Now that the mmu_notifier_invalidate_range() calls are in place, add the
callback to allow subsystems to register against it.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Jérôme Glisse <jglisse@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: Jay Cornwall <Jay.Cornwall@amd.com>
Cc: Oded Gabbay <Oded.Gabbay@amd.com>
Cc: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>
Add calls to the new mmu_notifier_invalidate_range() function to all
places in the VMM that need it.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Jérôme Glisse <jglisse@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: Jay Cornwall <Jay.Cornwall@amd.com>
Cc: Oded Gabbay <Oded.Gabbay@amd.com>
Cc: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>
This notifier closes an important gap in the current mmu_notifier
implementation, the existing callbacks are called too early or too late to
reliably manage a non-CPU TLB. Specifically, invalidate_range_start() is
called when all pages are still mapped and invalidate_range_end() when all
pages are unmapped and potentially freed.
This is fine when the users of the mmu_notifiers manage their own SoftTLB,
like KVM does. When the TLB is managed in software it is easy to wipe out
entries for a given range and prevent new entries to be established until
invalidate_range_end is called.
But when the user of mmu_notifiers has to manage a hardware TLB it can
still wipe out TLB entries in invalidate_range_start, but it can't make
sure that no new TLB entries in the given range are established between
invalidate_range_start and invalidate_range_end.
To avoid silent data corruption the entries in the non-CPU TLB need to be
flushed when the pages are unmapped (at this point in time no _new_ TLB
entries can be established in the non-CPU TLB) but not yet freed (as the
non-CPU TLB may still have _existing_ entries pointing to the pages about
to be freed).
To fix this problem we need to catch the moment when the Linux VMM flushes
remote TLBs (as a non-CPU TLB is not very CPU TLB), as this is the point
in time when the pages are unmapped but _not_ yet freed.
The mmu_notifier_invalidate_range() function aims to catch that moment.
IOMMU code will be one user of the notifier-callback. Currently this is
only the AMD IOMMUv2 driver, but its code is about to be more generalized
and converted to a generic IOMMU-API extension to fit the needs of similar
functionality in other IOMMUs as well.
The current attempt in the AMD IOMMUv2 driver to work around the
invalidate_range_start/end() shortcoming is to assign an empty page table
to the non-CPU TLB between any invalidata_range_start/end calls. With the
empty page-table assigned, every page-table walk to re-fill the non-CPU
TLB will cause a page-fault reported to the IOMMU driver via an interrupt,
possibly causing interrupt storms.
The page-fault handler in the AMD IOMMUv2 driver doesn't handle the fault
if an invalidate_range_start/end pair is active, it just reports back
SUCCESS to the device and let it refault the page. But existing hardware
(newer Radeon GPUs) that makes use of this feature don't re-fault
indefinitly, after a certain number of faults for the same address the
device enters a failure state and needs to be resetted.
To avoid the GPUs entering a failure state we need to get rid of the
empty-page-table workaround and use the mmu_notifier_invalidate_range()
function introduced with this patch.
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Jérôme Glisse <jglisse@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: Jay Cornwall <Jay.Cornwall@amd.com>
Cc: Oded Gabbay <Oded.Gabbay@amd.com>
Cc: Suravee Suthikulpanit <Suravee.Suthikulpanit@amd.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Oded Gabbay <oded.gabbay@amd.com>
Turned out to be much simpler on top of my latest atomic stuff than
what I've feared. Some details:
- Drop the modeset_lock_all snakeoil in drm_plane_init. Same
justification as for the equivalent change in drm_crtc_init done in
commit d0fa1af40e
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Mon Sep 8 09:02:49 2014 +0200
drm: Drop modeset locking from crtc init function
Without these the drm_modeset_lock_init would fall over the exact
same way.
- Since the atomic core code wraps the locking switching it to
per-plane locks was a one-line change.
- For the legacy ioctls add a plane argument to the locking helper so
that we can grab the right plane lock (cursor or primary). Since the
universal cursor plane might not be there, or someone really crazy
might forgoe the primary plane even accept NULL.
- Add some locking WARN_ON to the atomic helpers for good paranoid
measure and to check that it all works out.
Tested on my exynos atomic hackfest with full lockdep checks and ww
backoff injection.
v2: I've forgotten about the load-detect code in i915.
v3: Thierry reported that in latest 3.18-rc vmwgfx doesn't compile any
more due to
commit 21e88620aa
Author: Rob Clark <robdclark@gmail.com>
Date: Thu Oct 30 13:39:04 2014 -0400
drm/vmwgfx: fix lock breakage
Rebased and fix this up.
Cc: Thierry Reding <thierry.reding@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
v1: original
v2: danvet's kerneldoc nitpicks
Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJUX/DqAAoJEHm+PkMAQRiGLtQH/iAt3fRHlYDXjaJian/KG1Cb
wVP0I+HWZmvVmmd0PzyaxCZLgRNwdmmYHEH4QLy2JwZ3jZfFHlxhy+hDWCgz+67t
bIzkLs0Pf1T4kJ2+r8qW2kBEz9PWJHGTQw7NTqZ++Ts3rPptBA6Fg4mEJ6fQigXy
qRIY68DpipUkXV9BWBWijnTmrvP5tt7JtPzBr4DC8frMjvWct8+XwYhc2k2tEv2j
LwLYb1OW6PUpPv2BQBfWjqqH77vYNQVhJwuwGcDe2YZdI0UFkDheL24+RbbPcZ4f
OnrLjJSSgzv6lBWkAaXZK7/WJ/JZbXxEqHzWZQ3xXoQov97bm7lEYJqqi5gDasQ=
=6Qpa
-----END PGP SIGNATURE-----
Merge tag 'v3.18-rc4' into drm-next
backmerge to get vmwgfx locking changes into next as the
conflict with per-plane locking.
So here's my atomic series, finally all debugged&reviewed. Sean Paul has
done a full detailed pass over it all, and a lot of other people have
commented and provided feedback on some parts. Rob Clark also converted
msm over the w/e and seems happy. The only small thing is that Rob wants
to export the wait_for_vblank, which imo makes sense. Since there's other
stuff still to do I think we should apply Rob's patch (once it has grown
appropriate kerneldoc) later on top of this.
This is just the core<->driver interface plus a big pile of helpers. Short
recap of the main ideas:
- There are essentially three helper libraries in this patch set:
* Transitional helpers to use the new plane callbacks for legacy plane
updates and in the crtc helper's ->mode_set callback. These helpers are
only temporarily used to convert drivers to atomic, but they allow a
nice separation between changing the driver backend and switching to
the atomic commit logic.
* Legacy helpers to implement all the legacy driver entry points
(page_flip, set_config, plane vfuncs) on top of the new atomic driver
interface. These are completely driver agnostic. The reason for having
the legacy support as helpers is that drivers can switch step-by-step.
And they could e.g. even keep the legacy page_flip code around for some
old platforms where converting to full-blown atomic isn't worth it.
* Atomic helpers which implement the various new ->atomic_* driver
interfaces in terms of the revised crtc helper and new plane helper
hooks.
- The revised crtc helper implemenation essentially implements all the
lessons learned in the i915 modeset rework (when using the atomic helpers
only):
* Enable/disable sequence for a given config are always the same and
callbacks are always called in the same order. This contrast starkly
with the crtc helpers, where the sequence of operations is heavily
dependent on the previous config.
One corollary of this is that if the configuration of a crtc only
partially changes (e.g. a connector moves in a cloned config) the
helper code will still disable/enable the full display pipeline. This
is the only way to ensure that the enable/disable sequence is always
the same.
* It won't call disable or enable hooks more than once any more because
it lost track of state, thanks to the atomic state tracking. And if
drivers implement the ->reset hook properly (by either resetting the hw
or reading out the hw state into the atomic structures) this even
extends to the hardware state. So no more disable-me-harder kind of
nonsense.
* The only thing missing is the hw state readout/cross-check support, but
if drivers have hw state readout support in their ->reset handlers it's
simple to extend that to cross-check the hw state.
