]> git.baikalelectronics.ru Git - kernel.git/log
kernel.git
7 years agodrm/i915: Split intel_disable_ddi() into DP vs. HDMI variants
Ville Syrjälä [Tue, 10 Oct 2017 12:12:05 +0000 (15:12 +0300)]
drm/i915: Split intel_disable_ddi() into DP vs. HDMI variants

Untangle intel_disable_ddi() by splitting it into DP and HDMI specific
variants.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-8-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
7 years agodrm/i915: Remove useless eDP check from intel_ddi_pre_enable_dp()
Ville Syrjälä [Tue, 10 Oct 2017 12:12:04 +0000 (15:12 +0300)]
drm/i915: Remove useless eDP check from intel_ddi_pre_enable_dp()

intel_edp_panel_on() will itself do the is_edp() check, so the caller
doesn't have to bother. Pre-DDI code doesn't bother, so let's follow the
same approach for DDI.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-7-ville.syrjala@linux.intel.com
7 years agodrm/i915: Split intel_ddi_post_disable() into DP vs. HDMI variants
Ville Syrjälä [Tue, 10 Oct 2017 12:12:03 +0000 (15:12 +0300)]
drm/i915: Split intel_ddi_post_disable() into DP vs. HDMI variants

To clean up the mess in intel_ddi_post_disable() split it into two
clean variants for HDMI and DP.

v2: Rebase due to MST DPMS changes

Reviewed-by: Jani Nikula <jani.nikula@intel.com> #v1
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-6-ville.syrjala@linux.intel.com
7 years agodrm/i915: Inline the required bits of intel_ddi_post_disable() into intel_ddi_fdi_pos...
Ville Syrjälä [Tue, 10 Oct 2017 12:12:02 +0000 (15:12 +0300)]
drm/i915: Inline the required bits of intel_ddi_post_disable() into intel_ddi_fdi_post_disable()

To untangle the mess that is intel_ddi_post_disable() move the the bits
needed by FDI into intel_ddi_fdi_post_disable(). This way we can stop
worrying about FDI in intel_ddi_post_disable().

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-5-ville.syrjala@linux.intel.com
7 years agodrm/i915: Extract intel_disable_ddi_buf()
Ville Syrjälä [Tue, 10 Oct 2017 12:12:01 +0000 (15:12 +0300)]
drm/i915: Extract intel_disable_ddi_buf()

Extract the code to disable the DDI_BUF_CTL into small helper. This
will allows us to detangle the encoder type mess in
intel_ddi_post_disable().

v2: Keep using intel_ddi_get_encoder_port() for now

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-4-ville.syrjala@linux.intel.com
7 years agodrm/i915: Extract intel_ddi_clk_disable()
Ville Syrjälä [Tue, 10 Oct 2017 12:12:00 +0000 (15:12 +0300)]
drm/i915: Extract intel_ddi_clk_disable()

Pull the code to disable the port clock into a function. We already have
the intel_ddi_clk_select() counterpart.

v2: Keep using intel_ddi_get_encoder_port() for now (Chris)

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/20171010121207.570-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
7 years agodrm/i915: Dump 'output_types' in crtc state dump
Ville Syrjälä [Tue, 10 Oct 2017 12:11:59 +0000 (15:11 +0300)]
drm/i915: Dump 'output_types' in crtc state dump

To make it easier to debug things let's dump the output types bitmask in
the crtc state dump. And to make life that much better, let's pretty
print it as a a human reaadable string as well.

v2: Have the caller pass in the buffer (Chris)
    #undef OUTPUT_TYPE (Jani)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010121207.570-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
7 years agodrm/i915: Replace *_reference/unreference() or *_ref/unref with _get/put()
Harsha Sharma [Mon, 9 Oct 2017 12:06:43 +0000 (17:36 +0530)]
drm/i915: Replace *_reference/unreference() or *_ref/unref with _get/put()

Replace instances of drm_framebuffer_reference/unreference() with
*_get/put() suffixes and drm_dev_unref with *_put() suffix
because get/put is shorter and consistent with the
kernel use of *_get/put suffixes.
Done with following coccinelle semantic patch

@@
expression ex;
@@

(
-drm_framebuffer_unreference(ex);
+drm_framebuffer_put(ex);
|
-drm_dev_unref(ex);
+drm_dev_put(ex);
|
-drm_framebuffer_reference(ex);
+drm_framebuffer_get(ex);
)

Signed-off-by: Harsha Sharma <harshasharmaiitr@gmail.com>
[danvet: Drop the drm_dev_put change for now, to make the patch apply
with out a backmerge.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009120643.11953-1-harshasharmaiitr@gmail.com
7 years agodrm/i915: Get rid of hardcoded pipes
Mika Kahola [Fri, 13 Oct 2017 10:24:46 +0000 (13:24 +0300)]
drm/i915: Get rid of hardcoded pipes

Favor for_each_pipe() macro when looping through pipes.

Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507890286-16214-1-git-send-email-mika.kahola@intel.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
7 years agodrm/i915: Add retries for LSPCON detection
Shashank Sharma [Tue, 10 Oct 2017 10:07:44 +0000 (15:37 +0530)]
drm/i915: Add retries for LSPCON detection

We read the dp dual mode Adapter identifier to detect the
LSPCON device. It's been observed from the CI testing that in
few cases, this read can get delayed or fail. For such scenarios,
LSPCON vendors suggest to retry the read operation.

This patch adds retry in the probe function, while reading
LSPCON identifier.

V3: added this patch in the series

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102294
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102295
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102359
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103186
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630064-17908-4-git-send-email-shashank.sharma@intel.com
7 years agodrm/i915: Don't give up waiting on INVALID_MODE
Shashank Sharma [Tue, 10 Oct 2017 10:07:43 +0000 (15:37 +0530)]
drm/i915: Don't give up waiting on INVALID_MODE

Our current logic to read LSPCON's current mode, stops retries and
breaks wait-loop, if it gets LSPCON_MODE_INVALID as return from the
core function. This doesn't allow us to try reading the mode again.

This patch removes this condition and allows retries reading
the currnt mode until timeout.

This also fixes/prevents some of the noise in form of debug messages
while running IGT CI test cases.

V2: rebase, added r-b
V2: changed some debug message levels from debug->error and
    error->debug in lspcon_get_current_mode function.
V3: Rebase

Cc: Imre Deak <imre.deak@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102294
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102295
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102359
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103186
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Mahesh Kumar <Mahesh1.kumar@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630064-17908-3-git-send-email-shashank.sharma@intel.com
7 years agodrm: Add retries for lspcon mode detection
Shashank Sharma [Thu, 12 Oct 2017 16:40:08 +0000 (22:10 +0530)]
drm: Add retries for lspcon mode detection

From the CI builds, its been observed that during a driver
reload/insert, dp dual mode read function sometimes fails to
read from LSPCON device over i2c-over-aux channel.

This patch:
- adds some delay and few retries, allowing a scope for these
  devices to settle down and respond.
- changes one error log's level from ERROR->DEBUG as we want
  to call it an error only after all the retries are exhausted.

V2: Addressed review comments from Jani (for loop for retry)
V3: Addressed review comments from Imre (break on partial read too)
V3: Addressed review comments from Ville/Imre (Add the retries
    exclusively for LSPCON, not for all dp_dual_mode devices)
V4: Added r-b from Imre, sending it to dri-devel (Jani)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102294
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102295
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102359
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103186
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507826408-19322-1-git-send-email-shashank.sharma@intel.com
7 years agodrm/i915/bdw: Fix DP_AUX_CH_CTL_TIME_OUT setting
James Ausmus [Thu, 12 Oct 2017 21:30:37 +0000 (14:30 -0700)]
drm/i915/bdw: Fix DP_AUX_CH_CTL_TIME_OUT setting

Per BSpec, 400us is "BDW+ Do not use this setting." - not just PORT_A.
Set BDW to 600us unconditionally.

v2:
-Split in to two patches (Rodrigo)

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: James Ausmus <james.ausmus@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012213037.4245-2-james.ausmus@intel.com
7 years agodrm/i915: Fix DP_AUX_CH_CTL_TIME_OUT naming
James Ausmus [Thu, 12 Oct 2017 21:30:36 +0000 (14:30 -0700)]
drm/i915: Fix DP_AUX_CH_CTL_TIME_OUT naming

Rename DP_AUX_CH_CTL_TIME_OUT_1600us to DP_AUX_CH_CTL_TIME_OUT_MAX, as
the meaning of the (3 << 26) value varies per platform, but it's always the
maximum timeout for that platform. Pre-CNL it means 1600us, and for CNL
it means 3200us.

v2:
-Split in to two patches (Rodrigo)

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: James Ausmus <james.ausmus@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012213037.4245-1-james.ausmus@intel.com
7 years agodrm/i915/selftests: Exercise adding requests to a full GGTT
Chris Wilson [Thu, 12 Oct 2017 12:57:26 +0000 (13:57 +0100)]
drm/i915/selftests: Exercise adding requests to a full GGTT

A bug recently encountered involved the issue where are we were
submitting requests to different ppGTT, each would pin a segment of the
GGTT for its logical context and ring. However, this is invisible to
eviction as we do not tie the context/ring VMA to a request and so do
not automatically wait upon it them (instead they are marked as pinned,
preventing eviction entirely). Instead the eviction code must flush those
contexts by switching to the kernel context. This selftest tries to
fill the GGTT with contexts to exercise a path where the
switch-to-kernel-context failed to make forward progress and we fail
with ENOSPC.

v2: Make the hole in the filled GGTT explicit.
v3: Swap out the arbitrary timeout for a private notification from
i915_gem_evict_something()

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012125726.14736-3-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
7 years agodrm/i915/selftests: Wrap a timer into a i915_sw_fence
Chris Wilson [Thu, 12 Oct 2017 12:57:25 +0000 (13:57 +0100)]
drm/i915/selftests: Wrap a timer into a i915_sw_fence

For some selftests, we want to issue requests but delay them going to
hardware. Furthermore, we don't want those requests to block
indefinitely (or else we may hang the driver and block testing) so we
want to employ a timeout. So naturally we want a fence that is
automatically signaled by a timer.

v2: Add kselftests.
v3: Limit the API available to selftests; there isn't an overwhelming
reason to export it universally.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012125726.14736-2-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
7 years agodrm/i915: Fix eviction when the GGTT is idle but full
Chris Wilson [Thu, 12 Oct 2017 12:57:24 +0000 (13:57 +0100)]
drm/i915: Fix eviction when the GGTT is idle but full

In the full-ppgtt world, we can fill the GGTT full of context objects.
These context objects are currently implicitly tracked by the requests
that pin them i.e. they are only unpinned when the request is completed
and retired, but we do not have the link from the vma to the request
(anymore). In order to unpin those contexts, we have to issue another
request and wait upon the switch to the kernel context.

