The current enabling of bus mastering in the drm midlayer allows a large
race condition under kexec. When a kexec'ed kernel re-enables bus mastering
for the GPU, previously setup dma blocks may cause writes to random pieces
of memory. On radeon the writeback mechanism can cause these sorts of issues.

This patch doesn't fix the problem, but it moves the bus master enable under
the individual drivers control so they can move enabling it until later in
their load cycle and close the race.

Fix for radeon kms driver will be in a follow-up patch.
Signed-off-by: NDave Airlie <airlied@redhat.com>

We store stuff in texdw[7] so this array needs to have 8 elements.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: NJerome Glisse <jglisse@redhat.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

evergreen and northern island gpu needs more informations for 2D tiling
than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace
to provide those.

The v8 cs checking change to track color view on r6xx/r7xx doesn't
affect old userspace as old userspace always emited 0 for this register.

v2 fix r6xx/r7xx 2D tiling computation
v3 fix r6xx/r7xx height align for untiled surface & add support for
   tile split on evergreen and newer
v4 improve tiling debugging output
v5 fix tile split code for evergreen and newer
v6 set proper tile split for crtc register
v7 fix tile split limit value
v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs
   checking, update safe reg for r600, evergreen and cayman.
   Evergreen checking need some work around for stencil alignment
   issues
v9 fix tile split value range, fix compressed texture handling and
   mipmap calculation, allow evergreen check to be silencious in
   front of current broken userspace (depth/stencil alignment issue)
v10 fix eg 3d texture and compressed texture, fix r600 depth array,
    fix r600 color view computation, add support for evergreen stencil
    split
v11 more verbose debugging in some case
Signed-off-by: NJerome Glisse <jglisse@redhat.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

v2: agd5f: add strmout CS checking, copy_dw register checking

v3: agd5f: don't use cs_check_reg() for copy_dw checking as it
will incorrectly patch the command stream for certain regs.

v4: agd5f: add warning if safe reg check fails for copy_dw

v5: agd5f: add stricter checking for 6xx/7xx

v6: agd5f: add range checking for copy_dw on eg+,
add sx_surface_sync to safe reg list for 7xx.

v7: agd5f: add stricter checking for eg+
Signed-off-by: NMarek Olšák <maraeo@gmail.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The first parameter should be "number of elements" and the second parameter
should be "element size".
Signed-off-by: NAxel Lin <axel.lin@gmail.com>
Acked-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Bump driver minor to signal availability of the page-flip ioctl.
Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: NJakob Bornecrantz <jakob@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

And assign the initial width and height to the minimum in that case.
Strange values (-1) from these registers have been reported by users.
Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: NJakob Bornecrantz <jakob@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NJakob Bornecrantz <jakob@vmware.com>
Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NJakob Bornecrantz <jakob@vmware.com>
Reviewed-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NJakob Bornecrantz <jakob@vmware.com>
Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NJakob Bornecrantz <jakob@vmware.com>
Reviewed-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Pending events may have stale pointer references to struct drm_file objects
after a file has been closed, but before the event is supposed to be
attached to the drm file. Remove such events on file close.

Tested with "modetest".
Signed-off-by: NThomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: NJakob Bornecrantz <jakob@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NJakob Bornecrantz <jakob@vmware.com>
Reviewed-by: NThomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

CEA actually specifies an interlaced mode with even vtotal and
supplies a diagram showing how this is supposed to work.

Note that interlaced modes with an even vtotal seem to be a fairly
recent invention. All modelines lore I could dig up with googling says
that vtotal for interlaced modes _needs_ to be odd. But the even
modelines in CEA are not a spec-bug, there's a figure in CEA-861-E
called "Figure 5 Special Interlaced Video Format Timing (Even Vtotal)"
that explains how it's supposed to work. Furthermore intel Bspec
explicitly mentions that both odd and even interlaced vtotal are
supported (VTOTAL register in the south display engine of PCH split
chips).
Acked-by: NAdam Jackson <ajax@redhat.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The drm drivers set the fb_info->pixmap fields without setting
fb_info->pixmap.addr. If this is not set the fb core will overwrite
these all fb_info->pixmap fields anyway, so there is not much point
in setting them in the first place.

[airlied: dropped nvidiafb piece - not mine]
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Creating a range property is a common pattern, so create
a convenience function for this and use it where appropriate.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Creating an enum property is a common pattern, so create
a convenience function for this and use it where appropriate.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

info->fix.visual already is correctly set from drm_fb_helper_fill_fix.
info->fix.line_length is also set from drm_fb_helper_fill_fix,
so drm_fb_helper_set_par directly instead of a custom
exynos_drm_fbdev_set_par.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Tested-by: NInki Dae <inki.dae@samsung.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

list_for_each_entry_safe is for walking a list safe against removal
of entries. Here, no entries are removed, so use list_for_each_entry.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Several comments above functions say that the caller must hold the
mode_config lock, but the functions take the lock themselves. Fix
the comments.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

crtc_id is set but never used, so remove it from struct
drm_fb_helper_crtc.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

conn_limit is set but never used. Remove it from struct
drm_fb_helper.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

drm_fb_helper_on|off currently manually searches for encoders
to turn on/off. Make this simpler by using the helper function.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

drm_mode_crtc_set_gamma_size returns boolean true for success
and false for failure. This is not very kernel conform, so
change it to return 0 for success and a propert error code
otherwise. Noone checks the return value, so no users have to
be fixed.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

drm_setup_crtcs allocated modes using drm_mode_duplicate. Free
them in drm_fb_helper_crtc_free.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Modes are created using drm_mode_create which does a
drm_mode_object_get, so use drm_mode_destroy in drm_mode_remove
which does a drm_mode_object_put.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

to add the missing drm_mode_object_put for that mode.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

drm_mode_config_init initializes the idr with idr_init, so
add the missing counterparts in drm_mode_config_cleanup.
Signed-off-by: NSascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Not likely this will be implemented anytime soon, but for
completeness...
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Previous patch only updates r3xx+.  It's not likely
anyone will use this on r1xx/r2xx, but add it for consistency.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

This allows to avoid talking to a non-responding bus repeatedly until we
finally timeout after 15 attempts. We can do this by catching the -ENXIO
error, provided by i2c_algo_bit:bit_doAddress call.