* The crtc->mode_set callback is gone and its replacement only sets crtc
timings and no longer updates the primary plane state. This way we can
finally implement primary planes properly.
- The new plane helpers should be suitable enough for pretty much
everything, and a perfect fit for hardware with GO bits. Even if they
don't fit the atomic helper library is rather flexible and exports all
the functions for the individual steps to drivers. So drivers can pick
what matches and implement their own magic for everything else.
- A big difference compared to all previous atomic series is that this one
doesn't implement async commit in a generic way. Imo driver requirements
for that are too diverse to create anything reasonable sane which would
actually work on a reasonable amount of different drivers. Also, we've
never had a helper library for page_flips even, so it's really hard to
know what might work and what's stupid without a bit of experience in the form
of a few driver implementations.
I think with the current flexibility for drivers to pick individual
stages and existing helpers like drm_flip_queue it's rather easy though
to implement proper async commit.
- There's a few other differences of minor importance to earlier atomic
series:
* Common/generic properties are parsed in the callers/core and not in
drivers, and passed to drivers by directly setting the right members in
atomic state structures. That greatly simplifies all the transitional
and legacy helpers an removes a lot of boilerplate code.
* There's no crazy trylock mode used for the async commit since these
helpers don't do async commit. A simple ordered flip queue of atomic
state updates should be sufficient for preventing concurrent hw access
anyway, as long as synchronous updates stall correctly with e.g.
flush_work_queue or similar function. Abusing locks to enforce ordering
isn't a good idea imo anyway.
* These helpers reuse the existing ->mode_fixup hooks in the atomic_check
callback. Which means that drivers need to adapat and move a lot less code
into their atomic_check callbacks.
Now this isn't everything needed in the drm core and helpers for full
atomic support. But it's enough to start with converting drivers, and
except for actually testing multiplane and multicrtc updates also enough to
implement full atomic updates. Still missing are:
- Per-plane locking. Since these helpers here encapsulate the locking
completely this should be fairly easy to implement.
- fbdev support for atomic_check/commit, so that multi-pipe finally works
sanely in fbcon.
- Adding and decoding shared/core properties. That just needs to be rebased
from Rob's latest patch series, with minor adjustments so that the
decoding happens in the core instead of in drivers.
- Actually adding the atomic ioctl. Again just rebasing Rob's latest patch
should be all that's needed.
- Resolving how to deal with DPMS in atomic. Atomic is a good excuse to fix up
the crazy semantics dpms currently has. I'm floating an RFC about this topic
already.
- Finally I couldn't test connector/encoder stealing properly since my test
vehicle here doesn't allow a connector on different crtcs. So drivers
which support this might see some surprises in that area. There is no semantic
change though in how encoder stealing and assignment works (or at least no
intended one), so I think the risk is minimal.
As just mentioned I've done a fake conversion of an existing driver using
crtc helpers to debug the helper code and validate the smooth transition
approach. And that smooth transition was the really big motivation for
this. It seems to actually work and consists of 3 phases:
Phase 1: Rework driver backend for crtc/plane helpers
The requirement here is that universal plane support is already implement. If
universal plane support isn't implement yet it might be better though to just do
it as part of this phase, directly using the new plane helpers. There are two
big things to do:
- Split up the existing ->update/disable_plane hooks into check/commit
hooks and extract the crtc-wide prep/flush parts (like setting/clearing
GO bits).
- The other big change is to split the crtc->mode_set hook into the plane
update (done using the plane helpers) and the crtc setup in a new
->mode_set_nofb hook.
When phase 1 is complete the driver implements all the new callbacks which
push the software state into hardware, but still using all the legacy entry
points and crtc helpers. The transitional helpers serve as impendance
mismatch here.
Phase 2: Rework state handling
This consists of rolling out the state handling helpers for planes, crtcs
and connectors and reviewing all ->mode_fixup and similar hooks to make
sure they don't depend upon implicit global state which might change in the
atomic world. Any such code must be moved into ->atomic_check functions which
just rely on the free-standing atomic state update structures.
This phase also adds a few small pieces of fixup code to make sure the
atomic state doesn't get out of sync in the legacy driver callbacks.
Phase 3: Roll out atomic support
Now it's just about replacing vfuncs with the ones provided by the helper
and filling out the small missing pieces (like atomic_check logic or async
commit support needed for page_flips). Due to the prep work in phase 1 no
changes to the driver backend functions should be required, and because of
the prep work in phase 2 atomic implementations can be rolled out
step-by-step. So if async commit ins't implemented yet page_flip can be
implemented with the legacy functions without wreaking havoc in the other
operations.
* tag 'topic/atomic-helpers-2014-11-09' of git://anongit.freedesktop.org/drm-intel:
drm/atomic: Refcounting for plane_state->fb
drm: Docbook integration and over sections for all the new helpers
drm/atomic-helpers: functions for state duplicate/destroy/reset
drm/atomic-helper: implement ->page_flip
drm/atomic-helpers: document how to implement async commit
drm/atomic: Integrate fence support
drm/atomic-helper: implementatations for legacy interfaces
drm: Atomic crtc/connector updates using crtc/plane helper interfaces
drm/crtc-helper: Transitional functions using atomic plane helpers
drm/plane-helper: transitional atomic plane helpers
drm: Add atomic/plane helpers
drm: Global atomic state handling
drm: Add atomic driver interface definitions for objects
drm/modeset_lock: document trylock_only in kerneldoc
drm: fixup kerneldoc in drm_crtc.h
drm: Pull drm_crtc.h into the kerneldoc template
drm: Move drm_crtc_init from drm_crtc.h to drm_plane_helper.h
Another quiet week:
- A fix to silence edma probe error on non-supported platforms from Arnd
- A fix to enable the PL clock for Parallella, to make mainline usable with
the SDK.
- A somewhat verbose fix for the PLL clock tree on VF610
- Enabling of SD/MMC on one of the VF610-based boards (for testing)
- A fix for i.MX where CONFIG_SPI used to be implicitly enabled and now needs
to be added to the defconfig instead
- Another maintainer added for bcm2835: Lee Jones
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=bOlr
-----END PGP SIGNATURE-----
Merge tag 'armsoc-for-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull ARM SoC fixes from Olof Johansson:
"Another quiet week:
- a fix to silence edma probe error on non-supported platforms from
Arnd
- a fix to enable the PL clock for Parallella, to make mainline
usable with the SDK.
- a somewhat verbose fix for the PLL clock tree on VF610
- enabling of SD/MMC on one of the VF610-based boards (for testing)
- a fix for i.MX where CONFIG_SPI used to be implicitly enabled and
now needs to be added to the defconfig instead
- another maintainer added for bcm2835: Lee Jones"
* tag 'armsoc-for-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
ARM: dts: zynq: Enable PL clocks for Parallella
dma: edma: move device registration to platform code
ARM: dts: vf610: add SD node to cosmic dts
MAINTAINERS: update bcm2835 entry
ARM: imx: Fix the removal of CONFIG_SPI option
ARM: imx: clk-vf610: define PLL's clock tree
Pull devicetree bugfix from Grant Likely:
"One buffer overflow bug that shouldn't be left around"
* 'devicetree/merge' of git://git.kernel.org/pub/scm/linux/kernel/git/glikely/linux:
of: Fix overflow bug in string property parsing functions
Just various stuff all over from a bunch of people. Shortlog gives a beter
overview, it's really all misc drm patches.
* tag 'topic/core-stuff-2014-11-05' of git://anongit.freedesktop.org/drm-intel:
drm/edid: add #defines and helpers for ELD
drm/dp: Add counters in the drm_dp_aux struct for I2C NACKs and DEFERs
drm: Remove compiler BUG_ON() test
drm: Fix DRM_FORCE_ON_DIGITAL use
drm/gma500: Don't destroy DRM properties in the driver
drm/i915: Don't destroy DRM properties in the driver
drm: Add a note to drm_property_create() about property lifetime
gpu: drm: Fix warning caused by a parameter description in drm_crtc.c
drm/dp-helper: Move the legacy helpers to gma500
drm/crtc: Remove duplicated ioctl code
drm/crtc: Fix two typos
gpu:drm: Fix typo in Documentation/DocBook/drm.xml
gpu: drm: drm_dp_mst_topology.c: Fix improper use of strncat
drm: drm_err: Remove unnecessary __func__ argument
drm: Implement O_NONBLOCK support on /dev/dri/cardN
So my original plan was that the drm core refcounts framebuffers like
with the legacy ioctls. But that doesn't work for a bunch of reasons:
- State objects might live longer than until the next fb change
happens for a plane. For example delayed cleanup work only happens
_after_ the pageflip ioctl has completed. So this definitely doesn't
work without the plane state holding its own references.
