvt.c/gvt.h files while the common vGPU type related interfaces
will be called.

v2:
- intel_gvt_{init,cleanup}_vgpu_type_groups are initialized in
  gvt part. (Wang, Zhi)

Signed-off-by: fred gao <fred.gao@intel.com>
Reviewed-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move vGPU type related code into gvt file
fred gao [Thu, 28 Sep 2017 03:03:02 +0000 (11:03 +0800)]
drm/i915/gvt: Move vGPU type related code into gvt file

In this patch, all the vGPU type related code will be merged into
same gvt file and the common interface will be exposed to both
XenGT and KvmGT.

v2:
- remove the useless mdev_* gvt_ops.
  add get_gvt_attr ops for MPT module.
  intel_gvt_{init,cleanup}_vgpu_type_groups are initialized in
  gvt part. (Wang, Zhi)
- set gvt_vgpu_type_groups[i] to NULL. (Zhang,Xiong)

Signed-off-by: fred gao <fred.gao@intel.com>
Reviewed-by: Zhi Wang <zhi.a.wang@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move clean_workloads() into scheduler.c
Zhi Wang [Tue, 12 Sep 2017 17:58:35 +0000 (01:58 +0800)]
drm/i915/gvt: Move clean_workloads() into scheduler.c

Move clean_workloads() into scheduler.c since it's not specific to
execlist.

v2:

- Remove clean_workloads in intel_vgpu_select_submission_ops. (Zhenyu)

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Introduce intel_vgpu_reset_submission
Zhi Wang [Tue, 12 Sep 2017 17:41:35 +0000 (01:41 +0800)]
drm/i915/gvt: Introduce intel_vgpu_reset_submission

Introduce an generic API to reset vGPU virtual submission interface.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Introduce vGPU submission ops
Zhi Wang [Tue, 12 Sep 2017 16:31:29 +0000 (00:31 +0800)]
drm/i915/gvt: Introduce vGPU submission ops

Introduce vGPU submission ops to support easy switching submission mode
of one vGPU between different OSes.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Remove one extra declaration in scheduler.h
Zhi Wang [Tue, 12 Sep 2017 14:39:08 +0000 (22:39 +0800)]
drm/i915/gvt: Remove one extra declaration in scheduler.h

Now the function has been moved into scheduler.c. The extra declaration
is not necessary.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move common vGPU workload creation into scheduler.c
Zhi Wang [Tue, 12 Sep 2017 14:33:12 +0000 (22:33 +0800)]
drm/i915/gvt: Move common vGPU workload creation into scheduler.c

Move common vGPU workload creation functions into scheduler.c since
they are not specific to execlist emulation.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move common workload preparation into prepare_workload()
Zhi Wang [Tue, 12 Sep 2017 14:06:39 +0000 (22:06 +0800)]
drm/i915/gvt: Move common workload preparation into prepare_workload()

Move common workload preparation into prepare_workload() in scheduler.c,
as they are not specific to execlist emulation.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Factor out prepare_workload()
Zhi Wang [Tue, 12 Sep 2017 13:51:10 +0000 (21:51 +0800)]
drm/i915/gvt: Factor out prepare_workload()

Factor out prepare_workload() for the following re-factor.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Factor out vGPU workload creation/destroy
Zhi Wang [Tue, 12 Sep 2017 13:42:09 +0000 (21:42 +0800)]
drm/i915/gvt: Factor out vGPU workload creation/destroy

Factor out vGPU workload creation/destroy functions since they are not
specific to execlist emulation.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Use dyndbg for gvt debug info
Shuo Liu [Thu, 21 Sep 2017 00:02:31 +0000 (08:02 +0800)]
drm/i915/gvt: Use dyndbg for gvt debug info

It's better enable/disable and classify gvt debug info dynamically.
This patch change it to dyndbg so can be dynamically enable/disable
each item. All gvt log can be enabled by,
 $ echo 'file *gvt* +p' > /sys/kernel/debug/dynamic_debug/control

Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: ensure -ve return value is handled correctly
Colin Ian King [Tue, 19 Sep 2017 15:55:34 +0000 (16:55 +0100)]
drm/i915/gvt: ensure -ve return value is handled correctly

An earlier fix changed the return type from find_bb_size however the
integer return is being assigned to a unsigned int so the -ve error
check will never be detected. Make bb_size an int to fix this.

Detected by CoverityScan CID#1456886 ("Unsigned compared against 0")

Fixes: 1e3197d6ad73 ("drm/i915/gvt: Refine error handling for perform_bb_shadow")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Add VM healthy check for submit_context
fred gao [Tue, 19 Sep 2017 07:11:29 +0000 (15:11 +0800)]
drm/i915/gvt: Add VM healthy check for submit_context

When a scan error occurs in submit_context, this patch is to
decrease the mm ref count and free the workload struct before
the workload is abandoned.

v2:
- submit_context related code should be combined together. (Zhenyu)

v3:
- free all the unsubmitted workloads. (Zhenyu)

v4:
- refine the clean path. (Zhenyu)

v5:
- polish the title. (Zhenyu)

Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Add VM healthy check for workload_thread
fred gao [Tue, 19 Sep 2017 07:11:28 +0000 (15:11 +0800)]
drm/i915/gvt: Add VM healthy check for workload_thread

When a scan error occurs in dispatch_workload, this patch is to
check the healthy state and free all the queued workloads before
the failsafe mode is entered.

Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Change the return type during command scan
fred gao [Tue, 19 Sep 2017 21:36:47 +0000 (05:36 +0800)]
drm/i915/gvt: Change the return type during command scan

Generally, there are 3 types of errors during command scan: a) some
commands might be unknown with EBADRQC;  b) some cmd access invalid
address with EFAULT; c) some unexpected force nonpriv cmd with EPERM.
later the healthy state can be judged through the return error.

v2:
- remove some internal i915 errors rating.  (Zhenyu)

v3:
- the healthy state is judged through the internal defined return
  error. (Zhenyu)
- force non priv cmd error can be ignored. (Kevin)

v4:
- reuse standard defined errno instead of recreate, e.g EBADRQC for
  unknown cmd, EFAULT for invalid address, EPERM for nonpriv. (Zhenyu)

v5:
- remove some irrelevant code for the patch.
- fix typo of vgpu_is_vm_unhealthy. (Zhenyu)

v6:
- move the healthy check and failsafe code into another patch. (Zhenyu)

v7:
- polish title and commit message. (Zhenyu)

Signed-off-by: fred gao <fred.gao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915/gvt: Do not allocate initial ring scan buffer
Zhi Wang [Sun, 10 Sep 2017 14:01:10 +0000 (22:01 +0800)]
drm/i915/gvt: Do not allocate initial ring scan buffer

Theoretically, the largest bulk of commands in the ring buffer of an
engine might be the first submission, which usually contains a lot
of commands to initialize the HW. After removing the initial allocation
of the ring scan buffer and let krealloc() do everything we need, we
still have a big chance to get the buffer of suitable size in the first
submission.