The bug during eviction was that we assumed that a full GGTT meant we
would have requests on the GGTT timeline, and so we missed situations
where those requests where merely in flight (and when even they have not
yet been submitted to hw yet). The fix employed here is to change the
already-is-idle test to no look at the execution timeline, but count the
outstanding requests and then check that we have switched to the kernel
context. Erring on the side of overkill here just means that we stall a
little longer than may be strictly required, but we only expect to hit
this path in extreme corner cases where returning an erroneous error is
worse than the delay.

v2: Logical inversion when swapping over branches.

Fixes: 80b204bce8f2 ("drm/i915: Enable multiple timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012125726.14736-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Start tracking PSR state in crtc state
Ville Syrjälä [Thu, 12 Oct 2017 13:02:01 +0000 (16:02 +0300)]
drm/i915: Start tracking PSR state in crtc state

Add the minimal amount of PSR tracking into the crtc state. This allows
precomputing the possibility of using PSR correctly, and it means we can
safely call the psr enable/disable functions for any DP endcoder.

As a nice bonus we get rid of some more crtc->config usage, which we
want to kill off eventually.

v2: Fix 'goto unlock' fail in intel_psr_enable() (Jani)
    Check intel_dp_is_edp() in is_edp_psr() (Jani)

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171012130201.21318-1-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
7 years agodrm/i915: Update DRIVER_DATE to 20171012
Jani Nikula [Thu, 12 Oct 2017 18:05:11 +0000 (21:05 +0300)]
drm/i915: Update DRIVER_DATE to 20171012

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
7 years agodrm/i915: Simplify intel_sanitize_enable_ppgtt
Joonas Lahtinen [Tue, 10 Oct 2017 14:33:55 +0000 (17:33 +0300)]
drm/i915: Simplify intel_sanitize_enable_ppgtt

Remove dead code around has_aliasing_ppgtt condition.

Suggested-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010143355.16577-1-joonas.lahtinen@linux.intel.com
7 years agodrm/i915/userptr: Drop struct_mutex before cleanup
Chris Wilson [Wed, 11 Oct 2017 14:18:57 +0000 (15:18 +0100)]
drm/i915/userptr: Drop struct_mutex before cleanup

Purely to silence lockdep, as we know that no bo can exist at this time
and so the inversion is impossible. Nevertheless, lockdep currently
warns on unload:

[  137.522565] WARNING: possible circular locking dependency detected
[  137.522568] 4.14.0-rc4-CI-CI_DRM_3209+ #1 Tainted: G     U
[  137.522570] ------------------------------------------------------
[  137.522572] drv_module_relo/1532 is trying to acquire lock:
[  137.522574]  ("i915-userptr-acquire"){+.+.}, at: [<ffffffff8109a831>] flush_workqueue+0x91/0x540
[  137.522581]
               but task is already holding lock:
[  137.522583]  (&dev->struct_mutex){+.+.}, at: [<ffffffffa014fb3f>] i915_gem_fini+0x3f/0xc0 [i915]
[  137.522605]
               which lock already depends on the new lock.

[  137.522608]
               the existing dependency chain (in reverse order) is:
[  137.522611]
               -> #3 (&dev->struct_mutex){+.+.}:
[  137.522615]        __lock_acquire+0x1420/0x15e0
[  137.522618]        lock_acquire+0xb0/0x200
[  137.522621]        __mutex_lock+0x86/0x9b0
[  137.522623]        mutex_lock_interruptible_nested+0x1b/0x20
[  137.522640]        i915_mutex_lock_interruptible+0x51/0x130 [i915]
[  137.522657]        i915_gem_fault+0x20b/0x720 [i915]
[  137.522660]        __do_fault+0x1e/0x80
[  137.522662]        __handle_mm_fault+0xa08/0xed0
[  137.522664]        handle_mm_fault+0x156/0x300
[  137.522666]        __do_page_fault+0x2c5/0x570
[  137.522668]        do_page_fault+0x28/0x250
[  137.522671]        page_fault+0x22/0x30
[  137.522672]
               -> #2 (&mm->mmap_sem){++++}:
[  137.522677]        __lock_acquire+0x1420/0x15e0
[  137.522679]        lock_acquire+0xb0/0x200
[  137.522682]        down_read+0x3e/0x70
[  137.522699]        __i915_gem_userptr_get_pages_worker+0x141/0x240 [i915]
[  137.522701]        process_one_work+0x233/0x660
[  137.522704]        worker_thread+0x4e/0x3b0
[  137.522706]        kthread+0x152/0x190
[  137.522708]        ret_from_fork+0x27/0x40
[  137.522710]
               -> #1 ((&work->work)){+.+.}:
[  137.522714]        __lock_acquire+0x1420/0x15e0
[  137.522717]        lock_acquire+0xb0/0x200
[  137.522719]        process_one_work+0x206/0x660
[  137.522721]        worker_thread+0x4e/0x3b0
[  137.522723]        kthread+0x152/0x190
[  137.522725]        ret_from_fork+0x27/0x40
[  137.522727]
               -> #0 ("i915-userptr-acquire"){+.+.}:
[  137.522731]        check_prev_add+0x430/0x840
[  137.522733]        __lock_acquire+0x1420/0x15e0
[  137.522735]        lock_acquire+0xb0/0x200
[  137.522738]        flush_workqueue+0xb4/0x540
[  137.522740]        drain_workqueue+0xd4/0x1b0
[  137.522742]        destroy_workqueue+0x1c/0x200
[  137.522758]        i915_gem_cleanup_userptr+0x15/0x20 [i915]
[  137.522770]        i915_gem_fini+0x5f/0xc0 [i915]
[  137.522782]        i915_driver_unload+0x122/0x180 [i915]
[  137.522794]        i915_pci_remove+0x19/0x30 [i915]
[  137.522797]        pci_device_remove+0x39/0xb0
[  137.522800]        device_release_driver_internal+0x15d/0x220
[  137.522803]        driver_detach+0x40/0x80
[  137.522805]        bus_remove_driver+0x58/0xd0
[  137.522807]        driver_unregister+0x2c/0x40
[  137.522809]        pci_unregister_driver+0x36/0xb0
[  137.522828]        i915_exit+0x1a/0x8b [i915]
[  137.522831]        SyS_delete_module+0x18c/0x1e0
[  137.522834]        entry_SYSCALL_64_fastpath+0x1c/0xb1
[  137.522835]
               other info that might help us debug this:

[  137.522838] Chain exists of:
                 "i915-userptr-acquire" --> &mm->mmap_sem --> &dev->struct_mutex

[  137.522844]  Possible unsafe locking scenario:

[  137.522846]        CPU0                    CPU1
[  137.522848]        ----                    ----
[  137.522850]   lock(&dev->struct_mutex);
[  137.522852]                                lock(&mm->mmap_sem);
[  137.522854]                                lock(&dev->struct_mutex);
[  137.522857]   lock("i915-userptr-acquire");
[  137.522859]
                *** DEADLOCK ***

[  137.522862] 3 locks held by drv_module_relo/1532:
[  137.522864]  #0:  (&dev->mutex){....}, at: [<ffffffff8161d47b>] device_release_driver_internal+0x2b/0x220
[  137.522869]  #1:  (&dev->mutex){....}, at: [<ffffffff8161d489>] device_release_driver_internal+0x39/0x220
[  137.522873]  #2:  (&dev->struct_mutex){+.+.}, at: [<ffffffffa014fb3f>] i915_gem_fini+0x3f/0xc0 [i915]
[  137.522888]
               stack backtrace:
[  137.522891] CPU: 0 PID: 1532 Comm: drv_module_relo Tainted: G     U          4.14.0-rc4-CI-CI_DRM_3209+ #1
[  137.522894] Hardware name:                  /NUC7i5BNB, BIOS BNKBL357.86A.0048.2017.0704.1415 07/04/2017
[  137.522897] Call Trace:
[  137.522900]  dump_stack+0x68/0x9f
[  137.522902]  print_circular_bug+0x235/0x3c0
[  137.522905]  ? lockdep_init_map_crosslock+0x20/0x20
[  137.522908]  check_prev_add+0x430/0x840
[  137.522919]  ? i915_gem_fini+0x5f/0xc0 [i915]
[  137.522922]  ? __kernel_text_address+0x12/0x40
[  137.522925]  ? __save_stack_trace+0x66/0xd0
[  137.522928]  __lock_acquire+0x1420/0x15e0
[  137.522930]  ? __lock_acquire+0x1420/0x15e0
[  137.522933]  ? lockdep_init_map_crosslock+0x20/0x20
[  137.522936]  ? __this_cpu_preempt_check+0x13/0x20
[  137.522938]  lock_acquire+0xb0/0x200
[  137.522940]  ? flush_workqueue+0x91/0x540
[  137.522943]  flush_workqueue+0xb4/0x540
[  137.522945]  ? flush_workqueue+0x91/0x540
[  137.522948]  ? __mutex_unlock_slowpath+0x43/0x2c0
[  137.522951]  ? trace_hardirqs_on_caller+0xe3/0x1b0
[  137.522954]  drain_workqueue+0xd4/0x1b0
[  137.522956]  ? drain_workqueue+0xd4/0x1b0
[  137.522958]  destroy_workqueue+0x1c/0x200
[  137.522975]  i915_gem_cleanup_userptr+0x15/0x20 [i915]
[  137.522987]  i915_gem_fini+0x5f/0xc0 [i915]
[  137.523000]  i915_driver_unload+0x122/0x180 [i915]
[  137.523015]  i915_pci_remove+0x19/0x30 [i915]
[  137.523018]  pci_device_remove+0x39/0xb0
[  137.523021]  device_release_driver_internal+0x15d/0x220
[  137.523023]  driver_detach+0x40/0x80
[  137.523026]  bus_remove_driver+0x58/0xd0
[  137.523028]  driver_unregister+0x2c/0x40
[  137.523030]  pci_unregister_driver+0x36/0xb0
[  137.523049]  i915_exit+0x1a/0x8b [i915]
[  137.523052]  SyS_delete_module+0x18c/0x1e0
[  137.523055]  entry_SYSCALL_64_fastpath+0x1c/0xb1
[  137.523057] RIP: 0033:0x7f7bd0609287
[  137.523059] RSP: 002b:00007ffef694bc18 EFLAGS: 00000246 ORIG_RAX: 00000000000000b0
[  137.523062] RAX: ffffffffffffffda RBX: ffffffff81493f33 RCX: 00007f7bd0609287
[  137.523065] RDX: 0000000000000001 RSI: 0000000000000800 RDI: 0000564f999f9fc8
[  137.523067] RBP: ffffc90005c4ff88 R08: 0000000000000000 R09: 0000000000000080
[  137.523069] R10: 00007f7bd20ef8c0 R11: 0000000000000246 R12: 0000000000000000
[  137.523072] R13: 00007ffef694be00 R14: 0000000000000000 R15: 0000000000000000
[  137.523075]  ? __this_cpu_preempt_check+0x13/0x20