- The other issue is transition from legacy to atomic implementations,
where the driver works under a mix of both worlds. Which means
legacy paths might not properly update the ->fb pointer under
plane->state->fb. Which is a bit a problem when then someone comes
around and _does_ try to clean it up when it's long gone.
The second issue is just a bit a transition bug, since drivers should
update plane->state->fb in all the paths that aren't converted yet.
But a bit more robustness for the transition can't hurt - we pull
similar tricks with cleaning up the old fb in the transitional helpers
already.
The pattern for drivers that transition is
if (plane->state)
drm_atomic_set_fb_for_plane(plane->state, plane->fb);
inserted after the fb update has logically completed at the end of
->set_config (or ->set_base/mode_set if using the crtc helpers),
->page_flip, ->update_plane or any other entry point which updates
plane->fb.
v2: Update kerneldoc - copypasta fail.
v3: Fix spelling in the commit message (Sean).
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
The atomic users and helpers assume that there is always a obj->state
structure around. Which means drivers need to somehow create that at
driver load time. Also it should obviously reset hardware state, so
needs to be reset upon resume.
Finally the destroy/duplicate_state functions are an awful lot of
boilerplate if the driver doesn't need anything beyond the default
state objects.
So add helper functions for all of this.
v2: Somehow the plane/connector versions got lost in the first
version.
v3: Add kerneldoc.
v4: Make duplicate_state functions a bit more robust, which is useful
for debugging state tracking issues when transitioning to atomic.
v5: Clear temporary variables in the crtc state when duplicating it,
like ->mode_changed or ->planes_changed. If we don't do this stale
values for these might pollute the next atomic modeset.
v6: Also clear crtc_state->event in case the driver didn't (yet) clear
this out.
v7: Split out wrong squashed commit. Also improve the kerneldoc to
mention that obj->state can be NULL and when. Both suggested by
Daniel Thompson.
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently there is no way to implement async flips using atomic, that
essentially requires us to be able to cancel pending requests
mid-flight.
To be able to do that (and I guess we want this since vblank synced
updates which opportunistically cancel still pending updates seem to be
wanted) we'd need to add a mandatory cancellation mode. Depending upon
the exact semantics we decide upon that could mean that userspace will
not get completion events, or will get them all stacked up.
So reject async updates for now. Also async updates usually means not
vblank synced at all, and I guess for drivers which want to support
this they should simply add a special pageflip handler (since usually
you need a special flip cmd to achieve this). That kind of async flip
is pretty much exclusively just used for games and benchmarks where
dropping just one frame means you'll get a headshot or something bad
like that ... And so slight amounts of tearing is acceptable.
v2: Fixup kerneldoc, reported by Paulo.
v3: Use the set_crtc_for_plane function to assign the crtc, since
otherwise the book-keeping is off.
v4: Update crtc->primary->fb since ->page_flip is the only driver
callback where the core won't do this itself. We might want to fix
this inconsistency eventually.
v5: Use set_crtc_for_connector as suggested by Sean.
v6: Daniel Thompson noticed that my error handling is inconsistent
and that in a few cases I didn't handle fatal errors (i.e. not
-EDEADLK). Fix this by consolidate the ww mutex backoff handling
into one check in the fail: block and flatten the error control
flow everywhere else.
v7: Fix spelling mistake in the commit message (Sean).
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This patch is for enabling async commits. It replaces an earlier
approach which added an async boolean paramter to the ->prepare_fb
callbacks. The idea is that prepare_fb picks up the right fence to
synchronize against, which is then used by the synchronous commit
helper. For async commits drivers can either register a callback to
the fence or simply do the synchronous wait in their async work queue.
v2: Remove unused variable.
v3: Only wait for fences after the point of no return in the part
of the commit function which can be run asynchronously. This is after
the atomic state has been swapped in, hence now check
plane->state->fence.
Also add a WARN_ON to make sure we don't try to wait on a fence when
there's no fb, just as a sanity check.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Well, except page_flip since that requires async commit, which isn't
there yet.
For the functions which changes planes there's a bit of trickery
involved to keep the fb refcounting working. But otherwise fairly
straight-forward atomic updates.
The property setting functions are still a bit incomplete. Once we
have generic properties (e.g. rotation, but also all the properties
needed by the atomic ioctl) we need to filter those out and parse them
in the helper. Preferrably with the same function as used by the real
atomic ioctl implementation.
v2: Fixup kerneldoc, reported by Paulo.
v3: Add missing EXPORT_SYMBOL.
v4: We need to look at the crtc of the modeset, not some random
leftover one from a previous loop when udpating the connector->crtc
routing. Also push some local variables into inner loops to avoid
these kinds of bugs.
v5: Adjust semantics - drivers now own the atomic state upon
successfully synchronous commit.
v6: Use the set_crtc_for_plane function to assign the crtc, since
otherwise the book-keeping is off.
v7:
- Improve comments.
- Filter out the crtc of the ->set_config call when recomputing
crtc_state->enabled: We should compute the same state, but not doing
so will give us a good chance to catch bugs and inconsistencies -
the atomic helper's atomic_check function re-validates this again.
- Fix the set_config implementation logic when disabling the crtc: We
still need to update the output routing to disable all the
connectors properly in the state. Caught by the atomic_check
functions, so at least that part worked ;-) Also add some WARN_ONs
to ensure ->set_config preconditions all apply.
v8: Fixup an embarrassing h/vdisplay mixup.
v9: Shuffled bad squash to the right patch, spotted by Daniel
v10: Use set_crtc_for_connector as suggested by Sean.
v11: Daniel Thompson noticed that my error handling is inconsistent
and that in a few cases I didn't handle fatal errors (i.e. not
-EDEADLK). Fix this by consolidate the ww mutex backoff handling
into one check in the fail: block and flatten the error control
flow everywhere else.
v12: Review and discussion with Sean:
- One spelling fix.
- Correctly skip the crtc from the set_config set when recomputing
->enable state. That should allow us to catch any bugs in higher
levels in computing that state (which is supplied to the
->set_config implementation). I've screwed this up and Sean spotted
that the current code is pointless.
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
So this is finally the integration of the crtc and plane helper
interfaces into the atomic helper functions.
In the check function we now have a few steps:
- First we update the output routing and figure out which crtcs need a
full mode set. Suitable encoders are selected using ->best_encoder,
with the same semantics as the crtc helpers of implicitly disabling
all connectors currently using the encoder.
- Then we pull all other connectors into the state update which feed
from a crtc which changes. This must be done do catch mode changes
and similar updates - atomic updates are differences on top of the
current state.
- Then we call all the various ->mode_fixup to compute the adjusted
mode. Note that here we have a slight semantic difference compared
to the crtc helpers: We have not yet updated the encoder->crtc link
when calling the encoder's ->mode_fixup function. But that's a
requirement when converting to atomic since we want to prepare the
entire state completely contained with the over drm_atomic_state
structure. So this must be carefully checked when converting drivers
over to atomic helpers.
- Finally we do call the atomic_check functions on planes and crtcs.
The commit function is also quite a beast:
- The only step that can fail is done first, namely pinning the
framebuffers. After that we cross the point of no return, an async
commit would push all that into the worker thread.
- The disabling of encoders and connectors is a bit tricky, since
depending upon the final state we need to select different crtc
helper functions.