Tested on my SKL NUC.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move ring scan buffers into intel_vgpu_submission
Zhi Wang [Sun, 10 Sep 2017 13:58:11 +0000 (21:58 +0800)]
drm/i915/gvt: Move ring scan buffers into intel_vgpu_submission

Move ring scan buffers into intel_vgpu_submission since they belongs to
a part of vGPU submission stuffs.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Rename reserved ring buffer
Zhi Wang [Sun, 10 Sep 2017 13:46:06 +0000 (21:46 +0800)]
drm/i915/gvt: Rename reserved ring buffer

"reserved" means reserve something from somewhere. Actually they are
buffers used by command scanner. Rename it to ring_scan_buffer.

v2:

- Remove the usage of an extra variable. (Zhenyu)

Fixes: 0a53bc07f044 ("drm/i915/gvt: Separate cmd scan from request allocation")
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Fix a memory leak in cmd_parser.c
Zhi Wang [Sun, 10 Sep 2017 13:36:21 +0000 (21:36 +0800)]
drm/i915/gvt: Fix a memory leak in cmd_parser.c

The pointer points to the original memory can never take the return value
of krealloc().

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move tlb_handle_pending into intel_vgpu_submission
Zhi Wang [Sun, 10 Sep 2017 13:33:20 +0000 (21:33 +0800)]
drm/i915/gvt: Move tlb_handle_pending into intel_vgpu_submission

Move tlb_handle_pending into intel_vgpu_submssion since it belongs to a
part of vGPU submission stuffs

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Introduce intel_vgpu_submission
Zhi Wang [Sun, 10 Sep 2017 13:15:18 +0000 (21:15 +0800)]
drm/i915/gvt: Introduce intel_vgpu_submission

Introduce intel_vgpu_submission to hold all members related to submission
in struct intel_vgpu before.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Move workload cache init/clean into intel_vgpu_{setup, clean}_submission()
Zhi Wang [Sun, 10 Sep 2017 12:28:09 +0000 (20:28 +0800)]
drm/i915/gvt: Move workload cache init/clean into intel_vgpu_{setup, clean}_submission()

Move vGPU workload cache initialization/de-initialization into
intel_vgpu_{setup, clean}_submission() since they are not specific to
execlist stuffs.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Rename intel_vgpu_{init, clean}_gvt_context()
Zhi Wang [Sun, 10 Sep 2017 12:08:18 +0000 (20:08 +0800)]
drm/i915/gvt: Rename intel_vgpu_{init, clean}_gvt_context()

To move workload related functions into scheduler.c, an expected way is
to collect all the init/clean functions related to vGPU workload
submission into fewer functions.

Rename intel_vgpu_{init, clean}_gvt_context() for above usage in future.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Make elsp_dwords in the right order
Zhi Wang [Sun, 10 Sep 2017 08:40:04 +0000 (16:40 +0800)]
drm/i915/gvt: Make elsp_dwords in the right order

The context descriptors in elsp_dwords are stored in a reversed order and
the definition of context descriptor is also reversed. The revesred stuff
is hard to be used and might cause misunderstanding. Make them in the right
oder for following code re-factoring.

Tested on my SKL NUC.

Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
7 years agodrm/i915/gvt: Add support for opregion virtualization
Xiaolin Zhang [Fri, 8 Sep 2017 13:37:48 +0000 (21:37 +0800)]
drm/i915/gvt: Add support for opregion virtualization

opregion emulated with a copy from host which leads to some display
bugs such as guest resolution adjustment failure due to host opregion
fail to claim port D support. with a fake opregion table provided
to fully emulate opregion to meet guest port requirement.

v1 - initial patch
v2 - reforamt opregion arrary with 0x02x output
v3 - opregion array removed with opregion generation on host initizaiton
v4 - rebased v3 patch from stable branch to staging branch which also has
     different struct child_device_config and addressed v3 review comments.

Signed-off-by: Xiaolin Zhang <xiaolin.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
7 years agodrm/i915: Remove pre-production pooled-EU w/a for Broxton
Chris Wilson [Tue, 14 Nov 2017 13:51:16 +0000 (13:51 +0000)]
drm/i915: Remove pre-production pooled-EU w/a for Broxton

WaEnablePooledEuFor2x6 only applies to preproduction models, unsupported
since commit 0102ba1fd8af ("drm/i915: Add early BXT sdv to the list of
preproduction machines").

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114135116.30036-1-chris@chris-wilson.co.uk
Reviewed-by: David Weinehall <david.weinehall@linux.intel.com>
7 years agodrm/i915: Make request's wait-for-space explicit
Chris Wilson [Wed, 15 Nov 2017 15:12:04 +0000 (15:12 +0000)]
drm/i915: Make request's wait-for-space explicit

At the start of building a request, we would wait for roughly enough
space to fit the average request (to reduce the likelihood of having to
wait and abort partway through request construction). To achieve we
would try to begin a 0-length command packet, this just adds extra
confusion so make the wait-for-space explicit, as in the next patch we
want to move it from the backend to the i915_gem_request_alloc() so it
can ensure that the wait-for-space is the first operation in building a
new request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171115151204.8105-2-chris@chris-wilson.co.uk
7 years agodrm/i915/selftests: Increase size for mock ringbuffer
Chris Wilson [Wed, 15 Nov 2017 15:12:03 +0000 (15:12 +0000)]
drm/i915/selftests: Increase size for mock ringbuffer

We don't actually emit any commands into the ringbuffer, so we set it
very small. However, an upcoming change centralises the wait-for-space
into i915_gem_request_alloc() and that imposes a minimum size upon all
ringbuffers (mock or real) of MIN_SPACE_FOR_ADD_REQUEST. Grow the
mock ringbuffer such that we allocate a single page for the struct+buffer,
satisfying the new condition without wasting too much space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171115151204.8105-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Initialise entry in intel_ppat_get() for older compilers
Chris Wilson [Wed, 15 Nov 2017 13:17:05 +0000 (13:17 +0000)]
drm/i915: Initialise entry in intel_ppat_get() for older compilers

gcc-4.7.3 is confused by the guards inside intel_ppat_get() and reports:

drivers/gpu/drm/i915/i915_gem_gtt.c: In function ‘intel_ppat_get’:
drivers/gpu/drm/i915/i915_gem_gtt.c:3044:27: warning: ‘entry’ may be used uninitialized in this function [-Wmaybe-uninitialized]