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20171011141857.14161-1-chris@chris-wilson.co.uk
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/i915/dp: limit sink rates based on rate
Jani Nikula [Mon, 9 Oct 2017 09:29:59 +0000 (12:29 +0300)]
drm/i915/dp: limit sink rates based on rate

Get rid of redundant intel_dp_num_rates(). We can simply look at the
rate and limit based on that.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: <manasi.d.navare@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009092959.29021-3-jani.nikula@intel.com
7 years agodrm/i915/dp: centralize max source rate conditions more
Jani Nikula [Mon, 9 Oct 2017 09:29:58 +0000 (12:29 +0300)]
drm/i915/dp: centralize max source rate conditions more

Turn intel_dp_source_supports_hbr2() into a simple helper to query the
pre-filled source rates array, and move the conditions about which
platforms support which rates to the single point of truth in
intel_dp_set_source_rates().

This also reduces the code paths you have to think about in the source
rates initialization in intel_dp_set_source_rates(), making it easier to
grasp.

Cc: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009092959.29021-2-jani.nikula@intel.com
7 years agodrm/i915: Allow PCH platforms fall back to BIOS LVDS mode
Ville Syrjälä [Mon, 9 Oct 2017 16:19:51 +0000 (19:19 +0300)]
drm/i915: Allow PCH platforms fall back to BIOS LVDS mode

With intel_encoder_current_mode() using the normal state readout code it
actually works on PCH platforms as well. So let's nuke the PCH check from
intel_lvds_init(). I suppose there aren't any machines that actually
need this, but at least we get to eliminate a few lines of code, and one
FIXME.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009161951.22420-2-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Reuse normal state readout for LVDS/DVO fixed mode
Ville Syrjälä [Mon, 9 Oct 2017 16:19:50 +0000 (19:19 +0300)]
drm/i915: Reuse normal state readout for LVDS/DVO fixed mode

Reuse the normal state readout code to get the fixed mode for LVDS/DVO
encoders. This removes some partially duplicated state readout code
from LVDS/DVO encoders. The duplicated code wasn't actually even
populating the negative h/vsync flags, leading to possible state checker
complaints. The normal readout code populates that stuff fully.

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/20171009161951.22420-1-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Use rcu instead of stop_machine in set_wedged
Daniel Vetter [Wed, 11 Oct 2017 09:10:19 +0000 (11:10 +0200)]
drm/i915: Use rcu instead of stop_machine in set_wedged

stop_machine is not really a locking primitive we should use, except
when the hw folks tell us the hw is broken and that's the only way to
work around it.

This patch tries to address the locking abuse of stop_machine() from

commit 20e4933c478a1ca694b38fa4ac44d99e659941f5
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Nov 22 14:41:21 2016 +0000

    drm/i915: Stop the machine as we install the wedged submit_request handler

Chris said parts of the reasons for going with stop_machine() was that
it's no overhead for the fast-path. But these callbacks use irqsave
spinlocks and do a bunch of MMIO, and rcu_read_lock is _real_ fast.

To stay as close as possible to the stop_machine semantics we first
update all the submit function pointers to the nop handler, then call
synchronize_rcu() to make sure no new requests can be submitted. This
should give us exactly the huge barrier we want.

I pondered whether we should annotate engine->submit_request as __rcu
and use rcu_assign_pointer and rcu_dereference on it. But the reason
behind those is to make sure the compiler/cpu barriers are there for
when you have an actual data structure you point at, to make sure all
the writes are seen correctly on the read side. But we just have a
function pointer, and .text isn't changed, so no need for these
barriers and hence no need for annotations.

Unfortunately there's a complication with the call to
intel_engine_init_global_seqno:

- Without stop_machine we must hold the corresponding spinlock.

- Without stop_machine we must ensure that all requests are marked as
  having failed with dma_fence_set_error() before we call it. That
  means we need to split the nop request submission into two phases,
  both synchronized with rcu:

  1. Only stop submitting the requests to hw and mark them as failed.

  2. After all pending requests in the scheduler/ring are suitably
  marked up as failed and we can force complete them all, also force
  complete by calling intel_engine_init_global_seqno().

This should fix the followwing lockdep splat:

======================================================
WARNING: possible circular locking dependency detected
4.14.0-rc3-CI-CI_DRM_3179+ #1 Tainted: G     U
------------------------------------------------------
kworker/3:4/562 is trying to acquire lock:
 (cpu_hotplug_lock.rw_sem){++++}, at: [<ffffffff8113d4bc>] stop_machine+0x1c/0x40

but task is already holding lock:
 (&dev->struct_mutex){+.+.}, at: [<ffffffffa0136588>] i915_reset_device+0x1e8/0x260 [i915]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #6 (&dev->struct_mutex){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __mutex_lock+0x86/0x9b0
       mutex_lock_interruptible_nested+0x1b/0x20
       i915_mutex_lock_interruptible+0x51/0x130 [i915]
       i915_gem_fault+0x209/0x650 [i915]
       __do_fault+0x1e/0x80
       __handle_mm_fault+0xa08/0xed0
       handle_mm_fault+0x156/0x300
       __do_page_fault+0x2c5/0x570
       do_page_fault+0x28/0x250
       page_fault+0x22/0x30

-> #5 (&mm->mmap_sem){++++}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __might_fault+0x68/0x90
       _copy_to_user+0x23/0x70
       filldir+0xa5/0x120
       dcache_readdir+0xf9/0x170
       iterate_dir+0x69/0x1a0
       SyS_getdents+0xa5/0x140
       entry_SYSCALL_64_fastpath+0x1c/0xb1

-> #4 (&sb->s_type->i_mutex_key#5){++++}:
       down_write+0x3b/0x70
       handle_create+0xcb/0x1e0
       devtmpfsd+0x139/0x180
       kthread+0x152/0x190
       ret_from_fork+0x27/0x40

-> #3 ((complete)&req.done){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       wait_for_common+0x58/0x210
       wait_for_completion+0x1d/0x20
       devtmpfs_create_node+0x13d/0x160
       device_add+0x5eb/0x620
       device_create_groups_vargs+0xe0/0xf0
       device_create+0x3a/0x40
       msr_device_create+0x2b/0x40
       cpuhp_invoke_callback+0xc9/0xbf0
       cpuhp_thread_fun+0x17b/0x240
       smpboot_thread_fn+0x18a/0x280
       kthread+0x152/0x190
       ret_from_fork+0x27/0x40

-> #2 (cpuhp_state-up){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       cpuhp_issue_call+0x133/0x1c0
       __cpuhp_setup_state_cpuslocked+0x139/0x2a0
       __cpuhp_setup_state+0x46/0x60
       page_writeback_init+0x43/0x67
       pagecache_init+0x3d/0x42
       start_kernel+0x3a8/0x3fc
       x86_64_start_reservations+0x2a/0x2c
       x86_64_start_kernel+0x6d/0x70
       verify_cpu+0x0/0xfb

-> #1 (cpuhp_state_mutex){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __mutex_lock+0x86/0x9b0
       mutex_lock_nested+0x1b/0x20
       __cpuhp_setup_state_cpuslocked+0x53/0x2a0
       __cpuhp_setup_state+0x46/0x60
       page_alloc_init+0x28/0x30
       start_kernel+0x145/0x3fc
       x86_64_start_reservations+0x2a/0x2c
       x86_64_start_kernel+0x6d/0x70
       verify_cpu+0x0/0xfb

-> #0 (cpu_hotplug_lock.rw_sem){++++}:
       check_prev_add+0x430/0x840
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       cpus_read_lock+0x3d/0xb0
       stop_machine+0x1c/0x40
       i915_gem_set_wedged+0x1a/0x20 [i915]
       i915_reset+0xb9/0x230 [i915]
       i915_reset_device+0x1f6/0x260 [i915]
       i915_handle_error+0x2d8/0x430 [i915]
       hangcheck_declare_hang+0xd3/0xf0 [i915]
       i915_hangcheck_elapsed+0x262/0x2d0 [i915]
       process_one_work+0x233/0x660
       worker_thread+0x4e/0x3b0
       kthread+0x152/0x190
       ret_from_fork+0x27/0x40

other info that might help us debug this:

Chain exists of:
  cpu_hotplug_lock.rw_sem --> &mm->mmap_sem --> &dev->struct_mutex

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&dev->struct_mutex);
                               lock(&mm->mmap_sem);
                               lock(&dev->struct_mutex);
  lock(cpu_hotplug_lock.rw_sem);