Within the bit_doAddress we already try 3 times to get the edid data, so
if the routine tells us that bus is not responding, it is mostly pointless
to keep re-trying those attempts over and over again until we reach final
number of retries.

This change should fix https://bugs.freedesktop.org/show_bug.cgi?id=41059
and improve overall edid detection timing by 10-30% in most cases, and by
a much larger margin in case of phantom outputs (up to 30x in one worst
case).

Timing results for i915-powered machines for 'time xrandr' command:
Machine 1: from 0.840s to 0.290s
Machine 2: from 0.315s to 0.280s
Machine 3: from +/- 4s to 0.184s

Timing results for HD5770 with 'time xrandr' command:
Machine 4: from 3.210s to 1.060s
Reviewed-by: NChris Wilson <chris@hchris-wilson.co.uk>
Reviewed-by: NKeith Packard <keithp@keithp.com>
Tested-by: NSean Finney <seanius@seanius.net>
Tested-by: NSoren Hansen <soren@linux2go.dk>
Tested-by: NHernando Torque <sirius@sonnenkinder.org>
Tested-by: NMike Lothian <mike@fireburn.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=41059Signed-off-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Starting with DCE3 hardware, atom contains a general purpose
ProcessI2cChannelTransaction similar to ProcessAuxChannelTransaction.

Add an implementation using the atom tables for DCE3+ hardware.

This should be a little less CPU intensive than bit banging and
may work better in certain cases.

Enable it by setting the radeon hw_i2c module parameter to 1.  E.g.,
radeon.hw_i2c=1
on the kernel command line in grub.
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

The VESA specification suggests a 2.2 ms timeout on DDC channels.
Use exactly that (as the i915 driver does) instead of hard-coding a
jiffy count.
Signed-off-by: NJean Delvare <jdelvare@suse.de>
Reviewed-by: NKeith Packard <keithp@keithp.com>
Reviewed-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

A udelay value of 20 leads to an I2C bus running at only 25 kbps. I2C
devices can typically operate faster than this, 50 kbps should be fine
for all devices (and compliant devices can always stretch the clock if
needed.)

FWIW, the vast majority of framebuffer drivers set udelay to 10
already. So set it to 10 in DRM drivers too, this will make EDID block
reads faster. We might even lower the udelay value later if no problem
is reported.
Signed-off-by: NJean Delvare <jdelvare@suse.de>
Acked-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Cc: Keith Packard <keithp@keithp.com>
Reviewed-by: NAlex Deucher <alexdeucher@gmail.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

copy_blit operation works only on integral number of pages
so benchmarks shorter than one page size (4K) do not make sense

v2: use RADEON_GPU_PAGE_SIZE instead of "magic" 1024 number and
    sweep sizes between 1 * <page_size> to 16K * <page_size> doubling
    the size in each iteration; we get the same coverage, as
    in the original benchmark, but guarantee integer multiples
    of page size

v3: add whitespace between '*' operator per review received from
    zajec5@gmail.com
Signed-off-by: NIlija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

R600/700 and Evergreen/NI blit code have a few redundant
definitions in respective .c file. Move common definitions
into a separate (new) .h file.
Signed-off-by: NIlija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Its useful to be able to call the mode setting getter ioctls.
Not requiring master fd, enables writing a simple program which
can query the state of the video system.

Since these ioctls are only "getters" there is no security or
synchronization issues which would require master fd. Opening
an new fd is already protected by the file permissions on the
device file.
Signed-off-by: NMandeep Singh Baines <msb@chromium.org>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ilija Hadzic <ihadzic@research.bell-labs.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Stephane Marchesin <marcheu@chromium.org>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

The rbd_client structure uses a kref to arrange for cleaning up and
freeing an instance when its last reference is dropped.  The cleanup
routine is rbd_client_release(), and one of the things it does is
delete the rbd_client from rbd_client_list.  It acquires node_lock
to do so, but the way it is done is still not safe.

The problem is that when attempting to reuse an existing rbd_client,
the structure found might already be in the process of getting
destroyed and cleaned up.

Here's the scenario, with "CLIENT" representing an existing
rbd_client that's involved in the race:

 Thread on CPU A                | Thread on CPU B
 ---------------                | ---------------
 rbd_put_client(CLIENT)         | rbd_get_client()
   kref_put()                   |   (acquires node_lock)
     kref->refcount becomes 0   |   __rbd_client_find() returns CLIENT
     calls rbd_client_release() |   kref_get(&CLIENT->kref);
                                |   (releases node_lock)
       (acquires node_lock)     |
       deletes CLIENT from list | ...and starts using CLIENT...
       (releases node_lock)     |
       and frees CLIENT         | <-- but CLIENT gets freed here

Fix this by having rbd_put_client() acquire node_lock.  The result
could still be improved, but at least it avoids this problem.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

If an existing rbd client is found to be suitable for use in
rbd_get_client(), the rbd_options structure is not being
freed as it should.  Fix that.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>