Forgive the compiler this once, and rearrange the code so that entry is
always initialised.

v2: Flavour with a bit of NULL (instead of ERR_PTR(-ENOSPC))

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Zhi Wang <zhi.a.wang@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171115131705.16341-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915/selftests: Markup __iomem for igt_gem_coherency
Chris Wilson [Tue, 14 Nov 2017 19:18:42 +0000 (19:18 +0000)]
drm/i915/selftests: Markup __iomem for igt_gem_coherency

Silence sparse warnings by using __iomem markup and io accessors.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114191842.19063-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Use ELK stolen memory reserved detection for ILK
Ville Syrjälä [Thu, 2 Nov 2017 15:17:37 +0000 (17:17 +0200)]
drm/i915: Use ELK stolen memory reserved detection for ILK

While I have no solid proof that ILK follows the ELK path when it
comes to the stolen memory reserved area, there are some hints that
it might be the case. Unfortunately my ILK doesn't have this enabled,
and no way to enable it via the BIOS it seems.

So let's have ILK use the ELK code path, and let's toss in a WARN
into the code to see if we catch anyone with an ILK that has this
enabled to further analyze the situation.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171102151737.23336-3-ville.syrjala@linux.intel.com
Acked-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
7 years agodrm/i915: Make the report about a bogus stolen reserved area an error
Ville Syrjälä [Thu, 2 Nov 2017 15:17:36 +0000 (17:17 +0200)]
drm/i915: Make the report about a bogus stolen reserved area an error

Now that we should be properly filtering out the cases when the stolen
reserved area is disabled, let's convert the debug message about a
misplaced reserved area into an error.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171102151737.23336-2-ville.syrjala@linux.intel.com
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
7 years agodrm/i915: Check if the stolen memory "reserved" area is enabled or not
Ville Syrjälä [Thu, 2 Nov 2017 15:17:35 +0000 (17:17 +0200)]
drm/i915: Check if the stolen memory "reserved" area is enabled or not

Apparently there are some machines that put semi-sensible looking values
into the stolen "reserved" base and size, except those values are actually
outside the stolen memory. There is a bit in the register which
supposedly could tell us whether the reserved area is even enabled or
not. Let's check for that before we go trusting the base and size.

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/20171102151737.23336-1-ville.syrjala@linux.intel.com
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
7 years agodrm/i915: Fix kerneldocs for intel_audio.c
Ville Syrjälä [Tue, 14 Nov 2017 19:11:27 +0000 (21:11 +0200)]
drm/i915: Fix kerneldocs for intel_audio.c

Fix copy/paste fail in kerneldocs for intel_audio_codec_disable().

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114191127.16188-1-ville.syrjala@linux.intel.com
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
7 years agodrm/i915: Call uncore_suspend before platform suspend handlers
Hans de Goede [Tue, 14 Nov 2017 13:55:18 +0000 (14:55 +0100)]
drm/i915: Call uncore_suspend before platform suspend handlers

Quoting Ville: "the forcewake timer might still be active until the uncore
suspend, and having active forcewakes while we've already told the GT wake
stuff to stop acting normally doesn't seem quite right to me."

Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Suggested-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114135518.15981-3-hdegoede@redhat.com
7 years agodrm/i915: Re-register PMIC bus access notifier on runtime resume
Hans de Goede [Tue, 14 Nov 2017 13:55:17 +0000 (14:55 +0100)]
drm/i915: Re-register PMIC bus access notifier on runtime resume

intel_uncore_suspend() unregisters the uncore code's PMIC bus access
notifier and gets called on both normal and runtime suspend.

intel_uncore_resume_early() re-registers the notifier, but only on
normal resume. Add a new intel_uncore_runtime_resume() function which
only re-registers the notifier and call that on runtime resume.

Cc: stable@vger.kernel.org
Reported-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114135518.15981-2-hdegoede@redhat.com
7 years agodrm/i915: Fix false-positive assert_rpm_wakelock_held in i915_pmic_bus_access_notifier v2
Hans de Goede [Fri, 10 Nov 2017 15:03:01 +0000 (16:03 +0100)]
drm/i915: Fix false-positive assert_rpm_wakelock_held in i915_pmic_bus_access_notifier v2

assert_rpm_wakelock_held is triggered from i915_pmic_bus_access_notifier
even though it gets unregistered on (runtime) suspend, this is caused
by a race happening under the following circumstances:

intel_runtime_pm_put does:

   atomic_dec(&dev_priv->pm.wakeref_count);

   pm_runtime_mark_last_busy(kdev);
   pm_runtime_put_autosuspend(kdev);

And pm_runtime_put_autosuspend calls intel_runtime_suspend from
a workqueue, so there is ample of time between the atomic_dec() and
intel_runtime_suspend() unregistering the notifier. If the notifier
gets called in this windowd assert_rpm_wakelock_held falsely triggers
(at this point we're not runtime-suspended yet).

This commit adds disable_rpm_wakeref_asserts and
enable_rpm_wakeref_asserts calls around the
intel_uncore_forcewake_get(FORCEWAKE_ALL) call in
i915_pmic_bus_access_notifier fixing the false-positive WARN_ON.

Changes in v2:
-Reword comment explaining why disabling the wakeref asserts is
 ok and necessary

Cc: stable@vger.kernel.org
Reported-by: FKr <bugs-freedesktop@ubermail.me>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110150301.9601-2-hdegoede@redhat.com
7 years agodrm/i915/selftests: Always initialise err
Chris Wilson [Tue, 14 Nov 2017 22:33:46 +0000 (22:33 +0000)]
drm/i915/selftests: Always initialise err

smatch does not track initialised values as well as gcc, and this
triggers many warnings by smatch not presented by gcc. Silence smatch by
initialising the error values to -ENODEV, which we use to denote
internal errors. (If we see a selftest fail with a silent -ENODEV, we
know smatch was right!)

v2: smatch was right about igt_create_vma(), it may unlikely fail on the
first object allocation which we want to be loud about.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114223346.25958-1-chris@chris-wilson.co.uk
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
7 years agodrm/i915: Resume GuC before using GEM
Chris Wilson [Tue, 14 Nov 2017 13:03:00 +0000 (13:03 +0000)]
drm/i915: Resume GuC before using GEM

Resuming GEM presumes it can talk to hw, in particular to ensure the
kernel context is loaded upon resume for powersaving. If the GuC is
still asleep at this point, we upset the HW. Rearrange the resume such
that we restore the original order of init-hw, resume-guc, use-gem.

