If ->detect is called too soon after a hot plug event, the sink may not
be ready yet. So try up to 3 times with 1ms sleeps in between tries to
get the data (spec dictates that receivers must be ready to respond within
1ms and that sources should try 3 times).
See section 9.1 of the 1.1a DisplayPort spec.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
When a hotplug event is received, we need to check the receiver cap bits
in case they've changed (as they might with a hub or chain config).
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Especially after a hotplug or power status change, the sink may not
reply immediately to a link status query. So retry 3 times per the spec
to really make sure nothing is there.
See section 9.1 of the 1.1a DisplayPort spec.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
The Intel HDMI encoder can support 8bpc or 12bpc. Set the appropriate
value based on the pipe bpp when configuring the output.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
The pipe may be driving various bpp values depending on the display
configuration, so take that into account when calculating link bandwidth
requirements.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
Updating the planes is device specific, so create a new display callback
and use it in pipe_set_base. (In fact we could go even further, valid
display plane bits have changed with each generation, as has tiled
buffer handling.)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
Figuring out which pipe bpp to use is a bit painful. It depends on both
the encoder and display configuration attached to a pipe. For instance,
to drive a 24bpp framebuffer out to an 18bpp panel, we need to use 6bpc
on the pipe but also enable dithering. But driving that same
framebuffer to a DisplayPort output on another pipe means using 8bpc and
no dithering.
So split out and enhance the code to handle the various cases, returning
an appropriate pipe bpp as well as whether dithering should be enabled.
Save the resulting pipe bpp in the intel_crtc struct for use by encoders
in calculating bandwidth requirements (defaults to 24bpp on pre-ILK).
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
This may not be the default value, so pull the bpc out of the pipe reg
and write it to the DP transcoder so proper dithering and signaling
occurs.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
These bits are reserved on ILK+ (ILK+ provides this feature in the
transcoder and pipe configuration instead, which we already set).
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
This lets us make the various IRQ functions static and helps avoid
problems like the one fixed in "drm/i915: Use chipset-specific irq
installers" where one of the exported functions was called rather than
the chipset specific version.
This also fixes a UMS-mode bug -- the correct irq functions for IRL
and later chips were only getting loaded in the KMS path.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
As pointed out by Dan Carpenter, it was seemingly possible to hit an error
whilst mapping the buffer for the regs (except the only likely error
returns should not happen during init) and so leak a pin count on the
bo. To handle this we would need to reacquire the struct mutex, so for
simplicity rearrange for the lock to be held for the entire function.
For extra pedagogy, test that we only call init once.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
They use the same register interfaces, so we can simply enable the
existing code on IVB.
v2:
- resolve conflict with ring freq scaling, we can enable it too
v3:
- resolve conflict again, this time on drm-intel-next
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
Lots of register access in these functions, some of which requires the
struct mutex.
These functions now hold the struct mutex across the calls to
i915_save_display and i915_restore_display, and so the internal mutex
calls in those functions have been removed. To ensure that no-one else
was calling them (and hence violating the new required locking
invarient), those functions have been made static.
gen6_enable_rps locks the struct mutex, and so i915_restore_state
unlocks the mutex around calls to that function.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Provide a parameter to disable hanghcheck. This is useful mostly for
developers trying to debug known problems, and probably should not be
touched by normal users.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
The ring frequency scaling table tells the PCU to treat certain GPU
frequencies as if they were a given CPU frequency for purposes of
scaling the ring frequency. Normally the PCU will scale the ring
frequency based on the CPU P-state, but with the table present, it will
also take the GPU frequency into account.
The main downside of keeping the ring frequency high while the CPU is
at a low frequency (or asleep altogether) is increased power
consumption. But then if you're keeping your GPU busy, you probably
want the extra performance.
v2:
- add units to debug table header (from Eric)
- use tsc_khz as a fallback if the cpufreq driver doesn't give us a freq
(from Chris)
v3:
- fix comments & debug output
- remove unneeded force wake get/put
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
Konstantin Belousov pointed out that 4697995b98 replaced the generic
i915_driver_irq_*install() functions with chipset specific routines
accessible only through driver->irq_*install(). So update the sanity
check in i915_request_wait() to match.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
The failure is as follows:
1. Userspace gets forcewake lock, lock count >=1
2. GPU hang/reset occurs (forcewake bit is reset)
3. count is now incorrect
The failure can only occur when using the forcewake userspace lock.
This has the unfortunate consequence of messing up the driver as well as
userspace, unless userspace closes the debugfs file, the kernel will
never end up waking the GT since the refcount will be > 1.
The solution is to try to recover the correct forcewake state based on
the refcount. There is a period of time where userspace reads/writes may
occur after the reset, before the GT has been forcewaked. The interface
was never designed to be a perfect solution for userspace reads/writes,
and the kernel portion is fixed by this patch.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
When auditing the locking in i915_gem.c (for a prospective change which
I then abandoned), I noticed two places where struct_mutex is not held
across GEM object manipulations that would usually require it.