- Software tracking is a bit clarified compared to the crtc helpers:
We commit the software state before starting to touch the hardware,
like crtc helpers. But since we just swap them we still have the old
state (i.e. the current hw state) around, which is really handy to
write simple disable functions. So no more
drm_crtc_helper_disable_all_unused_functions kind of fun because
we're leaving unused crtcs/encoders behind. Everything gets shut
down in-order now, which is one of the key differences of the i915
helpers compared to crtc helpers and a really nice additional
guarantee.
- Like with the plane helpers the atomic commit function waits for one
vblank to pass before calling the framebuffer cleanup function.
Compared to Rob's helper approach there's a bunch of upsides:
- All the interfaces which can fail are called in the ->check hook
(i.e. ->best_match and the various ->mode_fixup hooks). This means
that drivers can just reuse those functions and don't need to move
everything into ->atomic_check callbacks. If drivers have no need
for additional constraint checking beyong their existing crtc
helper callbacks they don't need to do anything.
- The actual commit operation is properly stage: First we prepare
framebuffers, which can potentially still fail (due to memory
exhausting). This is important for the async case, where this must
be done synchronously to correctly return errors.
- The output configuration changes (done with crtc helper functions)
and the plane update (using atomic plane helpers) are correctly
interleaved: First we shut down any crtcs that need changing, then
we update planes and finally we enable everything again. Hardware
without GO bits must be more careful with ordering, which this
sequence enables.
- Also for hardware with shared output resources (like display PLLs)
we first must shut down the old configuration before we can enable
the new one. Otherwise we can hit an impossible intermediate state
where there's not enough PLLs (which is the point behind atomic
updates).
v2:
- Ensure that users of ->check update crtc_state->enable correctly.
- Update the legacy state in crtc/plane structures. Eventually we want
to remove that, but for now the drm core still expects this (especially
the plane->fb pointer).
v3: A few changes for better async handling:
- Reorder the software side state commit so that it happens all before
we touch the hardware. This way async support becomes very easy
since we can punt all the actual hw touching to a worker thread. And
as long as we synchronize with that thread (flushing or cancelling,
depending upon what the driver can handle) before we commit the next
software state there's no need for any locking in the worker thread
at all. Which greatly simplifies things.
And as long as we synchronize with all relevant threads we can have
a lot of them (e.g. per-crtc for per-crtc updates) running in
parallel.
- Expose pre/post plane commit steps separately. We need to expose the
actual hw commit step anyway for drivers to be able to implement
asynchronous commit workers. But if we expose pre/post and plane
commit steps individually we allow drivers to selectively use atomic
helpers.
- I've forgotten to call encoder/bridge ->mode_set functions, fix
this.
v4: Add debug output and fix a mixup between current and new state
that resulted in crtcs not getting updated correctly. And in an
Oops ...
v5:
- Be kind to driver writers in the vblank wait functions.. if thing
aren't working yet, and vblank irq will never come, then let's not
block forever.. especially under console-lock.
- Correctly clear connector_state->best_encoder when disabling.
Spotted while trying to understand a report from Rob Clark.
- Only steal encoder if it actually changed, otherwise hilarity ensues
if we steal from the current connector and so set the ->crtc pointer
unexpectedly to NULL. Reported by Rob Clark.
- Bail out in disable_outputs if an output currently doesn't have a
best_encoder - this means it's already disabled.
v6: Fixupe kerneldoc as reported by Paulo. And also fix up kerneldoc
in drm_crtc.h.
v7: Take ownership of the atomic state and clean it up with
drm_atomic_state_free().
v8 Various improvements all over:
- Polish code comments and kerneldoc.
- Improve debug output to make sure all failure cases are logged.
- Treat enabled crtc with no connectors as invalid input from userspace.
- Don't ignore the return value from mode_fixup().
v9:
- Improve debug output for crtc_state->mode_changed.
v10:
- Fixup the vblank waiting code to properly balance the vblank_get/put
calls.
- Better comments when checking/computing crtc->mode_changed
v11: Fixup the encoder stealing logic: We can't look at encoder->crtc
since that's not in the atomic state structures and might be updated
asynchronously in and async commit. Instead we need to inspect all the
connector states and check whether the encoder is currently in used
and if so, on which crtc.
v12: Review from Sean:
- A few spelling fixes.
- Flatten control flow indent by converting if blocks to early
continue/return in 2 places.
- Capture connectors_for_crtc return value in int num_connectors
instead of bool has_connectors and do an explicit int->bool
conversion with !!. I think the helper is more useful for drivers if
it returns the number of connectors (e.g. to detect cloning
configurations), so decided to keep that return value.
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Paulo Zanoni <przanoni@gmail.com>
Cc: Rob Clark <robdclark@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Enumeration
- Don't oops on virtual buses in acpi_pci_get_bridge_handle() (Yinghai Lu)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=n1pR
-----END PGP SIGNATURE-----
Merge tag 'pci-v3.18-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Pull PCI fix from Bjorn Helgaas:
"This fixes an oops when enabling SR-IOV VF devices. The oops is a
regression I added by configuring all devices during enumeration.
- Don't oops on virtual buses in acpi_pci_get_bridge_handle() (Yinghai Lu)"
* tag 'pci-v3.18-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
PCI: Don't oops on virtual buses in acpi_pci_get_bridge_handle()
acpi_pci_get_bridge_handle() returns the ACPI handle for the bridge device
(either a host bridge or a PCI-to-PCI bridge) leading to a PCI bus. But
SR-IOV virtual functions can be on a virtual bus with no bridge leading to
it. Return a NULL acpi_handle in this case instead of trying to
dereference the NULL pointer to the bridge.
This fixes a NULL pointer dereference oops in pci_get_hp_params() when
adding SR-IOV VF devices on virtual buses.
[bhelgaas: changelog, add comment in code]
Fixes: 6cd33649fa ("PCI: Add pci_configure_device() during enumeration")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=87591
Reported-by: Chao Zhou <chao.zhou@intel.com>
Reported-by: Joerg Roedel <joro@8bytes.org>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
These two functions allow drivers to reuse their atomic plane helpers
functions for the primary plane to implement the interfaces required
by the crtc helpers for the legacy ->set_config callback.
This is purely transitional and won't be used once the driver is fully
converted. But it allows partial conversions to the atomic plane
helpers which are functional.
v2:
- Use ->atomic_duplicate_state if available.
- Don't forget to run crtc_funcs->atomic_check.
v3: Shift source coordinates correctly for 16.16 fixed point.
v4: Don't forget to call ->atomic_destroy_state if available.
v5: Fixup kerneldoc.
v6: Reuse the plane_commit function from the transitional plane
helpers to avoid too much duplication.
v7:
- Remove some stale comment.
- Correctly handle the lack of plane->state object, necessary for
transitional use.
v8: Fixup an embarrassing h/vdisplay mixup.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Converting a driver to the atomic interface can be a daunting
undertaking. One of the prerequisites is to have full universal planes
support.
To make that transition a bit easier this patch provides plane helpers
which use the new atomic helper callbacks just only for the plane
changes. This way the plane update functionality can be tested without
being forced to convert everything at once.
Of course a real atomic update capable driver will implement the
all plane properties through the atomic interface, so these helpers
are mostly transitional. But they can be used to enable proper
universal plane support, especially once the crtc helpers have also
been adapted.
v2: Use ->atomic_duplicate_state if available.
v3: Don't forget to call ->atomic_destroy_state if available.
v4: Fixup kerneldoc, reported by Paulo.
v5: Extract a common plane_commit helper and fix some bugs in the
plane_state setup of the plane_disable implementation.
v6: Fix issues with the cleanup of the old fb. Since transitional
helpers can be mixed we need to assume that the old fb has been set up
by a legacy path (e.g. set_config or page_flip when the primary plane
is converted to use these functions already). Hence pass an additional
old_fb parameter to plane_commit to do that cleanup work correctly.
v7:
- Fix spurious WARNING (crtc helpers really love to disable stuff
harder) and fix array index bonghits.
- Correctly handle the lack of plane->state object, necessary for
transitional use.
- Don't indicate failure if drm_vblank_get doesn't work - that's
expected when the pipe is in dpms off mode.
v8: Review from Sean:
- s/fail/out/ to make the meaning of a label more clear.