 *** DEADLOCK ***

3 locks held by kworker/3:4/562:
 #0:  ("events_long"){+.+.}, at: [<ffffffff8109c64a>] process_one_work+0x1aa/0x660
 #1:  ((&(&i915->gpu_error.hangcheck_work)->work)){+.+.}, at: [<ffffffff8109c64a>] process_one_work+0x1aa/0x660
 #2:  (&dev->struct_mutex){+.+.}, at: [<ffffffffa0136588>] i915_reset_device+0x1e8/0x260 [i915]

stack backtrace:
CPU: 3 PID: 562 Comm: kworker/3:4 Tainted: G     U          4.14.0-rc3-CI-CI_DRM_3179+ #1
Hardware name:                  /NUC7i5BNB, BIOS BNKBL357.86A.0048.2017.0704.1415 07/04/2017
Workqueue: events_long i915_hangcheck_elapsed [i915]
Call Trace:
 dump_stack+0x68/0x9f
 print_circular_bug+0x235/0x3c0
 ? lockdep_init_map_crosslock+0x20/0x20
 check_prev_add+0x430/0x840
 ? irq_work_queue+0x86/0xe0
 ? wake_up_klogd+0x53/0x70
 __lock_acquire+0x1420/0x15e0
 ? __lock_acquire+0x1420/0x15e0
 ? lockdep_init_map_crosslock+0x20/0x20
 lock_acquire+0xb0/0x200
 ? stop_machine+0x1c/0x40
 ? i915_gem_object_truncate+0x50/0x50 [i915]
 cpus_read_lock+0x3d/0xb0
 ? stop_machine+0x1c/0x40
 stop_machine+0x1c/0x40
 i915_gem_set_wedged+0x1a/0x20 [i915]
 i915_reset+0xb9/0x230 [i915]
 i915_reset_device+0x1f6/0x260 [i915]
 ? gen8_gt_irq_ack+0x170/0x170 [i915]
 ? work_on_cpu_safe+0x60/0x60
 i915_handle_error+0x2d8/0x430 [i915]
 ? vsnprintf+0xd1/0x4b0
 ? scnprintf+0x3a/0x70
 hangcheck_declare_hang+0xd3/0xf0 [i915]
 ? intel_runtime_pm_put+0x56/0xa0 [i915]
 i915_hangcheck_elapsed+0x262/0x2d0 [i915]
 process_one_work+0x233/0x660
 worker_thread+0x4e/0x3b0
 kthread+0x152/0x190
 ? process_one_work+0x660/0x660
 ? kthread_create_on_node+0x40/0x40
 ret_from_fork+0x27/0x40
Setting dangerous option reset - tainting kernel
i915 0000:00:02.0: Resetting chip after gpu hang
Setting dangerous option reset - tainting kernel
i915 0000:00:02.0: Resetting chip after gpu hang

v2: Have 1 global synchronize_rcu() barrier across all engines, and
improve commit message.

v3: We need to protect the seqno update with the timeline spinlock (in
set_wedged) to avoid racing with other updates of the seqno, like we
already do in nop_submit_request (Chris).

v4: Use two-phase sequence to plug the race Chris spotted where we can
complete requests before they're marked up with -EIO.

v5: Review from Chris:
- simplify nop_submit_request.
- Add comment to rcu_read_lock section.
- Align comments with the new style.

v6: Remove unused variable to appease CI.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102886
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103096
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marta Lofstedt <marta.lofstedt@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171011091019.1425-1-daniel.vetter@ffwll.ch
7 years agodrm/i915: Introduce separate status variable for RC6 and LLC ring frequency setup
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:10 +0000 (22:30 +0100)]
drm/i915: Introduce separate status variable for RC6 and LLC ring frequency setup

Defined new struct intel_rc6 to hold RC6 specific state and
intel_ring_pstate to hold ring specific state.

v2: s/intel_ring_pstate/intel_llc_pstate. Removed checks from
autoenable_* functions. (Chris)

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-13-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-12-chris@chris-wilson.co.uk
7 years agodrm/i915: Create generic functions to control RC6, RPS
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:09 +0000 (22:30 +0100)]
drm/i915: Create generic functions to control RC6, RPS

Prepared generic functions intel_enable_rc6, intel_disable_rc6,
intel_enable_rps and intel_disable_rps functions to setup RC6/RPS
based on platforms.

v2: Make intel_enable/disable_rc6/rps static. (Chris)

v3: Added lockdep_assert_held(dev_priv->pcu_lock) in new generic
functions. (Chris)
Removed WARN_ON(&dev_priv->pcu_lock) from lower level functions as generic
function now has lockdep_assert. Rebase.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-12-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-11-chris@chris-wilson.co.uk
7 years agodrm/i915: Create generic function to setup LLC ring frequency table
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:08 +0000 (22:30 +0100)]
drm/i915: Create generic function to setup LLC ring frequency table

Prepared intel_update_ring_freq function to setup ring frequency
for applicable platforms determined by macro HAS_LLC.

v2: Replaced NEEDS_RING_FREQ_UPDATE with HAS_LLC macro. (Chris)
    Added check while calling from intel_enable_gt_powersave.

v3: s/intel_update_ring_freq/intel_enable_llc_pstate and created
new placeholder function intel_disable_llc_pstate. (Chris)

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-11-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-10-chris@chris-wilson.co.uk
7 years agodrm/i915: Rename intel_enable_rc6 to intel_rc6_enabled
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:07 +0000 (22:30 +0100)]
drm/i915: Rename intel_enable_rc6 to intel_rc6_enabled

This function gives the status of RC6, whether disabled or if
enabled then which state. intel_enable_rc6 will be used for
enabling RC6 in the next patch.

v2: Rebase.

v3: Rebase.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com> #1
Reviewed-by: Ewelina Musial <ewelina.musial@intel.com> #1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-10-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-9-chris@chris-wilson.co.uk
7 years agodrm/i915: Name structure in dev_priv that contains RPS/RC6 state as "gt_pm"
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:06 +0000 (22:30 +0100)]
drm/i915: Name structure in dev_priv that contains RPS/RC6 state as "gt_pm"

Prepared substructure rps for RPS related state. autoenable_work is
used for RC6 too hence it is defined outside rps structure. As we do
this lot many functions are refactored to use intel_rps *rps to access
rps related members. Hence renamed intel_rps_client pointer variables
to rps_client in various functions.

v2: Rebase.

v3: s/pm/gt_pm (Chris)
Refactored access to rps structure by declaring struct intel_rps * in
many functions.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com> #1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-9-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-8-chris@chris-wilson.co.uk
7 years agodrm/i915: Move rps.hw_lock to dev_priv and s/hw_lock/pcu_lock
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:05 +0000 (22:30 +0100)]
drm/i915: Move rps.hw_lock to dev_priv and s/hw_lock/pcu_lock

In order to separate GT PM related functionality into new structure
we are updating rps structure. hw_lock in it is used for display
related PCU communication too hence move it to dev_priv.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-8-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-7-chris@chris-wilson.co.uk
7 years agodrm/i915: Name i915_runtime_pm structure in dev_priv as "runtime_pm"
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:04 +0000 (22:30 +0100)]
drm/i915: Name i915_runtime_pm structure in dev_priv as "runtime_pm"

We were using dev_priv->pm for runtime power management related state.
This patch renames it to "runtime_pm" which looks more apt.

v2: s/rpm/runtime_pm (Chris)

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com> #1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-7-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-6-chris@chris-wilson.co.uk
7 years agodrm/i915: Separate RPS and RC6 handling for CHV
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:03 +0000 (22:30 +0100)]
drm/i915: Separate RPS and RC6 handling for CHV

This patch separates enable/disable of RC6 and RPS for CHV.

v2: Fixed comment.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-6-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-5-chris@chris-wilson.co.uk
7 years agodrm/i915: Separate RPS and RC6 handling for VLV
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:02 +0000 (22:30 +0100)]
drm/i915: Separate RPS and RC6 handling for VLV

This patch separates enable/disable of RC6 and RPS for VLV.

v2: Removed unnecessary comments about forcewakes while enabling
RC6/RPS. Added changes to output turbo control status for VLV in
i915_frequency_info.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-5-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-4-chris@chris-wilson.co.uk
7 years agodrm/i915: Separate RPS and RC6 handling for BDW
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:01 +0000 (22:30 +0100)]
drm/i915: Separate RPS and RC6 handling for BDW

This patch separates RC6 and RPS enabling for BDW.
RC6/RPS Disabling are handled through gen6 functions.
PM Programming guide recommends a sequence within forcewakes to
configure RC6, RPS and ring frequencies in sequence. With this
patch the order is still maintained.

v2: Update sequence numbers in RC6 programming and comment about
intent of reset_rps during gen8_enable_rps. (Radoslaw)

v3: Rebase.

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-4-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-3-chris@chris-wilson.co.uk
7 years agodrm/i915: Remove superfluous IS_BDW checks and non-BDW changes from gen8_enable_rps
Sagar Arun Kamble [Tue, 10 Oct 2017 21:30:00 +0000 (22:30 +0100)]
drm/i915: Remove superfluous IS_BDW checks and non-BDW changes from gen8_enable_rps

This patch removes all IS_BROADWELL checks and non-BDW changes from
gen8_enable_rps as it is called only for BROADWELL.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-3-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-2-chris@chris-wilson.co.uk
7 years agodrm/i915: Separate RPS and RC6 handling for gen6+
Sagar Arun Kamble [Tue, 10 Oct 2017 21:29:59 +0000 (22:29 +0100)]
drm/i915: Separate RPS and RC6 handling for gen6+

This patch separates enable/disable of RC6 and RPS for gen6+
platforms prior to VLV.

v2: Fixed checkpatch issue. (Sagar)

Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com> #1
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1507360055-19948-2-git-send-email-sagar.a.kamble@intel.com
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010213010.7415-1-chris@chris-wilson.co.uk
7 years agodrm/i915/selftests: ditch the kernel context
Matthew Auld [Tue, 10 Oct 2017 13:30:30 +0000 (14:30 +0100)]
drm/i915/selftests: ditch the kernel context

There's really no good reason to be using the kernel context for the
huge-page livetests. Also with the introduction of commit bef27bdb6cfb
("drm/i915: Assert we do not try to expand VMA for hugepage inside GGTT")
we start hitting the bug on in the selftests, since the kernel context
will always return true for i915_vma_is_ggtt(), so now seems like the
opportune time to instead create our own context.