Fixes: 37cd33006d02 ("drm/i915: Remove redundant intel_autoenable_gt_powersave()")
References: a1c419941453 ("drm/i915/guc: Add host2guc notification for suspend and resume")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Alex Dai <yu.dai@intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114130300.25677-2-chris@chris-wilson.co.uk
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
7 years agodrm/i915: Display WA #1185 WaDisableDARBFClkGating:cnl, glk
Rodrigo Vivi [Sat, 11 Nov 2017 00:03:19 +0000 (16:03 -0800)]
drm/i915: Display WA #1185 WaDisableDARBFClkGating:cnl, glk

Display is not sending a PMRsp when a PMReq is received
at the same time that all planes are turned off.
State machine in the dcprunit is stuck in the WAIT4DONE
state which means that there is no fill_done.

WA: disable arbiter clock gating, set bit [27] of 0x46530

v2: As Ville pointed out, based on the description the issue
    can happen when disabling the planes, similar to
    WaRsPkgCStateDisplayPMReq:hsw
    Also description of the issue was updated on commit
    message to make it more clear that we need this
    earlier.
v3: Restore comment about possibility to system hang
    to where we are sure about it, without speculation. (Ville).
v4: Remove doubled sob. Actually do v3 changes :/

Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
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/20171111000319.5040-1-rodrigo.vivi@intel.com
7 years agoMerge airlied/drm-next into drm-intel-next-queued
Rodrigo Vivi [Tue, 14 Nov 2017 15:43:00 +0000 (07:43 -0800)]
Merge airlied/drm-next into drm-intel-next-queued

Catchup with upstream.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
7 years agodrm/i915: Unconditionally apply the Broxton register workaround set
Chris Wilson [Tue, 14 Nov 2017 13:43:40 +0000 (13:43 +0000)]
drm/i915: Unconditionally apply the Broxton register workaround set

Having removed the preproduction Broxton support (see commit 0102ba1fd8af
("drm/i915: Add early BXT sdv to the list of preproduction machines")),
we know we then always need the production Broxton workaround set and do
not need a predicate upon revision.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114134340.5439-2-chris@chris-wilson.co.uk
7 years agodrm/i915: Remove pre-production Broxton register workarounds
Chris Wilson [Tue, 14 Nov 2017 13:43:39 +0000 (13:43 +0000)]
drm/i915: Remove pre-production Broxton register workarounds

We've begun excluding pre-production Broxton machines since commit
0102ba1fd8af ("drm/i915: Add early BXT sdv to the list of preproduction
machines"), now remove the list of workaround register values for those
early machines.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170927093325.24206-1-chris@chris-wilson.co.uk
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114134340.5439-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Unify SLICE_UNIT_LEVEL_CLKGATE w/a for cnl
Chris Wilson [Sat, 11 Nov 2017 10:03:36 +0000 (10:03 +0000)]
drm/i915: Unify SLICE_UNIT_LEVEL_CLKGATE w/a for cnl

gem_workarounds reports that the SLICE_UNIT_LEVEL_CLKGATE write isn't
sticking. Commit 0a60797a0efb ("drm/i915: Implement
ReadHitWriteOnlyDisable.") presumes that SLICE_UNIT_LEVEL_CLKGATE is a
masked register in the context image, but commit 90007bca6162
("drm/i915/cnl: Introduce initial Cannonlake Workarounds.") lists it as
an ordering unmasked register. The masked write will be losing the
default settings if we trust the original commit. That gem_workarounds
reports the value is lost entirely is more worrying though -- but it
clearly suggests that it is not a masked register in the context image,
so unify both w/a to use the original rmw.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103705
Fixes: 0a60797a0efb ("drm/i915: Implement ReadHitWriteOnlyDisable.")
References: 90007bca6162 ("drm/i915/cnl: Introduce initial Cannonlake Workarounds.")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Rafael Antognolli <rafael.antognolli@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171111100336.11020-1-chris@chris-wilson.co.uk
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
7 years agodrm/i915: fix 64bit divide
Lionel Landwerlin [Mon, 13 Nov 2017 23:34:53 +0000 (23:34 +0000)]
drm/i915: fix 64bit divide

ERROR: "__udivdi3" [drivers/gpu/drm/i915/i915.ko] undefined!
ERROR: "__divdi3" [drivers/gpu/drm/i915/i915.ko] undefined!

Store the frequency in kHz and drop 64bit divisions.

v2: Use div64_u64 (Matthew)

v3: store frequency in kHz to avoid 64bit divs (Chris/Ville)

Fixes: dab9178333 ("drm/i915: expose command stream timestamp frequency to userspace")
Reported-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171113233455.12085-3-lionel.g.landwerlin@intel.com
Reviewed-by: Ewelina Musial <ewelina.musial@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915: Generalize transcoder looping
Mika Kahola [Thu, 9 Nov 2017 08:37:50 +0000 (10:37 +0200)]
drm/i915: Generalize transcoder looping

To make looping through transcoders in intel_ddi.c more generic, let's switch
to use 'for_each_pipe()' macro to do this.

v2: Add a notion that we are dealing with transcoders instead of pipes (Jani)

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1510216670-16848-1-git-send-email-mika.kahola@intel.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915/glk: Refactor handling of PLANE_COLOR_CTL for GLK+
James Ausmus [Mon, 13 Nov 2017 18:11:28 +0000 (10:11 -0800)]
drm/i915/glk: Refactor handling of PLANE_COLOR_CTL for GLK+

Since GLK, some plane configuration settings have moved to the
PLANE_COLOR_CTL register. Refactor handling of the register to work like
PLANE_CTL. This also allows us to fix the set/read of the plane Alpha
Mode for GLK+.

v2: Adjust ordering of platform checks to be newest->oldest, drop
redundant comment about alpha blending. (Ville)

v3: Move Alpha Mode bits out of skl_plane_ctl_format into
skl_plane_ctl_alpha, and drop glk_plane_ctl_format, drop initialization
of state->color_ctl on platforms that don't use it, and drop color_ctl
local var. (Ville)

v4: Consolidate skl_plane_ctl_format switch statement on formats that
return the same settings. (Ville)