- spelling fix in the commit message.
Cc: Paulo Zanoni <przanoni@gmail.com>
Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This is the first cut of atomic helper code. As-is it's only useful to
implement a pure atomic interface for plane updates.
Later patches will integrate this with the crtc helpers so that full
atomic updates are possible. We also need a pile of helpers to aid
drivers in transitioning from the legacy world to the shiny new atomic
age. Finally we need helpers to implement legacy ioctls on top of the
atomic interface.
The design of the overall helpers<->driver interaction is fairly
simple, but has an unfortunate large interface:
- We have ->atomic_check callbacks for crtcs and planes. The idea is
that connectors don't need any checking, and if they do they can
adjust the relevant crtc driver-private state. So no connector hooks
should be needed. Also the crtc helpers integration will do the
->best_encoder checks, so no need for that.
- Framebuffer pinning needs to be done before we can commit to the hw
state. This is especially important for async updates where we must
pin all buffers before returning to userspace, so that really only
hw failures can happen in the asynchronous worker.
Hence we add ->prepare_fb and ->cleanup_fb hooks for this resources
management.
- The actual atomic plane commit can't fail (except hw woes), so has
void return type. It has three stages:
1. Prepare all affected crtcs with crtc->atomic_begin. Drivers can
use this to unset the GO bit or similar latches to prevent plane
updates.
2. Update plane state by looping over all changed planes and calling
plane->atomic_update. Presuming the hardware is sane and has GO
bits drivers can simply bash the state into the hardware in this
function. Other drivers might use this to precompute hw state for
the final step.
3. Finally latch the update for the next vblank with
crtc->atomic_flush. Note that this function doesn't need to wait
for the vblank to happen even for the synchronous case.
v2: Clear drm_<obj>_state->state to NULL when swapping in state.
v3: Add TODO that we don't short-circuit plane updates for now. Likely
no one will care.
v4: Squash in a bit of polish that somehow landed in the wrong (later)
patche.
v5: Integrate atomic functions into the drm docbook and fixup the
kerneldoc.
v6: Fixup fixup patch squashing fumble.
v7: Don't touch the legacy plane state plane->fb and plane->crtc. This
is only used by the legacy ioctl code in the drm core, and that code
already takes care of updating the pointers in all relevant cases.
This is in stark contrast to connector->encoder->crtc links on the
modeset side, which we still need to set since the core doesn't touch
them.
Also some more kerneldoc polish.
v8: Drop outdated comment.
v9: Handle the state->state pointer correctly: Only clearing the
->state pointer when assigning the state to the kms object isn't good
enough. We also need to re-link the swapped out state into the
drm_atomic_state structure.
v10: Shuffle the misplaced docbook template hunk around that Sean spotted.
Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Some differences compared to Rob's patches again:
- Dropped the committed and checked booleans. Checking will be
internally enforced by always calling ->atomic_check before
->atomic_commit. And async handling needs to be solved differently
because the current scheme completely side-steps ww mutex deadlock
avoidance (and so either reinvents a new deadlock avoidance wheel or
like the current code just deadlocks).
- State for connectors needed to be added, since now they have a
full-blown drm_connector_state (so that drivers have something to
attach their own stuff to).
- Refcounting is gone. I plane to solve async updates differently,
since the lock-passing scheme doesn't cut it (since it abuses ww
mutexes). Essentially what we need for async is a simple ownership
transfer from the caller to the driver. That doesn't need full-blown
refcounting.
- The acquire ctx is a pointer. Real atomic callers should have that
on their stack, legacy entry points need to put the right one
(obtained by drm_modeset_legacy_acuire_ctx) in there.
- I've dropped all hooks except check/commit. All the begin/end
handling is done by core functions and is the same.
- commit/check are just thin wrappers that ensure that ->check is
always called.
- To help out with locking in the legacy implementations I've added a
helper to just grab all locks in the backoff case.
v2: Add notices that check/commit can fail with EDEADLK.
v3:
- More consistent naming for state_alloc.
- Add state_clear which is needed for backoff and retry.
v4: Planes/connectors can switch between crtcs, and we need to be
careful that we grab the state (and locks) for both the old and new
crtc. Improve the interface functions to ensure this.
v5: Add functions to grab affected connectors for a crtc and to recompute
the crtc->enable state. This is useful for both helper and atomic ioctl
code when e.g. removing a connector.
v6: Squash in fixup from Fengguang to use ERR_CAST.
v7: Add debug output.
v8: Make checkpatch happy about kcalloc argument ordering.
v9: Improve kerneldoc in drm_crtc.h
v10:
- Fix another kcalloc argument misorder I've missed.
- More polish for kerneldoc.
v11: Clarify the ownership rules for the state object. The new rule is
that a successful drm_atomic_commit (whether synchronous or asnyc)
always inherits the state and is responsible for the clean-up. That
way async and sync ->commit functions are more similar.
v12: A few bugfixes:
- Assign state->state pointers correctly when grabbing state objects -
we need to link them up with the global state.
- Handle a NULL crtc in set_crtc_for_plane to simplify code flow a bit
for the callers of this function.
v13: Review from Sean:
- kerneldoc spelling fixes
- Don't overallocate states->planes.
- Handle NULL crtc in set_crtc_for_connector.
v14: Sprinkle __must_check over all functions which do wait/wound
locking to make sure callers don't forget this. Since I have ;-)
v15: Be more explicit in the kerneldoc when functions can return
-EDEADLK what to do. And that every other -errno is fatal.
v16: Indent with tabs instead of space, spotted by Ander.
v17: Review from Thierry, small kerneldoc and other naming polish.
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: Daniel Thompson <daniel.thompson@linaro.org>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Heavily based upon Rob Clark's atomic series.
- Dropped the connector state from the crtc state, instead opting for a
full-blown connector state. The only thing it has is the desired
crtc, but drivers which have connector properties have now a
data-structure to subclass.
- Rename create_state to duplicate_state. Especially for legacy ioctls
we want updates on top of existing state, so we need a way to get at
the current state. We need to be careful to clear the backpointers
to the global state correctly though.
- Drop property values. Drivers with properties simply need to
subclass the datastructures and track the decoded values in there. I
also think that common properties (like rotation) should be decoded
and stored in the core structures.
- Create a new set of ->atomic_set_prop functions, for smoother
transitions from legacy to atomic operations.
- Pass the ->atomic_set_prop ioctl the right structure to avoid
chasing pointers in drivers.
- Drop temporary boolean state for now until we resurrect them with
the helper functions.
- Drop invert_dimensions. For now we don't need any checking since
that's done by the higher-level legacy ioctls. But even then we
should also add rotation/flip tracking to the core drm_crtc_state,
not just whether the dimensions are inverted.
- Track crtc state with an enable/disable. That's equivalent to
mode_valid, but a bit clearer that it means the entire crtc.
The global interface will follow in subsequent patches.
v2: We need to allow drivers to somehow set up the initial state and
clear it on resume. So add a plane->reset callback for that. Helpers
will be provided with default behaviour for all these.
v3: Split out the plane->reset into a separate patch.
v4: Improve kerneldoc in drm_crtc.h
v5: Remove unused inline functions for handling state objects, those
callbacks are now mandatory for full atomic support.
v6: Fix commit message nit Sean noticed.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I've forgotten to do this in:
commit cb597bb3a2
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Sun Jul 27 19:09:33 2014 +0200
drm: trylock modest locking for fbdev panics
Oops, fix this asap.
In my defense kerneldoc is really awful and there's no way it can pick
up structured comments per struct member. Which means we need both
since people won't scroll up even a few lines.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
In the interest of reducing magic numbers and having to cross check with
the specs all the time.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
These counters are used for Displayort compliance testing to detect error
conditions when executing tests 4.2.2.4 and 4.2.2.5 in the Displayport Link
CTS specificaiton. They determine whether to use the preferred/requested
mode or the failsafe mode during these tests.