Fixes: 4049866f0913 ("drm/i915/selftests: huge page tests")
Fixes: bef27bdb6cfb ("drm/i915: Assert we do not try to expand VMA for hugepage inside GGTT")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010133030.12112-1-matthew.auld@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: Silently fallback to 4k scratch
Chris Wilson [Tue, 10 Oct 2017 11:10:05 +0000 (12:10 +0100)]
drm/i915: Silently fallback to 4k scratch

If we fail to allocate a 64k hugepage for scratch, we try again with a
normal 4k page (with some loss of efficiency at runtime). As we handle
this gracefully, we do not need a noisy allocation failure warning.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010111005.13625-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Cleanup South Error Interrupts
Mika Kahola [Tue, 10 Oct 2017 10:17:06 +0000 (13:17 +0300)]
drm/i915: Cleanup South Error Interrupts

Cleanup and parametrize the handling of South Error Interrupts (SERR_INT).

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630626-23806-6-git-send-email-mika.kahola@intel.com
7 years agodrm/i915: Favor for_each_pipe() macro
Mika Kahola [Tue, 10 Oct 2017 10:17:05 +0000 (13:17 +0300)]
drm/i915: Favor for_each_pipe() macro

Favor for_each_pipe() macro when looping through pipes.

v2: use 'enum pipe pipe' instead of 'i'

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630626-23806-5-git-send-email-mika.kahola@intel.com
7 years agodrm/i915: Fold IRQ pipe masks
Mika Kahola [Tue, 10 Oct 2017 10:17:04 +0000 (13:17 +0300)]
drm/i915: Fold IRQ pipe masks

Fold IRQ pipe masks into one loop instead of hardcoding per pipe.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630626-23806-4-git-send-email-mika.kahola@intel.com
7 years agodrm/i915: Remove I915_MAX_PIPES dependency for DDB allocation
Mika Kahola [Tue, 10 Oct 2017 10:17:03 +0000 (13:17 +0300)]
drm/i915: Remove I915_MAX_PIPES dependency for DDB allocation

Remove dependency for I915_MAX_PIPES by replacing it with
for_each_pipe() macro.

v2: use 'enum pipe pipe' instead of 'i'

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630626-23806-3-git-send-email-mika.kahola@intel.com
7 years agodrm/i915: Don't relay on I915_MAX_PIPES
Mika Kahola [Tue, 10 Oct 2017 10:17:02 +0000 (13:17 +0300)]
drm/i915: Don't relay on I915_MAX_PIPES

Let's remove the dependency on I915_MAX_PIPES. Instead, get the number
of pipes from platform information.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1507630626-23806-2-git-send-email-mika.kahola@intel.com
7 years agodrm/i915: Nuke the bogus kernel doc for i9xx_disable_pll()
Ville Syrjälä [Wed, 13 Sep 2017 14:08:57 +0000 (17:08 +0300)]
drm/i915: Nuke the bogus kernel doc for i9xx_disable_pll()

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170913140900.6972-6-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Pass crtc state to i9xx_enable_pll()
Ville Syrjälä [Wed, 13 Sep 2017 14:08:56 +0000 (17:08 +0300)]
drm/i915: Pass crtc state to i9xx_enable_pll()

Pass the crtc state to i9xx_enable_pll() and use it rather than
crtc->config.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170913140900.6972-5-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Parametrize CBR_DPLLBMD_PIPE defines
Ville Syrjälä [Wed, 13 Sep 2017 14:08:54 +0000 (17:08 +0300)]
drm/i915: Parametrize CBR_DPLLBMD_PIPE defines

Apply a bit of polish by parametrizing the CBR_DPLLBMD_PIPE defines.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170913140900.6972-3-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Use enum pipe for PCH transcoders
Ville Syrjälä [Tue, 10 Oct 2017 12:55:56 +0000 (15:55 +0300)]
drm/i915: Use enum pipe for PCH transcoders

One more place where we've failed to switch to enum pipe when talking
about PCH transcoders.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010125556.25086-1-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
7 years agodrm/i915: Increase atomic update vblank evasion time with lockdep
Daniel Vetter [Tue, 10 Oct 2017 09:18:16 +0000 (11:18 +0200)]
drm/i915: Increase atomic update vblank evasion time with lockdep

All our mmio writes take forever with lockdep due to the constant
lock acquire&dropping we do. Ville has some patches to only acquire
the mmio spinlocks once instead for every single mmio, but those
aren't ready yet.

As an interim solution just extend our budget slightly when lockdep is
enabled, to avoid the rare and sporadic noise in CI.

v2: I forgot to add the FIXME comment ...

Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=103169
References: https://bugs.freedesktop.org/show_bug.cgi?id=103124
References: https://bugs.freedesktop.org/show_bug.cgi?id=102403
References: https://bugs.freedesktop.org/show_bug.cgi?id=103020
References: https://bugs.freedesktop.org/show_bug.cgi?id=103019
References: https://bugs.freedesktop.org/show_bug.cgi?id=102723
References: https://bugs.freedesktop.org/show_bug.cgi?id=102544
References: https://bugs.freedesktop.org/show_bug.cgi?id=103180
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010091816.26898-1-daniel.vetter@ffwll.ch
7 years agodrm/i915: Use execlists_num_ports instead of size of array
Mika Kuoppala [Tue, 10 Oct 2017 11:48:57 +0000 (14:48 +0300)]
drm/i915: Use execlists_num_ports instead of size of array

There is function to tell how many ports we have, so use it.
We still have direct relationship with array size and port count,
so no harm was done.

Fixes: 76e70087d360 ("drm/i915: Make execlist port count variable")
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171010114857.13108-1-mika.kuoppala@intel.com
7 years agodrm/i915: Preallocate our mmu notifier workequeu to unbreak cpu hotplug deadlock
Daniel Vetter [Mon, 9 Oct 2017 16:44:00 +0000 (18:44 +0200)]
drm/i915: Preallocate our mmu notifier workequeu to unbreak cpu hotplug deadlock

4.14-rc1 gained the fancy new cross-release support in lockdep, which
seems to have uncovered a few more rules about what is allowed and
isn't.

This one here seems to indicate that allocating a work-queue while
holding mmap_sem is a no-go, so let's try to preallocate it.

Of course another way to break this chain would be somewhere in the
cpu hotplug code, since this isn't the only trace we're finding now
which goes through msr_create_device.

Full lockdep splat:

======================================================
WARNING: possible circular locking dependency detected
4.14.0-rc1-CI-CI_DRM_3118+ #1 Tainted: G     U
------------------------------------------------------
prime_mmap/1551 is trying to acquire lock:
 (cpu_hotplug_lock.rw_sem){++++}, at: [<ffffffff8109dbb7>] apply_workqueue_attrs+0x17/0x50

but task is already holding lock:
 (&dev_priv->mm_lock){+.+.}, at: [<ffffffffa01a7b2a>] i915_gem_userptr_init__mmu_notifier+0x14a/0x270 [i915]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #6 (&dev_priv->mm_lock){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __mutex_lock+0x86/0x9b0
       mutex_lock_nested+0x1b/0x20
       i915_gem_userptr_init__mmu_notifier+0x14a/0x270 [i915]
       i915_gem_userptr_ioctl+0x222/0x2c0 [i915]
       drm_ioctl_kernel+0x69/0xb0
       drm_ioctl+0x2f9/0x3d0
       do_vfs_ioctl+0x94/0x670
       SyS_ioctl+0x41/0x70
       entry_SYSCALL_64_fastpath+0x1c/0xb1

-> #5 (&mm->mmap_sem){++++}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __might_fault+0x68/0x90
       _copy_to_user+0x23/0x70
       filldir+0xa5/0x120
       dcache_readdir+0xf9/0x170
       iterate_dir+0x69/0x1a0
       SyS_getdents+0xa5/0x140
       entry_SYSCALL_64_fastpath+0x1c/0xb1

-> #4 (&sb->s_type->i_mutex_key#5){++++}:
       down_write+0x3b/0x70
       handle_create+0xcb/0x1e0
       devtmpfsd+0x139/0x180
       kthread+0x152/0x190
       ret_from_fork+0x27/0x40

-> #3 ((complete)&req.done){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       wait_for_common+0x58/0x210
       wait_for_completion+0x1d/0x20
       devtmpfs_create_node+0x13d/0x160
       device_add+0x5eb/0x620
       device_create_groups_vargs+0xe0/0xf0
       device_create+0x3a/0x40
       msr_device_create+0x2b/0x40
       cpuhp_invoke_callback+0xa3/0x840
       cpuhp_thread_fun+0x7a/0x150
       smpboot_thread_fn+0x18a/0x280
       kthread+0x152/0x190
       ret_from_fork+0x27/0x40

-> #2 (cpuhp_state){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       cpuhp_issue_call+0x10b/0x170
       __cpuhp_setup_state_cpuslocked+0x134/0x2a0
       __cpuhp_setup_state+0x46/0x60
       page_writeback_init+0x43/0x67
       pagecache_init+0x3d/0x42
       start_kernel+0x3a8/0x3fc
       x86_64_start_reservations+0x2a/0x2c
       x86_64_start_kernel+0x6d/0x70
       verify_cpu+0x0/0xfb

-> #1 (cpuhp_state_mutex){+.+.}:
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       __mutex_lock+0x86/0x9b0
       mutex_lock_nested+0x1b/0x20
       __cpuhp_setup_state_cpuslocked+0x52/0x2a0
       __cpuhp_setup_state+0x46/0x60
       page_alloc_init+0x28/0x30
       start_kernel+0x145/0x3fc
       x86_64_start_reservations+0x2a/0x2c
       x86_64_start_kernel+0x6d/0x70
       verify_cpu+0x0/0xfb

-> #0 (cpu_hotplug_lock.rw_sem){++++}:
       check_prev_add+0x430/0x840
       __lock_acquire+0x1420/0x15e0
       lock_acquire+0xb0/0x200
       cpus_read_lock+0x3d/0xb0
       apply_workqueue_attrs+0x17/0x50
       __alloc_workqueue_key+0x1d8/0x4d9
       i915_gem_userptr_init__mmu_notifier+0x1fb/0x270 [i915]
       i915_gem_userptr_ioctl+0x222/0x2c0 [i915]
       drm_ioctl_kernel+0x69/0xb0
       drm_ioctl+0x2f9/0x3d0
       do_vfs_ioctl+0x94/0x670
       SyS_ioctl+0x41/0x70
       entry_SYSCALL_64_fastpath+0x1c/0xb1

other info that might help us debug this:

Chain exists of:
  cpu_hotplug_lock.rw_sem --> &mm->mmap_sem --> &dev_priv->mm_lock

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&dev_priv->mm_lock);
                               lock(&mm->mmap_sem);
                               lock(&dev_priv->mm_lock);
  lock(cpu_hotplug_lock.rw_sem);

 *** DEADLOCK ***

2 locks held by prime_mmap/1551:
 #0:  (&mm->mmap_sem){++++}, at: [<ffffffffa01a7b18>] i915_gem_userptr_init__mmu_notifier+0x138/0x270 [i915]
 #1:  (&dev_priv->mm_lock){+.+.}, at: [<ffffffffa01a7b2a>] i915_gem_userptr_init__mmu_notifier+0x14a/0x270 [i915]

stack backtrace:
CPU: 4 PID: 1551 Comm: prime_mmap Tainted: G     U          4.14.0-rc1-CI-CI_DRM_3118+ #1
Hardware name: Dell Inc. XPS 8300  /0Y2MRG, BIOS A06 10/17/2011
Call Trace:
 dump_stack+0x68/0x9f
 print_circular_bug+0x235/0x3c0
 ? lockdep_init_map_crosslock+0x20/0x20
 check_prev_add+0x430/0x840
 __lock_acquire+0x1420/0x15e0
 ? __lock_acquire+0x1420/0x15e0
 ? lockdep_init_map_crosslock+0x20/0x20
 lock_acquire+0xb0/0x200
 ? apply_workqueue_attrs+0x17/0x50
 cpus_read_lock+0x3d/0xb0
 ? apply_workqueue_attrs+0x17/0x50
 apply_workqueue_attrs+0x17/0x50
 __alloc_workqueue_key+0x1d8/0x4d9
 ? __lockdep_init_map+0x57/0x1c0
 i915_gem_userptr_init__mmu_notifier+0x1fb/0x270 [i915]
 i915_gem_userptr_ioctl+0x222/0x2c0 [i915]
 ? i915_gem_userptr_release+0x140/0x140 [i915]
 drm_ioctl_kernel+0x69/0xb0
 drm_ioctl+0x2f9/0x3d0
 ? i915_gem_userptr_release+0x140/0x140 [i915]
 ? __do_page_fault+0x2a4/0x570
 do_vfs_ioctl+0x94/0x670
 ? entry_SYSCALL_64_fastpath+0x5/0xb1
 ? __this_cpu_preempt_check+0x13/0x20
 ? trace_hardirqs_on_caller+0xe3/0x1b0
 SyS_ioctl+0x41/0x70
 entry_SYSCALL_64_fastpath+0x1c/0xb1
RIP: 0033:0x7fbb83c39587
RSP: 002b:00007fff188dc228 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: ffffffff81492963 RCX: 00007fbb83c39587
RDX: 00007fff188dc260 RSI: 00000000c0186473 RDI: 0000000000000003
RBP: ffffc90001487f88 R08: 0000000000000000 R09: 00007fff188dc2ac
R10: 00007fbb83efcb58 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000003 R14: 00000000c0186473 R15: 00007fff188dc2ac
 ? __this_cpu_preempt_check+0x13/0x20

Note that this also has the minor benefit of slightly reducing the
critical section where we hold mmap_sem.

v2: Set ret correctly when we raced with another thread.

v3: Use Chris' diff. Attach the right lockdep splat.

v4: Repaint in Tvrtko's colors (aka don't report ENOMEM if we race and
some other thread managed to not also get an ENOMEM and successfully
install the mmu notifier. Note that the kernel guarantees that small
allocations succeed, so this never actually happens).

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sasha Levin <alexander.levin@verizon.com>
Cc: Marta Lofstedt <marta.lofstedt@intel.com>
Cc: Tejun Heo <tj@kernel.org>
References: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_3180/shard-hsw3/igt@prime_mmap@test_userptr.html
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102939
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009164401.16035-1-daniel.vetter@ffwll.ch
7 years agodrm/i915/bios: don't pass bdb to parsers that don't parse VBT directly
Jani Nikula [Thu, 28 Sep 2017 08:22:04 +0000 (11:22 +0300)]
drm/i915/bios: don't pass bdb to parsers that don't parse VBT directly

Hint that you're not supposed to look at VBT in these functions.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/b82c326be8c796a70bdc2bd1c479bbb6159f5cb0.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: parse SDVO device mapping from pre-parsed child devices
Jani Nikula [Thu, 28 Sep 2017 08:22:03 +0000 (11:22 +0300)]
drm/i915/bios: parse SDVO device mapping from pre-parsed child devices

We parse and store the child devices in
parse_general_definitions(). There is no need to parse the VBT block
again for SDVO device mapping. Do the same as we do in
parse_ddi_ports().

We no longer have access to child device size at this stage, but we also
don't need to worry about reading past the child device anymore. Instead
of a child device size check, do a mild optimization by limiting the
parsing to gens 3 through 7.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/c918d4173dd38a165295f1270cb16c2c01bd8cd1.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: merge parse_device_mapping() into parse_general_definitions()
Jani Nikula [Thu, 28 Sep 2017 08:22:02 +0000 (11:22 +0300)]
drm/i915/bios: merge parse_device_mapping() into parse_general_definitions()

They're both parsing the same block, and there's no need for them to be
split. The former also benefits from the range checks in the latter.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/64a292606ecbb0b8602e6c5523c5746573ec3944.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: cleanup comments and useless return
Jani Nikula [Thu, 28 Sep 2017 08:22:01 +0000 (11:22 +0300)]
drm/i915/bios: cleanup comments and useless return

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/4a95fb9d23d980830e3158d3c57258e6539965ce.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: remove an unnecessary temp variable
Jani Nikula [Thu, 28 Sep 2017 08:22:00 +0000 (11:22 +0300)]
drm/i915/bios: remove an unnecessary temp variable

Prepare for merging parse_device_mapping() into
parse_general_definitions(). No functional changes.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/f1c3621e2622f4debdfb4a2f5c1959845754ac04.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: don't initialize fields based on vbt version
Jani Nikula [Thu, 28 Sep 2017 08:21:59 +0000 (11:21 +0300)]
drm/i915/bios: don't initialize fields based on vbt version

In theory, these might clobber information for older VBT versions.

We might have to store the BDB version for later parsing, but currently
all code accessing these fields will only use them on newer platforms
with new enough BDB versions.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/0232d9cb258e8f83c4180cdb8aad1459a312ec2a.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: refactor parse general definitions
Jani Nikula [Thu, 28 Sep 2017 08:21:58 +0000 (11:21 +0300)]
drm/i915/bios: refactor parse general definitions

Early return on failures. Rename the variable for later merging with
parse_device_mappings().

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/785abb904a572752fec68d90d34efeb67774dc1f.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915/bios: parse DDI ports also for CHV for HDMI DDC pin and DP AUX channel
Jani Nikula [Thu, 28 Sep 2017 08:21:57 +0000 (11:21 +0300)]
drm/i915/bios: parse DDI ports also for CHV for HDMI DDC pin and DP AUX channel

While technically CHV isn't DDI, we do look at the VBT based DDI port
info for HDMI DDC pin and DP AUX channel. (We call these "alternate",
but they're really just something that aren't platform defaults.)

In commit e4ab73a13291 ("drm/i915: Respect alternate_ddc_pin for all DDI
ports") Ville writes, "IIRC there may be CHV system that might actually
need this."

I'm not sure why there couldn't be even more platforms that need this,
but start conservative, and parse the info for CHV in addition to DDI.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100553
Reported-by: Marek Wilczewski <mw@3cte.pl>
Cc: stable@vger.kernel.org
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/d0815082cb98487618429b62414854137049b888.1506586821.git.jani.nikula@intel.com
7 years agodrm/i915: avoid division by zero on cnl_calc_wrpll_link
Paulo Zanoni [Thu, 5 Oct 2017 21:38:42 +0000 (18:38 -0300)]
drm/i915: avoid division by zero on cnl_calc_wrpll_link

If for some unexpected reason the registers all read zero it's better
to WARN and return instead of dividing by zero and completely freezing
the machine.

I don't expect this to happen in the wild with the current code, but I
accidentally triggered the division by zero while doing some debugging
in an unusual environment.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171005213842.11423-2-paulo.r.zanoni@intel.com
7 years agodrm/i915: add the BXT and CNL DPLL registers to pipe_config_compare
Paulo Zanoni [Fri, 22 Sep 2017 20:53:42 +0000 (17:53 -0300)]
drm/i915: add the BXT and CNL DPLL registers to pipe_config_compare

Looks like we were missing them.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170922205343.16006-2-paulo.r.zanoni@intel.com
7 years agodrm/i915: s/sg_mask/sg_page_sizes/
Matthew Auld [Mon, 9 Oct 2017 11:00:24 +0000 (12:00 +0100)]
drm/i915: s/sg_mask/sg_page_sizes/

It's a little unclear what the sg_mask actually is, so prefer the more
meaningful name of sg_page_sizes.