Signed-off-by: James Ausmus <james.ausmus@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171113181128.2926-1-james.ausmus@intel.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915: Introduce GEM proxy
Tina Zhang [Tue, 14 Nov 2017 10:25:13 +0000 (10:25 +0000)]
drm/i915: Introduce GEM proxy

GEM proxy is a kind of GEM, whose backing physical memory is pinned
and produced by guest VM and is used by host as read only. With GEM
proxy, host is able to access guest physical memory through GEM object
interface. As GEM proxy is such a special kind of GEM, a new flag
I915_GEM_OBJECT_IS_PROXY is introduced to ban host from changing the
backing storage of GEM proxy.

v3:
- update "Reviewed-by". (Joonas)

v2:
- return -ENXIO when pin and map pages of GEM proxy to kernel space.
  (Chris)

Here are the histories of this patch in "Dma-buf support for Gvt-g"
patch-set:

v14:
- return -ENXIO when gem proxy object is banned by ioctl.
  (Chris) (Daniel)

v13:
- add comments to GEM proxy. (Chris)
- don't ban GEM proxy in i915_gem_sw_finish_ioctl. (Chris)
- check GEM proxy bar after finishing i915_gem_object_wait. (Chris)
- remove GEM proxy bar in i915_gem_madvise_ioctl.

v6:
- add gem proxy barrier in the following ioctls. (Chris)
  i915_gem_set_caching_ioctl
  i915_gem_set_domain_ioctl
  i915_gem_sw_finish_ioctl
  i915_gem_set_tiling_ioctl
  i915_gem_madvise_ioctl

Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-2-git-send-email-tina.zhang@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-2-chris@chris-wilson.co.uk
7 years agodrm/i915: Object w/o backing storage is banned by -ENXIO
Tina Zhang [Tue, 14 Nov 2017 10:25:12 +0000 (10:25 +0000)]
drm/i915: Object w/o backing storage is banned by -ENXIO

-ENXIO should be returned when operations are banned from changing
backing storage of objects without backing storage.

v4:
- update "Reviewed-by". (Joonas)

v3:
- separate this patch from "Introduce GEM proxy" patch-set. (Joonas)

v2:
- update the patch description and subject to just mention objects w/o
  backing storage, instead of "GEM proxy". (Joonas)

Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-1-git-send-email-tina.zhang@intel.com
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Fix function name in comment
Lucas De Marchi [Tue, 14 Nov 2017 00:46:38 +0000 (16:46 -0800)]
drm/i915: Fix function name in comment

Commit 78597996370c (drm/i915/bxt: Fix PPS lost state after suspend
breaking eDP link training) renamed the function to
intel_power_sequencer_reset() but forgot to update comment.

Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171114004638.5186-1-lucas.demarchi@intel.com
7 years agoMerge tag 'exynos-drm-next-for-v4.15' of git://git.kernel.org/pub/scm/linux/kernel...
Dave Airlie [Tue, 14 Nov 2017 04:12:43 +0000 (14:12 +1000)]
Merge tag 'exynos-drm-next-for-v4.15' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

- Improved HDMI and Mixer drivers
  . It moves mode setup and plane update code to commit
    like other CRTC drivers
  . It makes mode commit to be called in enable callback only one time
  . some cleanup and fixup to HDMI and Mixer drivers.
  . It adds 1024x768, 1280x1024 and 1366x768 modes support
- Added HDMI audio interface driver
  . As of now, HDMI audio worked on boards with external audio codec connected
    in parallel with the HDMI audio transmitter's I2S interface.
    This patch is required to support HDMI audio properly.

* tag 'exynos-drm-next-for-v4.15' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm: exynos: Add driver for HDMI audio interface
  drm/exynos/hdmi: add 85.5MHz pixel clock for v14 HDMI PHY
  drm/exynos/mixer: enable support for 1024x768 and 1280x1024 modes
  drm/exynos/hdmi: quirk for support mode timings conversion
  drm/exynos/mixer: pass actual mode on MIXER to encoder
  drm/exynos: add mode_fixup callback to exynos_drm_crtc_ops
  drm/exynos/hdmi: remove redundant mode field
  drm/exynos/mixer: remove mixer_resources sub-structure
  drm/exynos/mixer: fix mode validation code
  drm/exynos/mixer: move resolution configuration to single function
  drm/exynos/mixer: move mode commit to enable callback
  drm/exynos/mixer: abstract out output mode setup code

7 years agoMerge branch 'linus-4.14-rc4-acp-prereq' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Mon, 13 Nov 2017 19:53:39 +0000 (05:53 +1000)]
Merge branch 'linus-4.14-rc4-acp-prereq' of git://people.freedesktop.org/~agd5f/linux into drm-next

This is a shared tree between drm and audio for some amd bits.

* 'linus-4.14-rc4-acp-prereq' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu Moving amdgpu asic types to a separate file
  ASoC: AMD: Added asic_type as ACP DMA driver platform data
  drm/amd/amdgpu: Added asic_type as ACP DMA driver platform data

7 years agoMerge tag 'drm-misc-fixes-2017-11-13' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Mon, 13 Nov 2017 19:29:34 +0000 (05:29 +1000)]
Merge tag 'drm-misc-fixes-2017-11-13' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

Driver Changes:
- qxl: Use a shadow bo as primary and blit to it to fix flicker (Gerd)
- rockchip: Convert psr spinlock to mutex (Emil)

Cc: Emil Renner Berthing <kernel@esmil.dk>
Cc: Gerd Hoffmann <kraxel@redhat.com>
* tag 'drm-misc-fixes-2017-11-13' of git://anongit.freedesktop.org/drm/drm-misc:
  drm/rockchip: analogix_dp: Use mutex rather than spinlock

7 years agoMerge tag 'drm-intel-next-fixes-2017-11-10' of git://anongit.freedesktop.org/drm...
Dave Airlie [Mon, 13 Nov 2017 19:18:21 +0000 (05:18 +1000)]
Merge tag 'drm-intel-next-fixes-2017-11-10' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

drm/i915 fixes for v4.15

* tag 'drm-intel-next-fixes-2017-11-10' of git://anongit.freedesktop.org/drm/drm-intel:
  drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU
  drm/i915: Move init_clock_gating() back to where it was
  drm/i915: Prune the reservation shared fence array
  drm/i915: Idle the GPU before shinking everything
  drm/i915: Lock llist_del_first() vs llist_del_all()
  drm/i915: Calculate ironlake intermediate watermarks correctly, v2.
  drm/i915: Disable lazy PPGTT page table optimization for vGPU
  drm/i915/execlists: Remove the priority "optimisation"
  drm/i915: Filter out spurious execlists context-switch interrupts