V2:
- Addressed previous review feedback
- Updated commit message
- Changed from uint8_t to uint32_t
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Todd Previte <tprevite@gmail.com>
[danvet: s/uint32_t/unsigned/ for clearer intent. Also drop the i915
from the subject, it's all core stuff.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I've tried to cc all the people who have recently added new stuff
but forgotten to update documentation.
I've also decided not to bother documenting the massive property list
in struct drm_mode_config. If that beast keeps on growing we might want
to extract it into a separate structure which we won't document.
Cc: Thomas Wood <thomas.wood@intel.com>
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
While writing atomic docs I've noticed that I don't get any errors
for my screw-ups in drm_crtc.h. Fix this immediately.
This just does the bare minimum to get starts, lots of stuff isn't
properly documented yet unfortunately.
v2: Fix adjacent spelling error Sean noticed.
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Cc: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Just a bit of OCD cleanup on headers - this function isn't the core
interface any more but just a helper for drivers who haven't yet
transitioned to universal planes. Put the declaration at the right
spot and sprinkle necessary #includes over all drivers.
Maybe this helps to encourage driver maintainers to do the switch.
v2: Fix #include ordering for tegra, reported by 0-day builder.
v3: Include required headers, reported by Thierry.
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Thierry Reding <treding@nvidia.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
The string property read helpers will run off the end of the buffer if
it is handed a malformed string property. Rework the parsers to make
sure that doesn't happen. At the same time add new test cases to make
sure the functions behave themselves.
The original implementations of of_property_read_string_index() and
of_property_count_strings() both open-coded the same block of parsing
code, each with it's own subtly different bugs. The fix here merges
functions into a single helper and makes the original functions static
inline wrappers around the helper.
One non-bugfix aspect of this patch is the addition of a new wrapper,
of_property_read_string_array(). The new wrapper is needed by the
device_properties feature that Rafael is working on and planning to
merge for v3.19. The implementation is identical both with and without
the new static inline wrapper, so it just got left in to reduce the
churn on the header file.
Signed-off-by: Grant Likely <grant.likely@linaro.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Darren Hart <darren.hart@intel.com>
Cc: <stable@vger.kernel.org> # v3.3+: Drop selftest hunks that don't apply
So far, the required PLL's (PLL1/PLL2/PLL5) have been initialized
by boot loader and the kernel code defined fixed rates according
to those default configurations. Beginning with the USB PLL7 the
code started to initialize the PLL's itself (using imx_clk_pllv3).
However, since commit dc4805c2e7
(ARM: imx: remove ENABLE and BYPASS bits from clk-pllv3 driver)
imx_clk_pllv3 no longer takes care of the ENABLE and BYPASS bits,
hence the USB PLL were not configured correctly anymore.
This patch not only fixes those USB PLL's, but also makes use of
the imx_clk_pllv3 for all PLL's and alignes the code with the PLL
support of the i.MX6 series.
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Pull CMA and DMA-mapping fixes from Marek Szyprowski:
"This contains important fixes for recently introduced highmem support
for default contiguous memory region used for dma-mapping subsystem"
* 'fixes-for-v3.18' of git://git.linaro.org/people/mszyprowski/linux-dma-mapping:
mm, cma: make parameters order consistent in func declaration and definition
mm: cma: Use %pa to print physical addresses
mm: cma: Ensure that reservations never cross the low/high mem boundary
mm: cma: Always consider a 0 base address reservation as dynamic
mm: cma: Don't crash on allocation if CMA area can't be activated
* A regression from 3.16 which was noticed in 3.17. With the restructuring of
the m25p80.c driver and the SPI NOR library framework, we omitted proper
listing of the SPI device IDs. This means m25p80.c wouldn't auto-load
(modprobe) properly when built as a module. For now, we duplicate the device
IDs into both modules.
* The OMAP / ELM modules were depending on an implicit link ordering. Use
deferred probing so that the new link order (in 3.18-rc) can still allow for
successful probing.
* Fix suspend/resume support for LH28F640BF NOR flash
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQIcBAABAgAGBQJUVq20AAoJEFySrpd9RFgt8XQQAI4oIygz4zGQ6n0y4HOqwOBy
F4ZPtOuzuCYA86x2zORFgj4A9JGVjDQwTfnMQnn1NG+XEEmZMJfG2IwqlUxsZd5A
KkAS5XUoi/Fvq95Qi95KQYXqm1dniXEGKsRFsHKXIsnnDmbqRK5fBn6Ve5PAwcau
uru5FwrZ2Ve0EwF9/Z/bxAatRirdAhwgMGlaXdXLmL7S13NQGmXP9QI7CbxSZ38R
GJ+A6PhiYs6Sml6Ou5bovNXyFGfx4J35pk6nTWoWe5MfZHRQk447OQwBPbsrM119
Boq8F/6diXyJfuXdxvF6JiDmDzaw/fBY+Xuq1O6p+JzLONN16x93KlpAPhzy4a15
PwFHCBzg5khY49if/dmrPJ+kLkU+9wIHUib8m6HSImCKBT5Bv/VJXoQ1g4s1IJ8/
Di3mz/8pWm/cscABIkuEqb9TwwUrSHzVXgGH/p4CY0eUo8DbQQA1zDsig8aRIX36
FlAReaHH8QivdnghkMX9Px7SIo7XoMZZEi+55k8FrIVqjqEHNGx+w+BKhxgtFggN
nAg0l7NrLdQHpigK1SZjLFGIYi7MmarvbatUjVPGagiRqoQ0mCSS7eKX1DEs4EAo
P2g64BSJickGAhUiAV9ZO1EBoaOU6olIPpc33J+uG/8qBU1cNClx3FJ1UPWX27JQ
+FBsD1mec4FuoZ2SoE7r
=IxjM
-----END PGP SIGNATURE-----
Merge tag 'for-linus-20141102' of git://git.infradead.org/linux-mtd
Pull MTD fixes from Brian Norris:
"Three main MTD fixes for 3.18:
- A regression from 3.16 which was noticed in 3.17. With the
restructuring of the m25p80.c driver and the SPI NOR library
framework, we omitted proper listing of the SPI device IDs. This
means m25p80.c wouldn't auto-load (modprobe) properly when built as
a module. For now, we duplicate the device IDs into both modules.
- The OMAP / ELM modules were depending on an implicit link ordering.
Use deferred probing so that the new link order (in 3.18-rc) can
still allow for successful probing.
- Fix suspend/resume support for LH28F640BF NOR flash"
* tag 'for-linus-20141102' of git://git.infradead.org/linux-mtd:
mtd: cfi_cmdset_0001.c: fix resume for LH28F640BF chips
mtd: omap: fix mtd devices not showing up
mtd: m25p80,spi-nor: Fix module aliases for m25p80
mtd: spi-nor: make spi_nor_scan() take a chip type name, not spi_device_id
mtd: m25p80: get rid of spi_get_device_id
This is a set of six patches consisting of two MAINTAINER updates, two scsi-mq
fixs for the old parallel interface (not every request is tagged and we need
to set the right flags to populate the SPI tag message) and a fix for a memory
leak in scatterlist traversal caused by a preallocation update in 3.17) and an
ipv6 fix for cxgbi.
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
iQEcBAABAgAGBQJUVqYyAAoJEDeqqVYsXL0MDwQIAI4aCdC3LUelt6IzywToiJ9u
GTUZU7eULzV1KuaExcBM8g9DyE3wtnhfS0WA5TqvfKf8ImMKntan0loK0juIjj8K
UBtbSDapwkKRRAIoRHvBzcHU+Jzc//naO95ojsFt2ELHB91jta368+DiPb8m+wQx
IKeYO6vFaSwZZCKVNaWwyCpxEKmMO+ib80zfuANBzWjau3IFxMiUofSvtPGIWb+E
Zxi9qvOyqjfFE8OLfL9ckH5NPiIYcwmarAXyuYcJstiw6VX2deGBaJalueSSC0JV
TAqEtT5k1AraFS2HM0TRxar+HK5iBj3s8f0pFBfgD2+oagUln6L/486nMh2Ktac=
=l6wS
-----END PGP SIGNATURE-----
Merge tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"This is a set of six patches consisting of:
- two MAINTAINER updates
- two scsi-mq fixs for the old parallel interface (not every request
is tagged and we need to set the right flags to populate the SPI
tag message)
- a fix for a memory leak in scatterlist traversal caused by a
preallocation update in 3.17
- an ipv6 fix for cxgbi"
[ The scatterlist fix also came in separately through the block layer tree ]
* tag 'scsi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
MAINTAINERS: ufs - remove self
MAINTAINERS: change hpsa and cciss maintainer
libcxgbi : support ipv6 address host_param
scsi: set REQ_QUEUE for the blk-mq case
Revert "block: all blk-mq requests are tagged"
lib/scatterlist: fix memory leak with scsi-mq
Pull drm fixes from Dave Airlie:
"Nothing too astounding or major: radeon, i915, vmwgfx, armada and
exynos.