Suggested-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110024.29114-1-matthew.auld@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
7 years agodrm/i915: Early rejection of mappable GGTT pin attempts for large bo
Chris Wilson [Mon, 9 Oct 2017 08:44:01 +0000 (09:44 +0100)]
drm/i915: Early rejection of mappable GGTT pin attempts for large bo

Currently, we reject attempting to pin a large bo into the mappable
aperture, but only after trying to create the vma. Under debug kernels,
repeatedly creating and freeing that vma for an oversized bo consumes
one-third of the runtime for pwrite/pread tests as it is spent on
kmalloc/kfree tracking. If we move the rejection to before creating that
vma, we lose some accuracy of checking against the fence_size as opposed
to object size, though the fence can never be smaller than the object.
Note that the vma creation itself will reject an attempt to create a vma
larger than the GTT so we can remove one redundant test.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-7-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Avoid evicting user fault mappable vma for pread/pwrite
Chris Wilson [Mon, 9 Oct 2017 08:44:00 +0000 (09:44 +0100)]
drm/i915: Avoid evicting user fault mappable vma for pread/pwrite

Both pread/pwrite GTT paths provide a fast fallback in case we cannot
map the whole object at a time. Currently, we use the fallback for very
large objects and for active objects that would require remapping, but
we can also add active fault mappable objects to the list that we want
to avoid evicting. The rationale is that such fault mappable objects are
in active use and to evict requires tearing down the CPU PTE and forcing
a page fault on the next access; more costly, and intefers with other
processes, than our per-page GTT fallback.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-6-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Try a minimal attempt to insert the whole object for relocations
Chris Wilson [Mon, 9 Oct 2017 08:43:59 +0000 (09:43 +0100)]
drm/i915: Try a minimal attempt to insert the whole object for relocations

As we have a lightweight fallback to insert a single page into the
aperture, try to avoid any heavier evictions when attempting to insert
the entire object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-5-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Check PIN_NONFAULT overlaps in evict_for_node
Chris Wilson [Mon, 9 Oct 2017 08:43:58 +0000 (09:43 +0100)]
drm/i915: Check PIN_NONFAULT overlaps in evict_for_node

If the caller says that he doesn't want to evict any other faulting
vma, honour that flag. The logic was used in evict_something, but not
the more specific evict_for_node, now being used as a preliminary probe
since commit 606fec956c0e ("drm/i915: Prefer random replacement before
eviction search").

Fixes: 606fec956c0e ("drm/i915: Prefer random replacement before eviction search")
Fixes: 821188778b9b ("drm/i915: Choose not to evict faultable objects from the GGTT")
References: https://bugs.freedesktop.org/show_bug.cgi?id=102490
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/20171009084401.29090-4-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Track user GTT faulting per-vma
Chris Wilson [Mon, 9 Oct 2017 08:43:57 +0000 (09:43 +0100)]
drm/i915: Track user GTT faulting per-vma

We don't wish to refault the entire object (other vma) when unbinding
one partial vma. To do this track which vma have been faulted into the
user's address space.

v2: Use a local vma_offset to tidy up a multiline unmap_mapping_range().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-3-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Consolidate get_fence with pin_fence
Chris Wilson [Mon, 9 Oct 2017 08:43:56 +0000 (09:43 +0100)]
drm/i915: Consolidate get_fence with pin_fence

Following the pattern now used for obj->mm.pages, use just pin_fence and
unpin_fence to control access to the fence registers. I.e. instead of
calling get_fence(); pin_fence(), we now just need to call pin_fence().
This will make it easier to reduce the locking requirements around
fence registers.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Pin fence for iomap
Chris Wilson [Mon, 9 Oct 2017 08:43:55 +0000 (09:43 +0100)]
drm/i915: Pin fence for iomap

Acquire the fence register for the iomap in i915_vma_pin_iomap() on
behalf of the caller.

We probably want for the caller to specify whether the fence should be
pinned for their usage, but at the moment all callers do want the
associated fence, or none, so take it on their behalf.

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/20171009084401.29090-1-chris@chris-wilson.co.uk
7 years agodrm/i915: Provide an assert for when we expect forcewake to be held
Chris Wilson [Mon, 9 Oct 2017 11:03:01 +0000 (12:03 +0100)]
drm/i915: Provide an assert for when we expect forcewake to be held

Add assert_forcewakes_active() (the complementary function to
assert_forcewakes_inactive) that documents the requirement of a
function for its callers to be holding the forcewake ref (i.e. the
function is part of a sequence over which RC6 must be prevented).

One such example is during ringbuffer reset, where RC6 must be held
across the whole reinitialisation sequence.

v2: Include debug information in the WARN so we know which fw domain is
missing.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110301.21705-5-chris@chris-wilson.co.uk
7 years agodrm/i915/selftests: Hold the rpm wakeref for the reset tests
Chris Wilson [Mon, 9 Oct 2017 11:03:00 +0000 (12:03 +0100)]
drm/i915/selftests: Hold the rpm wakeref for the reset tests

The lowlevel reset functions expect the caller to be holding the rpm
wakeref for the device access across the reset. We were not explicitly
doing this in the sefltest, so for simplicity acquire the wakeref for
the duration of all subtests.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110301.21705-4-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
7 years agodrm/i915: Hold forcewake for the duration of reset+restart
Chris Wilson [Mon, 9 Oct 2017 11:02:59 +0000 (12:02 +0100)]
drm/i915: Hold forcewake for the duration of reset+restart

Resetting the engine requires us to hold the forcewake wakeref to
prevent RC6 trying to happen in the middle of the reset sequence. The
consequence of an unwanted RC6 event in the middle is that random state
is then saved to the powercontext and restored later, which may
overwrite the mmio state we need to preserve (e.g. PD_DIR_BASE in the
legacy ringbuffer reset_ring_common()).

This was noticed in the live_hangcheck selftests when Haswell would
sporadically fail to restart during igt_reset_queue().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110301.21705-3-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
7 years agodrm/i915/selftests: Pretty print engine state when requests fail to start
Chris Wilson [Mon, 9 Oct 2017 11:02:58 +0000 (12:02 +0100)]
drm/i915/selftests: Pretty print engine state when requests fail to start

During hangcheck testing, we try to execute requests following the GPU
reset, and in particular want to try and debug when those fail.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110301.21705-2-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
7 years agodrm/i915: Make i915_engine_info pretty printer to standalone
Chris Wilson [Mon, 9 Oct 2017 11:02:57 +0000 (12:02 +0100)]
drm/i915: Make i915_engine_info pretty printer to standalone

We can use drm_printer to hide the differences between printk and
seq_printf, and so make the i915_engine_info pretty printer able to be
called from different contexts and not just debugfs. For instance, I
want to use the pretty printer to debug kselftests.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009110301.21705-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
7 years agodrm/i915: Assert we do not try to expand VMA for hugepage inside GGTT
Chris Wilson [Mon, 9 Oct 2017 09:20:19 +0000 (10:20 +0100)]
drm/i915: Assert we do not try to expand VMA for hugepage inside GGTT

We only apply the hugepage PD redirection inside the ppGTT, so during
i915_vma_insert() we want to exclude the GGTT from the additional
alignment constraints (thereby avoiding the extra GTT pressure from
fragmentation). Add an assert to document that intention alongside the
comment.

v2: After discussion with Matthew, make it a blanket GGTT ban
(previously we allowed the expansion for appgtt, and so indirectly
ggtt). There are issues we need to fix before allowing the current
appgtt to be used with hugepages, and if we do, we probably want more
care over when to expand/align, as the mappable aperture inside the ggtt
is precious.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009092019.20747-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
7 years agodrm/i915: Use intel_get_pipe_timings() and intel_mode_from_pipe_config() in intel_crt...
Ville Syrjälä [Fri, 1 Apr 2016 18:48:50 +0000 (21:48 +0300)]
drm/i915: Use intel_get_pipe_timings() and intel_mode_from_pipe_config() in intel_crtc_mode_get()

Eliminate the duplicate code for pipe timing readout in
intel_crtc_mode_get() by using the functions we use for the normal state
readout.

v2: Store dotclock in adjusted_mode instead of the final mode

Cc: dri-devel@lists.freedesktop.org
Cc: Rob Kramer <rob@solution-space.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1459536530-17754-1-git-send-email-ville.syrjala@linux.intel.com
7 years agodrm/i915: Read timings from the correct transcoder in intel_crtc_mode_get()
Ville Syrjälä [Fri, 1 Apr 2016 15:37:25 +0000 (18:37 +0300)]
drm/i915: Read timings from the correct transcoder in intel_crtc_mode_get()

intel_crtc->config->cpu_transcoder isn't yet filled out when
intel_crtc_mode_get() gets called during output probing, so we should
not use it there. Instead intel_crtc_mode_get() figures out the correct
transcoder on its own, and that's what we should use.

If the BIOS boots LVDS on pipe B, intel_crtc_mode_get() would actually
end up reading the timings from pipe A instead (since PIPE_A==0),
which clearly isn't what we want.

It looks to me like this may have been broken by
commit eccb140bca67 ("drm/i915: hw state readout&check support for cpu_transcoder")
as that one removed the early initialization of cpu_transcoder from
intel_crtc_init().

Cc: stable@vger.kernel.org
Cc: dri-devel@lists.freedesktop.org
Cc: Rob Kramer <rob@solution-space.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reported-by: Rob Kramer <rob@solution-space.com>
Fixes: eccb140bca67 ("drm/i915: hw state readout&check support for cpu_transcoder")
References: https://lists.freedesktop.org/archives/dri-devel/2016-April/104142.html
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1459525046-19425-1-git-send-email-ville.syrjala@linux.intel.com
7 years agodrm/i915: enable platform support for 2M pages
Matthew Auld [Fri, 6 Oct 2017 22:18:33 +0000 (23:18 +0100)]
drm/i915: enable platform support for 2M pages

For gen8+ platforms which support the 48b PPGTT, enable platform level
support for 2M pages. Also enable for mock testing.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-22-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-21-chris@chris-wilson.co.uk
7 years agodrm/i915: enable platform support for 64K pages
Matthew Auld [Fri, 6 Oct 2017 22:18:32 +0000 (23:18 +0100)]
drm/i915: enable platform support for 64K pages

For gen9+ enable platform level support for 64K pages. Also enable for
mock testing.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-21-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-20-chris@chris-wilson.co.uk
7 years agodrm/i915: disable platform support for vGPU huge gtt pages
Matthew Auld [Fri, 6 Oct 2017 22:18:31 +0000 (23:18 +0100)]
drm/i915: disable platform support for vGPU huge gtt pages

Currently gvt gtt handling doesn't support huge page entries, so disable
for now.

v2: remove useless 48b PPGTT check

Suggested-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-20-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-19-chris@chris-wilson.co.uk
7 years agodrm/i915/selftests: mix huge pages
Matthew Auld [Fri, 6 Oct 2017 22:18:30 +0000 (23:18 +0100)]
drm/i915/selftests: mix huge pages

Try to mix sg page sizes for 4K, 64K and 2M pages.