7 years agoMerge tag 'drm-misc-fixes-2017-11-02' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Mon, 13 Nov 2017 19:17:23 +0000 (05:17 +1000)]
Merge tag 'drm-misc-fixes-2017-11-02' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

Driver Changes:
- qxl: Use a shadow bo as primary and blit to it to fix flicker (Gerd)

* tag 'drm-misc-fixes-2017-11-02' of git://anongit.freedesktop.org/drm/drm-misc:
  qxl: alloc & use shadow for dumb buffers
  drm/qxl: replace QXL_INFO with DRM_DEBUG_DRIVER

7 years agodrm/i915: There is only one fault register from GEN8 onwards
Michel Thierry [Mon, 13 Nov 2017 17:36:28 +0000 (09:36 -0800)]
drm/i915: There is only one fault register from GEN8 onwards

Until Haswell/Baytrail, the hardware used to have a per engine fault
register (e.g. 0x4094 - render fault register, 0x4194 - media fault
register and so on). But since Broadwell, all these registers were
combined into a singe one and the engine id stored in bits 14:12.

Not only we should not been reading (and writing to) registers that do
not exist, in platforms with VCS2 (SKL), the address that would belong
this engine (0x4494, VCS2_HW = 4) is already assigned to other register.

v2: use less controversial function names (Chris).
v3: make non-exported functions static, remove now obsolete check for
engine presence before posting_read (Chris).

References: IHD-OS-BDW-Vol 2c-11.15, page 75.
References: IHD-OS-SKL-Vol 2c-05.16, page 350.
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171113173628.11689-1-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Clear per-engine fault register as early as possible
Michel Thierry [Sat, 11 Nov 2017 00:44:47 +0000 (16:44 -0800)]
drm/i915: Clear per-engine fault register as early as possible

From gen6, the hardware tracks address lookup failures and we should
clear those registers upon startup to prevent false positives. However,
this was happening before we have the engines defined (intel_uncore_init())
and the for_each_engine loop was just a nop. The earliest we can call
this is inside intel_engines_init_mmio().

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171111004448.12360-1-michel.thierry@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: expose command stream timestamp frequency to userspace
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:44 +0000 (19:08 +0000)]
drm/i915: expose command stream timestamp frequency to userspace

We use to have this fixed per generation, but starting with CNL userspace
cannot tell just off the PCI ID. Let's make this information available. This
is particularly useful for performance monitoring where much of the
normalization work is done using those timestamps (this include pipeline
statistics in both GL & Vulkan as well as OA reports).

v2: Use variables for 24MHz/19.2MHz values (Ewelina)
    Renamed function & coding style (Sagar)

v3: Fix frequency read on Broadwell (Sagar)
    Fix missing divide by 4 on <= gen4 (Sagar)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Tested-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-7-lionel.g.landwerlin@intel.com
7 years agodrm/i915/perf: enable perf support on CNL
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:43 +0000 (19:08 +0000)]
drm/i915/perf: enable perf support on CNL

This adds new registers to the whitelist to configs emitted from userspace.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-6-lionel.g.landwerlin@intel.com
7 years agodrm/i915: fix register naming
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:42 +0000 (19:08 +0000)]
drm/i915: fix register naming

This name was added with the whitelisting of registers for building up OA
configs. It is contained in a range gen8 whitelist :

   addr >= RPM_CONFIG0.reg && addr <= NOA_CONFIG(8).reg

Hence why the name isn't used anywhere.

v2: Fix register name again RPC->RCP (Matthew)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-5-lionel.g.landwerlin@intel.com
7 years agodrm/i915/perf: refactor perf setup
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:41 +0000 (19:08 +0000)]
drm/i915/perf: refactor perf setup

Gen8/9 aren't very different and we can merge some of this code.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-4-lionel.g.landwerlin@intel.com
7 years agodrm/i915/perf: add support for Coffeelake GT3
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:40 +0000 (19:08 +0000)]
drm/i915/perf: add support for Coffeelake GT3

We can enable GT3 as well as GT2.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-3-lionel.g.landwerlin@intel.com
7 years agodrm/i915/perf: complete whitelisting for OA programming on HSW
Lionel Landwerlin [Fri, 10 Nov 2017 19:08:39 +0000 (19:08 +0000)]
drm/i915/perf: complete whitelisting for OA programming on HSW

We were missing some registers and also can name one for which we only had
the offset.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-2-lionel.g.landwerlin@intel.com
7 years agodrm/i915: Handle locking better in i915_sink_crc.
Maarten Lankhorst [Fri, 10 Nov 2017 11:34:58 +0000 (12:34 +0100)]
drm/i915: Handle locking better in i915_sink_crc.

Lock the bare minimum, instead of the entire world, and
use interruptible locking because we can.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-6-maarten.lankhorst@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/i915: Handle adjust better in intel_pipe_config_compare
Maarten Lankhorst [Fri, 10 Nov 2017 11:34:56 +0000 (12:34 +0100)]
drm/i915: Handle adjust better in intel_pipe_config_compare

Some parameters use CHECK_BOOL, but should really use
CHECK_BOOL_INCOMPLETE. We cannot currently check whether
the inherited infoframes and audio are set up correctly.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-4-maarten.lankhorst@linux.intel.com
[mlankhorst: Add danvet's comment about why this is needed.]
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/i915: Check boolean options in intel_pipe_config_compare with its own macro
Maarten Lankhorst [Fri, 10 Nov 2017 11:34:55 +0000 (12:34 +0100)]
drm/i915: Check boolean options in intel_pipe_config_compare with its own macro

Add PIPE_CONF_CHECK_BOOL for boolean options, which are printed with yesno.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-3-maarten.lankhorst@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/i915: Remove bogus ips_enabled check.
Maarten Lankhorst [Fri, 10 Nov 2017 11:34:54 +0000 (12:34 +0100)]
drm/i915: Remove bogus ips_enabled check.