Biggest ones:
- vmwgfx has one big locking regression fix
- i915 has come displayport fixes
- radeon has some stability and a memory alloc failure
- armada and exynos have some vblank fixes"
* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (24 commits)
drm/exynos: correct connector->dpms field before resuming
drm/exynos: enable vblank after DPMS on
drm/exynos: init kms poll at the end of initialization
drm/exynos: propagate plane initialization errors
drm/exynos: vidi: fix build warning
drm/exynos: remove explicit encoder/connector de-initialization
drm/exynos: init vblank with real number of crtcs
drm/vmwgfx: Filter out modes those cannot be supported by the current VRAM size.
drm/vmwgfx: Fix hash key computation
drm/vmwgfx: fix lock breakage
drm/i915/dp: only use training pattern 3 on platforms that support it
drm/radeon: remove some buggy dead code
drm/i915: Ignore VBT backlight check on Macbook 2, 1
drm/radeon: remove invalid pci id
drm/radeon: dpm fixes for asrock systems
radeon: clean up coding style differences in radeon_get_bios()
drm/radeon: Use drm_malloc_ab instead of kmalloc_array
drm/radeon/dpm: disable ulv support on SI
drm/i915: Fix GMBUSFREQ on vlv/chv
drm/i915: Ignore long hpds on eDP ports
...
Pull VFS fixes from Al Viro:
"A bunch of assorted fixes, most of them followups to overlayfs merge"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
ovl: initialize ->is_cursor
Return short read or 0 at end of a raw device, not EIO
isofs: don't bother with ->d_op for normal case
isofs_cmp(): we'll never see a dentry for . or ..
overlayfs: fix lockdep misannotation
ovl: fix check for cursor
overlayfs: barriers for opening upper-layer directory
rcu: Provide counterpart to rcu_dereference() for non-RCU situations
staging: android: logger: Fix log corruption regression
Pull networking fixes from David Miller:
"A bit has accumulated, but it's been a week or so since my last batch
of post-merge-window fixes, so...
1) Missing module license in netfilter reject module, from Pablo.
Lots of people ran into this.
2) Off by one in mac80211 baserate calculation, from Karl Beldan.
3) Fix incorrect return value from ax88179_178a driver's set_mac_addr
op, which broke use of it with bonding. From Ian Morgan.
4) Checking of skb_gso_segment()'s return value was not all
encompassing, it can return an SKB pointer, a pointer error, or
NULL. Fix from Florian Westphal.
This is crummy, and longer term will be fixed to just return error
pointers or a real SKB.
6) Encapsulation offloads not being handled by
skb_gso_transport_seglen(). From Florian Westphal.
7) Fix deadlock in TIPC stack, from Ying Xue.
8) Fix performance regression from using rhashtable for netlink
sockets. The problem was the synchronize_net() invoked for every
socket destroy. From Thomas Graf.
9) Fix bug in eBPF verifier, and remove the strong dependency of BPF
on NET. From Alexei Starovoitov.
10) In qdisc_create(), use the correct interface to allocate
->cpu_bstats, otherwise the u64_stats_sync member isn't
initialized properly. From Sabrina Dubroca.
11) Off by one in ip_set_nfnl_get_byindex(), from Dan Carpenter.
12) nf_tables_newchain() was erroneously expecting error pointers from
netdev_alloc_pcpu_stats(). It only returna a valid pointer or
NULL. From Sabrina Dubroca.
13) Fix use-after-free in _decode_session6(), from Li RongQing.
14) When we set the TX flow hash on a socket, we mistakenly do so
before we've nailed down the final source port. Move the setting
deeper to fix this. From Sathya Perla.
15) NAPI budget accounting in amd-xgbe driver was counting descriptors
instead of full packets, fix from Thomas Lendacky.
16) Fix total_data_buflen calculation in hyperv driver, from Haiyang
Zhang.
17) Fix bcma driver build with OF_ADDRESS disabled, from Hauke
Mehrtens.
18) Fix mis-use of per-cpu memory in TCP md5 code. The problem is
that something that ends up being vmalloc memory can't be passed
to the crypto hash routines via scatter-gather lists. From Eric
Dumazet.
19) Fix regression in promiscuous mode enabling in cdc-ether, from
Olivier Blin.
20) Bucket eviction and frag entry killing can race with eachother,
causing an unlink of the object from the wrong list. Fix from
Nikolay Aleksandrov.
21) Missing initialization of spinlock in cxgb4 driver, from Anish
Bhatt.
22) Do not cache ipv4 routing failures, otherwise if the sysctl for
forwarding is subsequently enabled this won't be seen. From
Nicolas Cavallari"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (131 commits)
drivers: net: cpsw: Support ALLMULTI and fix IFF_PROMISC in switch mode
drivers: net: cpsw: Fix broken loop condition in switch mode
net: ethtool: Return -EOPNOTSUPP if user space tries to read EEPROM with lengh 0
stmmac: pci: set default of the filter bins
net: smc91x: Fix gpios for device tree based booting
mpls: Allow mpls_gso to be built as module
mpls: Fix mpls_gso handler.
r8152: stop submitting intr for -EPROTO
netfilter: nft_reject_bridge: restrict reject to prerouting and input
netfilter: nft_reject_bridge: don't use IP stack to reject traffic
netfilter: nf_reject_ipv6: split nf_send_reset6() in smaller functions
netfilter: nf_reject_ipv4: split nf_send_reset() in smaller functions
netfilter: nf_tables_bridge: update hook_mask to allow {pre,post}routing
drivers/net: macvtap and tun depend on INET
drivers/net, ipv6: Select IPv6 fragment idents for virtio UFO packets
drivers/net: Disable UFO through virtio
net: skb_fclone_busy() needs to detect orphaned skb
gre: Use inner mac length when computing tunnel length
mlx4: Avoid leaking steering rules on flow creation error flow
net/mlx4_en: Don't attempt to TX offload the outer UDP checksum for VXLAN
...
Pull scheduler fixes from Ingo Molnar:
"Various scheduler fixes all over the place: three SCHED_DL fixes,
three sched/numa fixes, two generic race fixes and a comment fix"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/dl: Fix preemption checks
sched: Update comments for CLONE_NEWNS
sched: stop the unbound recursion in preempt_schedule_context()
sched/fair: Fix division by zero sysctl_numa_balancing_scan_size
sched/fair: Care divide error in update_task_scan_period()
sched/numa: Fix unsafe get_task_struct() in task_numa_assign()
sched/deadline: Fix races between rt_mutex_setprio() and dl_task_timer()
sched/deadline: Don't replenish from a !SCHED_DEADLINE entity
sched: Fix race between task_group and sched_task_group
Pull perf fixes from Ingo Molnar:
"Mostly tooling fixes, plus on the kernel side:
- a revert for a newly introduced PMU driver which isn't complete yet
and where we ran out of time with fixes (to be tried again in
v3.19) - this makes up for a large chunk of the diffstat.