v2: s/BIT(x) >> 12/BIT(x) >> PAGE_SHIFT/

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-19-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-18-chris@chris-wilson.co.uk
7 years agodrm/i915/selftests: huge page tests
Matthew Auld [Fri, 6 Oct 2017 22:18:29 +0000 (23:18 +0100)]
drm/i915/selftests: huge page tests

v2: mock test page support configurations and add MI_STORE_DWORD test

v3: run all mockable huge page tests on all platforms via the mock_device

v4: add pin_update regression test
    various improvements suggested by Chris

v5: fix issues reported by kbuild
    test single sg spanning multiple page sizes
    don't explode when running the live-tests through the appgtt

v6: lots of improvements from Chris

v7: run on each engine for igt_write_huge
    add simple tmpfs fallback test

v8: size_t is bad
    don't break the i386 build

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-18-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-17-chris@chris-wilson.co.uk
7 years agodrm/i915/debugfs: include some gtt page size metrics
Matthew Auld [Fri, 6 Oct 2017 22:18:28 +0000 (23:18 +0100)]
drm/i915/debugfs: include some gtt page size metrics

Good to know, mostly for debugging purposes.

v2: some improvements from Chris

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-17-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-16-chris@chris-wilson.co.uk
7 years agodrm/i915: accurate page size tracking for the ppgtt
Matthew Auld [Fri, 6 Oct 2017 22:18:27 +0000 (23:18 +0100)]
drm/i915: accurate page size tracking for the ppgtt

Now that we support multiple page sizes for the ppgtt, it would be
useful to track the real usage for debugging purposes.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-16-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-15-chris@chris-wilson.co.uk
7 years agodrm/i915: support 64K pages for the 48b PPGTT
Matthew Auld [Fri, 6 Oct 2017 22:18:26 +0000 (23:18 +0100)]
drm/i915: support 64K pages for the 48b PPGTT

Support inserting 64K pages into the 48b PPGTT.

v2: check for 64K scratch

v3: we should only have to re-adjust maybe_64K at every sg interval

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-15-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-14-chris@chris-wilson.co.uk
7 years agodrm/i915: add support for 64K scratch page
Matthew Auld [Fri, 6 Oct 2017 22:18:25 +0000 (23:18 +0100)]
drm/i915: add support for 64K scratch page

Before we can fully enable 64K pages, we need to first support a 64K
scratch page if we intend to support the case where we have object sizes
< 2M, since any scratch PTE must also point to a 64K region.  Without
this our 64K usage is limited to objects which completely fill the
page-table, and therefore don't need any scratch.

v2: add reminder about why 48b PPGTT

Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-14-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-13-chris@chris-wilson.co.uk
7 years agodrm/i915: support 2M pages for the 48b PPGTT
Matthew Auld [Fri, 6 Oct 2017 22:18:24 +0000 (23:18 +0100)]
drm/i915: support 2M pages for the 48b PPGTT

Support inserting 2M gtt pages into the 48b PPGTT.

v2: sanity check sg->length against page_size

v3: don't recalculate rem on each loop
    whitespace breakup

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-13-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-12-chris@chris-wilson.co.uk
7 years agodrm/i915: disable GTT cache for 2M pages
Matthew Auld [Fri, 6 Oct 2017 22:18:23 +0000 (23:18 +0100)]
drm/i915: disable GTT cache for 2M pages

When SW enables the use of 2M/1G pages, it must disable the GTT cache.

v2: don't disable for Cherryview which doesn't even support 48b PPGTT!

v3: explicitly check that the system does support 2M/1G pages

v4: split WA and decision logic

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-12-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-11-chris@chris-wilson.co.uk
7 years agodrm/i915: enable IPS bit for 64K pages
Matthew Auld [Fri, 6 Oct 2017 22:18:22 +0000 (23:18 +0100)]
drm/i915: enable IPS bit for 64K pages

Before we can enable 64K pages through the IPS bit, we must first enable
it through MMIO, otherwise the page-walker will simply ignore it.

v2: add comment mentioning that 64K is BDW+

v3: move to more suitable home

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-11-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-10-chris@chris-wilson.co.uk
7 years agodrm/i915: align 64K objects to 2M
Matthew Auld [Fri, 6 Oct 2017 22:18:21 +0000 (23:18 +0100)]
drm/i915: align 64K objects to 2M

We can't mix 64K and 4K pte's in the same page-table, so for now we
align 64K objects to 2M to avoid any potential mixing. This is
potentially wasteful but in reality shouldn't be too bad since this only
applies to the virtual address space of a 48b PPGTT.

v2: don't separate logically connected ops

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-10-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-9-chris@chris-wilson.co.uk
7 years agodrm/i915: align the vma start to the largest gtt page size
Matthew Auld [Fri, 6 Oct 2017 22:18:20 +0000 (23:18 +0100)]
drm/i915: align the vma start to the largest gtt page size

For the 48b PPGTT try to align the vma start address to the required
page size boundary to guarantee we use said page size in the gtt. If we
are dealing with multiple page sizes, we can't guarantee anything and
just align to the largest. For soft pinning and objects which need to be
tightly packed into the lower 32bits we don't force any alignment.

v2: various improvements suggested by Chris

v3: use set_pages and better placement of page_sizes

v4: prefer upper_32_bits()

v5: assign vma->page_sizes = vma->obj->page_sizes directly
    prefer sizeof(vma->page_sizes)

v6: fixup checking of end to exclude GGTT (which are assumed to be
    limited to 4G).

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-9-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-8-chris@chris-wilson.co.uk
7 years agodrm/i915: introduce vm set_pages/clear_pages
Matthew Auld [Fri, 6 Oct 2017 22:18:19 +0000 (23:18 +0100)]
drm/i915: introduce vm set_pages/clear_pages

Move the setting/clearing of the vma->pages to a vm operation. Doing so
neatens things up a little, but more importantly gives us a sane place
to also set/clear the vma->pages_sizes, which we introduce later in
preparation for supporting huge-pages.

v2: remove redundant vma->pages check

v3: GEM_BUG_ON(vma->pages) following i915_vma_remove

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-8-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-7-chris@chris-wilson.co.uk
7 years agodrm/i915: introduce page_size members
Matthew Auld [Fri, 6 Oct 2017 22:18:18 +0000 (23:18 +0100)]
drm/i915: introduce page_size members

In preparation for supporting huge gtt pages for the ppgtt, we introduce
page size members for gem objects.  We fill in the page sizes by
scanning the sg table.

v2: pass the sg_mask to set_pages

v3: calculate the sg_mask inline with populating the sg_table where
possible, and pass to set_pages along with the pages.

v4: bunch of improvements from Joonas

v5: fix num_pages blunder
    introduce i915_sg_page_sizes helper

v6: prefer GEM_BUG_ON(sizes == 0)

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel@ffwll.ch>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-7-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-6-chris@chris-wilson.co.uk
7 years agodrm/i915: push set_pages down to the callers
Matthew Auld [Fri, 6 Oct 2017 22:18:17 +0000 (23:18 +0100)]
drm/i915: push set_pages down to the callers

Each backend is now responsible for calling __i915_gem_object_set_pages
upon successfully gathering its backing storage. This eliminates the
inconsistency between the async and sync paths, which stands out even
more when we start throwing around an sg_mask in a later patch.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-6-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-5-chris@chris-wilson.co.uk
7 years agodrm/i915: introduce page_sizes field to dev_info
Matthew Auld [Fri, 6 Oct 2017 22:18:16 +0000 (23:18 +0100)]
drm/i915: introduce page_sizes field to dev_info

In preparation for huge gtt pages expose page_sizes as part of the
device info, to indicate the page sizes supported by the HW.  Currently
only 4K is supported.

v2: s/page_size_mask/page_sizes/

v3: introduce I915_GTT_MAX_PAGE_SIZE

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-5-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-4-chris@chris-wilson.co.uk
7 years agodrm/i915/gemfs: enable THP
Matthew Auld [Fri, 6 Oct 2017 22:18:15 +0000 (23:18 +0100)]
drm/i915/gemfs: enable THP

Enable transparent-huge-pages through gemfs by mounting with
huge=within_size.

v2: sprinkle within_size comment

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-4-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-3-chris@chris-wilson.co.uk
7 years agodrm/i915: introduce simple gemfs
Matthew Auld [Fri, 6 Oct 2017 22:18:14 +0000 (23:18 +0100)]
drm/i915: introduce simple gemfs

Not a fully blown gemfs, just our very own tmpfs kernel mount. Doing so
moves us away from the shmemfs shm_mnt, and gives us the much needed
flexibility to do things like set our own mount options, namely huge=
which should allow us to enable the use of transparent-huge-pages for
our shmem backed objects.

v2: various improvements suggested by Joonas

v3: move gemfs instance to i915.mm and simplify now that we have
file_setup_with_mnt

v4: fallback to tmpfs shm_mnt upon failure to setup gemfs

v5: make tmpfs fallback kinder

v5: better gemfs failure message
    flags variable

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: linux-mm@kvack.org
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-3-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-2-chris@chris-wilson.co.uk
7 years agomm/shmem: introduce shmem_file_setup_with_mnt
Matthew Auld [Fri, 6 Oct 2017 22:18:13 +0000 (23:18 +0100)]
mm/shmem: introduce shmem_file_setup_with_mnt

We are planning to use our own tmpfs mnt in i915 in place of the
shm_mnt, such that we can control the mount options, in particular
huge=, which we require to support huge-gtt-pages. So rather than roll
our own version of __shmem_file_setup, it would be preferred if we could
just give shmem our mnt, and let it do the rest.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: linux-mm@kvack.org
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006145041.21673-2-matthew.auld@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171006221833.32439-1-chris@chris-wilson.co.uk
7 years agodrm/i915/execlists: Add a comment for the extra MI_ARB_ENABLE
Chris Wilson [Thu, 5 Oct 2017 19:10:05 +0000 (20:10 +0100)]
drm/i915/execlists: Add a comment for the extra MI_ARB_ENABLE

Michel Thierry noticed that we were applying WaDisableCtxRestoreArbitration
even to gen9, which does not require the w/a. The rationale is that we
need to enable MI arbitration for execlists to work, and to be safe we
do that before every batch (in addition to every context switch into the
batch). Since this is not clear from the single line comment suggesting
the MI_ARB_ENABLE is solely for the w/a, add a little more detail.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171005191005.13462-1-chris@chris-wilson.co.uk
Reviewed-by: Michel Thierry <michel.thierry@intel.com>