The flag just tells us IPS can be enabled, if the primary plane
is not enabled it means IPS might not be. This never triggered
in CI because we don't have a haswell ULT there, but can be
reproduced easily with kms_atomic_transitions.plane-all-modeset-transition

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-2-maarten.lankhorst@linux.intel.com
[mlankhorst: Remove from haswell_get_pipe_config too. (danvet)]
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/rockchip: analogix_dp: Use mutex rather than spinlock
Emil Renner Berthing [Wed, 4 Oct 2017 17:53:46 +0000 (19:53 +0200)]
drm/rockchip: analogix_dp: Use mutex rather than spinlock

On the Samsung Chromebook Plus I get this error with 4.14-rc3:

BUG: scheduling while atomic: kworker/3:1/50/0x00000002
Modules linked in:
CPU: 3 PID: 50 Comm: kworker/3:1 Not tainted 4.14.0-0.rc3-kevin #2
Hardware name: Google Kevin (DT)
Workqueue: events analogix_dp_psr_work
Call trace:
[<ffffff80080873b0>] dump_backtrace+0x0/0x320
[<ffffff80080876e4>] show_stack+0x14/0x20
[<ffffff8008606d38>] dump_stack+0x9c/0xbc
[<ffffff80080c6b5c>] __schedule_bug+0x4c/0x70
[<ffffff80086188c0>] __schedule+0x3f0/0x458
[<ffffff8008618960>] schedule+0x38/0xa0
[<ffffff800861c20c>] schedule_hrtimeout_range_clock+0x84/0xe8
[<ffffff800861c2a0>] schedule_hrtimeout_range+0x10/0x18
[<ffffff800861bcec>] usleep_range+0x64/0x78
[<ffffff8008415a6c>] analogix_dp_transfer+0x16c/0x340
[<ffffff8008412550>] analogix_dpaux_transfer+0x10/0x18
[<ffffff80083ceb14>] drm_dp_dpcd_access+0x4c/0xf0
[<ffffff80083cf614>] drm_dp_dpcd_write+0x1c/0x28
[<ffffff8008413b98>] analogix_dp_disable_psr+0x60/0xa8
[<ffffff800840da3c>] analogix_dp_psr_work+0x4c/0x90
[<ffffff80080bb09c>] process_one_work+0x1d4/0x348
[<ffffff80080bb258>] worker_thread+0x48/0x478
[<ffffff80080c11fc>] kthread+0x12c/0x130
[<ffffff8008084290>] ret_from_fork+0x10/0x18

Changing rockchip_dp_device::psr_lock to a mutex rather
than spinlock seems to fix the issue.

Signed-off-by: Emil Renner Berthing <kernel@esmil.dk>
Tested-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Signed-off-by: Mark Yao <mark.yao@rock-chips.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171004175346.11956-1-kernel@esmil.dk
7 years agodrm/i915: Remove Gen9 WAs with no effect
Oscar Mateo [Wed, 8 Nov 2017 23:59:44 +0000 (15:59 -0800)]
drm/i915: Remove Gen9 WAs with no effect

GEN8_CONFIG0 (0xD00) is a protected by a lock (bit 31) which is set by
the BIOS, so there is no way we can enable the three chicken bits
mandated by the WA (the BIOS should be doing it instead).

v2: Rebased
v3: Standalone patch

References: b033bb6d5d3a ("drm/i915/gen9: Enable must set chicken bits in config0 reg")
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1510185589-9100-2-git-send-email-oscar.mateo@intel.com
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Remove redundant intel_autoenable_gt_powersave()
Chris Wilson [Sun, 12 Nov 2017 11:27:38 +0000 (11:27 +0000)]
drm/i915: Remove redundant intel_autoenable_gt_powersave()

Now that we always execute a context switch upon module load, there is
no need to queue a delayed task for doing so. The purpose of the delayed
task is to enable GT powersaving, for which we need the HW state to be
valid (i.e. having loaded a context and initialised basic state). We
used to defer this operation as historically it was slow (due to slow
register polling, fixed with commit 1758b90e38f5 ("drm/i915: Use a hybrid
scheme for fast register waits")) but now we have a requirement to save
the default HW state.

v2: Load the kernel context (to provide the power context) upon resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171112112738.1463-3-chris@chris-wilson.co.uk
7 years agodrm/i915: Update watermark state correctly in sanitize_watermarks
Maarten Lankhorst [Fri, 10 Nov 2017 11:34:53 +0000 (12:34 +0100)]
drm/i915: Update watermark state correctly in sanitize_watermarks

We no longer use intel_crtc->wm.active for watermarks any more,
which was incorrect. But this uncovered a bug in sanitize_watermarks(),
which meant that we wrote the correct watermarks, but the next
update would still use the wrong hw watermarks for calculating.
This caused all further updates to fail with -EINVAL and the
log would reveal an error like the one below:

[   10.043902] [drm:ilk_validate_wm_level.part.8 [i915]] Sprite WM0 too large 56 (max 0)
[   10.043960] [drm:ilk_validate_pipe_wm [i915]] LP0 watermark invalid
[   10.044030] [drm:intel_crtc_atomic_check [i915]] No valid intermediate pipe watermarks are possible

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: b6b178a77210 ("drm/i915: Calculate ironlake intermediate watermarks correctly, v2.")
Cc: stable@vger.kernel.org #v4.8+
Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-1-maarten.lankhorst@linux.intel.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915/selftests: Yet another forgotten mock_i915->mm initialiser
Chris Wilson [Fri, 10 Nov 2017 23:24:47 +0000 (23:24 +0000)]
drm/i915/selftests: Yet another forgotten mock_i915->mm initialiser

Move all of the i915->mm initialisation to a private function that can
be reused by the mock i915 device to save forgetting any more steps.

For example,
<7>[ 1542.046332] [IGT] drv_selftest: starting subtest mock_objects
<4>[ 1542.123924] Setting dangerous option mock_selftests - tainting kernel
<6>[ 1542.167941] i915: Performing mock selftests with st_random_seed=0x246f5ab5 st_timeout=1000
<4>[ 1542.178012] INFO: trying to register non-static key.
<4>[ 1542.178027] the code is fine but needs lockdep annotation.
<4>[ 1542.178032] turning off the locking correctness validator.
<4>[ 1542.178041] CPU: 3 PID: 6008 Comm: kworker/3:7 Tainted: G     U          4.14.0-rc8-CI-CI_DRM_3332+ #1
<4>[ 1542.178049] Hardware name:                  /NUC6CAYB, BIOS AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017
<4>[ 1542.178144] Workqueue: events __i915_gem_free_work [i915]
<4>[ 1542.178152] Call Trace:
<4>[ 1542.178163]  dump_stack+0x68/0x9f
<4>[ 1542.178170]  register_lock_class+0x3fd/0x580
<4>[ 1542.178177]  ? unwind_next_frame+0x14/0x20
<4>[ 1542.178184]  ? __save_stack_trace+0x73/0xd0
<4>[ 1542.178191]  __lock_acquire+0xa4/0x1b00
<4>[ 1542.178254]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178261]  ? __lock_acquire+0x4ab/0x1b00
<4>[ 1542.178268]  lock_acquire+0xb0/0x200
<4>[ 1542.178273]  ? lock_acquire+0xb0/0x200
<4>[ 1542.178336]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178344]  _raw_spin_lock+0x32/0x50
<4>[ 1542.178405]  ? __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178468]  __i915_gem_free_work+0x28/0xa0 [i915]
<4>[ 1542.178476]  process_one_work+0x221/0x650
<4>[ 1542.178483]  worker_thread+0x4e/0x3c0
<4>[ 1542.178489]  kthread+0x114/0x150
<4>[ 1542.178494]  ? process_one_work+0x650/0x650
<4>[ 1542.178499]  ? kthread_create_on_node+0x40/0x40
<4>[ 1542.178506]  ret_from_fork+0x27/0x40