Since one is in initial setup and the other in driver unload, I'm
guessing the mutex is not required for either; but post a patch in case
it is.
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The interface to ->truncate_range is changing very slightly: once "tmpfs:
take control of its truncate_range" has been applied, this can be applied.
For now there is only a slight inefficiency while this remains unapplied,
but it will soon become essential for managing shmem's use of swap.
Change i915_gem_object_truncate() to use shmem_truncate_range() directly:
which should also spare i915 later change if we switch from
inode_operations->truncate_range to file_operations->fallocate.
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Keith Packard <keithp@keithp.com>
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Soon tmpfs will stop supporting ->readpage and read_cache_page_gfp(): once
"tmpfs: add shmem_read_mapping_page_gfp" has been applied, this patch can
be applied to ease the transition.
Make i915_gem_object_get_pages_gtt() use shmem_read_mapping_page_gfp() in
the one place it's needed; elsewhere use shmem_read_mapping_page(), with
the mapping's gfp_mask properly initialized.
Forget about __GFP_COLD: since tmpfs initializes its pages with memset,
asking for a cold page is counter-productive.
Include linux/shmem_fs.h also in drm_gem.c: with shmem_file_setup() now
declared there too, we shall remove the prototype from linux/mm.h later.
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Keith Packard <keithp@keithp.com>
Cc: Dave Airlie <airlied@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When auditing the locking in i915_gem.c (for a prospective change which I
then abandoned), I noticed two places where struct_mutex is not held
across GEM object manipulations that would usually require it. Since one
is in initial setup and the other in driver unload, I'm guessing the mutex
is not required for either; but post a patch in case it is.
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
This makes things a little clearer and prevents us from running old code
on a new chipset that may not be supported.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewied-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Even if the object is no longer in the GTT domain, there may still be
a user space mapping which needs to be released.
Without this fix, render-based text (mostly in firefox) would
occasionally get corrupted when the system was under load.
Signed-off-by: Keith Packard <keithp@keithp.com>
This reverts commit 4a684a4117.
Userland has always been required to set the object's domain to GTT
before using it through a GTT mapping, it's not something that the
kernel is supposed to enforce. (The pagefault support is so that we
can handle multiple mappings without userland having to pin across
them, not so that userland can use GTT after GPU domains without
telling the kernel).
Fixes 19.2% +/- 0.8% (n=6) performance regression in cairo-gl
firefox-talos-gfx on my T420 latop.
Signed-off-by: Keith Packard <keithp@keithp.com>
The failure appeared in dmesg as:
[drm:i915_hangcheck_ring_idle] *ERROR* Hangcheck timer elapsed... blt
ring idle [waiting on 35064155, at 35064155], missed IRQ?
This works around that problem on by making the blitter command
streamer write interrupt state to the Hardware Status Page when a
MI_USER_INTERRUPT command is decoded, which appears to force the seqno
out to memory before the interrupt happens.
v1->v2: Moved to prior interrupt handler installation and RMW flags as
per feedback.
v2->v3: Removed RMW of flags (by anholt)
Cc: stable@kernel.org
Signed-off-by: Daniel J Blueman <daniel.blueman@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk> [v1]
Tested-by: Eric Anholt <eric@anholt.net> [v1,v3]
(incidence of the bug with a testcase went from avg 2/1000 to
0/12651 in the latest test run (plus more for v1))
Tested-by: Kenneth Graunke <kenneth@whitecape.org> [v1]
Tested-by: Robert Hooker <robert.hooker@canonical.com> [v1]
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=33394
Signed-off-by: Dave Airlie <airlied@redhat.com>
It seems to me that we are leaking 'user_pages' in
drivers/gpu/drm/i915/i915_gem.c::i915_gem_shmem_pread_slow() if
read_cache_page_gfp() fails.
Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Improves full-screen openarena on my laptop 20.3% +/- 4.0% (n=3)
Improves 800x600 nexuiz on my laptop 12.3% +/- 0.1% (n=3)
We have more room to improve with doing LLC caching for display using
GFDT, and in doing LLC+MLC caching, but this was an easy performance
win and incremental improvement toward those two.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The simplest and common method for ensuring scanout coherency on all
chipsets is to mark the scanout buffers as uncached (and for
userspace to remember to flush the render cache every so often).
We can improve upon this for later generations by marking scanout
objects as GFDT and only flush those cachelines when required. However,
we start simple.
[v2: Move the set to uncached above the clflush. Otherwise, we'd skip
the clflush and try to scan out data that was still sitting in the
cache.]
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We need to perform a few operations in order to move the object into the
display plane (where it can be accessed coherently by the display
engine) that are important for future safety to forbid whilst pinned. As a
result, we want to need to perform some of the operations before pinning,
but some are required once we have been bound into the GTT. So combine
the pinning performed by all the callers with set_to_display_plane(), so
this complication is contained within the single function.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