- compilation warning fixes
- a printk message fix
- event_idx usage fixes/cleanups"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf probe: Trivial typo fix for --demangle
perf tools: Fix report -F dso_from for data without branch info
perf tools: Fix report -F dso_to for data without branch info
perf tools: Fix report -F symbol_from for data without branch info
perf tools: Fix report -F symbol_to for data without branch info
perf tools: Fix report -F mispredict for data without branch info
perf tools: Fix report -F in_tx for data without branch info
perf tools: Fix report -F abort for data without branch info
perf tools: Make CPUINFO_PROC an array to support different kernel versions
perf callchain: Use global caching provided by libunwind
perf/x86/intel: Revert incomplete and undocumented Broadwell client support
perf/x86: Fix compile warnings for intel_uncore
perf: Fix typos in sample code in the perf_event.h header
perf: Fix and clean up initialization of pmu::event_idx
perf: Fix bogus kernel printk
perf diff: Add missing hists__init() call at tool start
Pull core fixes from Ingo Molnar:
"The tree contains two RCU fixes and a compiler quirk comment fix"
* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
rcu: Make rcu_barrier() understand about missing rcuo kthreads
compiler/gcc4+: Remove inaccurate comment about 'asm goto' miscompiles
rcu: More on deadlock between CPU hotplug and expedited grace periods
Pablo Neira Ayuso says:
====================
netfilter/ipvs fixes for net
The following patchset contains fixes for netfilter/ipvs. This round of
fixes is larger than usual at this stage, specifically because of the
nf_tables bridge reject fixes that I would like to see in 3.18. The
patches are:
1) Fix a null-pointer dereference that may occur when logging
errors. This problem was introduced by 4a4739d56b ("ipvs: Pull
out crosses_local_route_boundary logic") in v3.17-rc5.
2) Update hook mask in nft_reject_bridge so we can also filter out
packets from there. This fixes 36d2af5 ("netfilter: nf_tables: allow
to filter from prerouting and postrouting"), which needs this chunk
to work.
3) Two patches to refactor common code to forge the IPv4 and IPv6
reject packets from the bridge. These are required by the nf_tables
reject bridge fix.
4) Fix nft_reject_bridge by avoiding the use of the IP stack to reject
packets from the bridge. The idea is to forge the reject packets and
inject them to the original port via br_deliver() which is now
exported for that purpose.
5) Restrict nft_reject_bridge to bridge prerouting and input hooks.
the original skbuff may cloned after prerouting when the bridge stack
needs to flood it to several bridge ports, it is too late to reject
the traffic.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
That can be reused by the reject bridge expression to build the reject
packet. The new functions are:
* nf_reject_ip6_tcphdr_get(): to sanitize and to obtain the TCP header.
* nf_reject_ip6hdr_put(): to build the IPv6 header.
* nf_reject_ip6_tcphdr_put(): to build the TCP header.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
That can be reused by the reject bridge expression to build the reject
packet. The new functions are:
* nf_reject_ip_tcphdr_get(): to sanitize and to obtain the TCP header.
* nf_reject_iphdr_put(): to build the IPv4 header.
* nf_reject_ip_tcphdr_put(): to build the TCP header.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Author: David Jeffery <djeffery@redhat.com>
Changes to the basic direct I/O code have broken the raw driver when reading
to the end of a raw device. Instead of returning a short read for a read that
extends partially beyond the device's end or 0 when at the end of the device,
these reads now return EIO.
The raw driver needs the same end of device handling as was added for normal
block devices. Using blkdev_read_iter, which has the needed size checks,
prevents the EIO conditions at the end of the device.
Signed-off-by: David Jeffery <djeffery@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
UFO is now disabled on all drivers that work with virtio net headers,
but userland may try to send UFO/IPv6 packets anyway. Instead of
sending with ID=0, we should select identifiers on their behalf (as we
used to).
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Fixes: 916e4cf46d ("ipv6: reuse ip6_frag_id from ip6_ufo_append_data")
Signed-off-by: David S. Miller <davem@davemloft.net>
Some drivers are unable to perform TX completions in a bound time.
They instead call skb_orphan()
Problem is skb_fclone_busy() has to detect this case, otherwise
we block TCP retransmits and can freeze unlucky tcp sessions on
mostly idle hosts.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: 1f3279ae0c ("tcp: avoid retransmits of TCP packets hanging in host queues")
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull two RCU fixes from Paul E. McKenney:
" - Complete the work of commit dd56af42bd (rcu: Eliminate deadlock
between CPU hotplug and expedited grace periods), which was
intended to allow synchronize_sched_expedited() to be safely
used when holding locks acquired by CPU-hotplug notifiers.
This commit makes the put_online_cpus() avoid the deadlock
instead of just handling the get_online_cpus().
- Complete the work of commit 35ce7f29a4 (rcu: Create rcuo
kthreads only for onlined CPUs), which was intended to allow
RCU to avoid allocating unneeded kthreads on systems where the
firmware says that there are more CPUs than are really present.
This commit makes rcu_barrier() aware of the mismatch, so that
it doesn't hang waiting for non-existent CPUs. "
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Merge misc fixes from Andrew Morton:
"21 fixes"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (21 commits)
mm/balloon_compaction: fix deflation when compaction is disabled
sh: fix sh770x SCIF memory regions
zram: avoid NULL pointer access in concurrent situation
mm/slab_common: don't check for duplicate cache names
ocfs2: fix d_splice_alias() return code checking
mm: rmap: split out page_remove_file_rmap()
mm: memcontrol: fix missed end-writeback page accounting
mm: page-writeback: inline account_page_dirtied() into single caller
lib/bitmap.c: fix undefined shift in __bitmap_shift_{left|right}()
drivers/rtc/rtc-bq32k.c: fix register value
memory-hotplug: clear pgdat which is allocated by bootmem in try_offline_node()
drivers/rtc/rtc-s3c.c: fix initialization failure without rtc source clock
kernel/kmod: fix use-after-free of the sub_info structure
drivers/rtc/rtc-pm8xxx.c: rework to support pm8941 rtc
mm, thp: fix collapsing of hugepages on madvise
drivers: of: add return value to of_reserved_mem_device_init()
mm: free compound page with correct order
gcov: add ARM64 to GCOV_PROFILE_ALL
fsnotify: next_i is freed during fsnotify_unmount_inodes.
mm/compaction.c: avoid premature range skip in isolate_migratepages_range
...
Commit 0a31bc97c8 ("mm: memcontrol: rewrite uncharge API") changed
page migration to uncharge the old page right away. The page is locked,
unmapped, truncated, and off the LRU, but it could race with writeback
ending, which then doesn't unaccount the page properly:
test_clear_page_writeback() migration
wait_on_page_writeback()
TestClearPageWriteback()
mem_cgroup_migrate()
clear PCG_USED
mem_cgroup_update_page_stat()
if (PageCgroupUsed(pc))
decrease memcg pages under writeback
release pc->mem_cgroup->move_lock
The per-page statistics interface is heavily optimized to avoid a
function call and a lookup_page_cgroup() in the file unmap fast path,
which means it doesn't verify whether a page is still charged before
clearing PageWriteback() and it has to do it in the stat update later.
Rework it so that it looks up the page's memcg once at the beginning of
the transaction and then uses it throughout. The charge will be
verified before clearing PageWriteback() and migration can't uncharge
the page as long as that is still set. The RCU lock will protect the
memcg past uncharge.
As far as losing the optimization goes, the following test results are
from a microbenchmark that maps, faults, and unmaps a 4GB sparse file
three times in a nested fashion, so that there are two negative passes
that don't account but still go through the new transaction overhead.
There is no actual difference:
old: 33.195102545 seconds time elapsed ( +- 0.01% )
new: 33.199231369 seconds time elapsed ( +- 0.03% )
The time spent in page_remove_rmap()'s callees still adds up to the
same, but the time spent in the function itself seems reduced:
# Children Self Command Shared Object Symbol
old: 0.12% 0.11% filemapstress [kernel.kallsyms] [k] page_remove_rmap
new: 0.12% 0.08% filemapstress [kernel.kallsyms] [k] page_remove_rmap
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Cc: <stable@vger.kernel.org> [3.17.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A follow-up patch would have changed the call signature. To save the
trouble, just fold it instead.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Cc: <stable@vger.kernel.org> [3.17.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Joerg Roedel