v2: Fish out i915->mm.object_stat_lock which was being inited over in
i915_drv.c (Matthew)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110232447.21618-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
7 years agodrm/i915: Kerneldoc typo s/rps/rps_client/
Chris Wilson [Thu, 9 Nov 2017 14:06:44 +0000 (14:06 +0000)]
drm/i915: Kerneldoc typo s/rps/rps_client/

Update the kerneldoc parameter name to match the real parameter name.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109140644.10805-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
7 years agodrm/i915: Stop caching the "golden" renderstate
Chris Wilson [Fri, 10 Nov 2017 14:26:34 +0000 (14:26 +0000)]
drm/i915: Stop caching the "golden" renderstate

As we now record the default HW state and so only emit the "golden"
renderstate once to prepare the HW, there is no advantage in keeping the
renderstate batch around as it will never be used again.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-8-chris@chris-wilson.co.uk
7 years agodrm/i915: Record the default hw state after reset upon load
Chris Wilson [Fri, 10 Nov 2017 14:26:33 +0000 (14:26 +0000)]
drm/i915: Record the default hw state after reset upon load

Take a copy of the HW state after a reset upon module loading by
executing a context switch from a blank context to the kernel context,
thus saving the default hw state over the blank context image.
We can then use the default hw state to initialise any future context,
ensuring that each starts with the default view of hw state.

v2: Unmap our default state from the GTT after stealing it from the
context. This should stop us from accidentally overwriting it via the
GTT (and frees up some precious GTT space).

Testcase: igt/gem_ctx_isolation
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-7-chris@chris-wilson.co.uk
7 years agodrm/i915: Mark the context state as dirty/written
Chris Wilson [Fri, 10 Nov 2017 14:26:32 +0000 (14:26 +0000)]
drm/i915: Mark the context state as dirty/written

In the next few patches, we will want to both copy out of the context
image and write a valid image into a new context. To be completely safe,
we should then couple in our domain tracking to ensure that we don't
have any issues with stale data remaining in unwanted cachelines.

Historically, we omitted the .write=true from the call to set-gtt-domain
in i915_switch_context() in order to avoid a stall between every request
as we would want to wait for the previous context write from the gpu.
Since then, we limit the set-gtt-domain to only occur when we first bind
the vma, so once in use we will never stall, and we are sure to flush
the context following a load from swap.

Equally we never applied the lessons learnt from ringbuffer submission
to execlists; so time to apply the flush of the lrc after load as well.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-6-chris@chris-wilson.co.uk
7 years agodrm/i915: Inline intel_modeset_gem_init()
Chris Wilson [Fri, 10 Nov 2017 14:26:31 +0000 (14:26 +0000)]
drm/i915: Inline intel_modeset_gem_init()

intel_modeset_gem_init() now only sets up the legacy overlay, so let's
remove the function and call the setup directly during driver load. This
should help us find a better point in the initialisation sequence for it
later.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-5-chris@chris-wilson.co.uk
7 years agodrm/i915: Move intel_init_clock_gating() to i915_gem_init()
Chris Wilson [Fri, 10 Nov 2017 14:26:30 +0000 (14:26 +0000)]
drm/i915: Move intel_init_clock_gating() to i915_gem_init()

Despite its name intel_init_clock_gating applies both display clock gating
workarounds; GT mmio workarounds and the occasional GT power context
workaround. Worse, sometimes it includes a context register workaround
which we need to apply before we record the default HW state for all
contexts.

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>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-4-chris@chris-wilson.co.uk
7 years agodrm/i915: Move GT powersaving init to i915_gem_init()
Chris Wilson [Fri, 10 Nov 2017 14:26:29 +0000 (14:26 +0000)]
drm/i915: Move GT powersaving init to i915_gem_init()

GT powersaving is tightly coupled to the request infrastructure. To
avoid complications with the order of initialisation in the next patch
(where we want to send requests to hw during GEM init) move the
powersaving initialisation into the purview of i915_gem_init().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-3-chris@chris-wilson.co.uk
7 years agodrm/i915: Force the switch to the i915->kernel_context
Chris Wilson [Fri, 10 Nov 2017 14:26:28 +0000 (14:26 +0000)]
drm/i915: Force the switch to the i915->kernel_context

In the next few patches, we will have a hard requirement that we emit a
context-switch to the perma-pinned i915->kernel_context (so that we can
save the HW state using that context-switch). As the first context
itself may be classed as a kernel context, we want to be explicit in our
comparison. For an extra-layer of finesse, we can check the last
unretired context on the engine; as well as the last retired context
when idle.

v2: verbose verbosity
v3: Always force the switch, even when the engine is idle, and update
the assert that this happens before suspend.

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> #v1
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-2-chris@chris-wilson.co.uk
7 years agodrm/i915: Define an engine class enum for the uABI
Tvrtko Ursulin [Fri, 10 Nov 2017 14:26:27 +0000 (14:26 +0000)]
drm/i915: Define an engine class enum for the uABI

We want to be able to report back to userspace details about an engine's
class, and in return for userspace to be able to request actions
regarding certain classes of engines. To isolate the uABI from any
variations between hw generations, we define an abstract class for the
engines and internally map onto the hw.

v2: Remove MAX from the uABI; keep it internal if we need it, but don't
let userspace make the mistake of using it themselves.
v3: s/OTHER/INVALID/
  The use of OTHER is ill-defined, so remove it from the uABI as any
  future new type of engine can define a class to suit it. But keep a
  reserved value for an invalid class, so that we can always
  unambiguously express when something doesn't belong to the
  classification.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> #v2
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-1-chris@chris-wilson.co.uk