Add secure-boot for the dGPU set of GM20X chips, using the PMU as the
high-secure falcon.

This work is based on Deepak Goyal's initial port of Secure Boot to
Nouveau.

v2. use proper memory target function
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Secure falcons' firmware is managed by secboot. Do not load it in GR for
them.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Start securely-managed falcons using secboot functions since the process
for them is different from just writing CPUCTL.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

On GM200 and later GPUs, firmware for some essential falcons (notably
GR ones) must be authenticated by a NVIDIA-produced signature and
loaded by a high-secure falcon in order to be able to access privileged
registers, in a process known as Secure Boot.

Secure Boot requires building a binary blob containing the firmwares
and signatures of the falcons to be loaded. This blob is then given to
a high-secure falcon running a signed loader firmware that copies the
blob into a write-protected region, checks that the signatures are
valid, and finally loads the verified firmware into the managed falcons
and switches them to privileged mode.

This patch adds infrastructure code to support this process on chips
that require it.

v2:
- The IRQ mask of the PMU falcon was left - replace it with the proper
  irq_mask variable.
- The falcon reset procedure expecting a falcon in an initialized state,
  which was accidentally provided by the PMU subdev. Make sure that
  secboot can manage the falcon on its own.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Load firmware and bundles in GM200's constructor. The previously called
GF100 function did not care about the bundles.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

There functions are going to be used by other chips that rely on
NVIDIA-provided firmware. Export them.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Make these functions easier to use by handling memory management from
within.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

The firmwares required by GR may vary from chip to chip, especially with
the introduction of secure boot and NVIDIA-provided firmwares. Move the
firmware loading outside of gf100_gr_ctor so other chips may still call
it while managing their firmwares themselves.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Some members of gf100_gr were freed by the gk20a driver. That's not
where it should be done - free them in gf100 so other chips that use
NVIDIA-provided firmware free these structures properly.

This also removes the need for a GK20A-specific destructor.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Add memcpy functions to copy a buffer to a gpuobj and vice-versa. This
will be used by the secure boot code.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Most of the per-chipset differences will go away when we fully switch
to using the register lists provided by the firmware files, which will
leave all the remaining code "belonging" to GM200.

This is a preemptive rename from GM204 to GM200.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Upon encountering an unknown condition code, the script interpreter
is supposed to skip 'size' bytes and continue at the next devinit
token.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

It is not advisable to perform devinit if it has already been done.
VBIOS will very likely have invoked devinit if the GPU is the primary
graphics device, but there is no accurate way to detect this fact yet.

This patch adds such a method for gf100 and later chips, by means of the
NV_PTOP_SCRATCH1_DEVINIT_COMPLETED bit. This bit is set to 1 by devinit,
and reset to 0 when the GPU is powered.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

We never use any nv50-specific member in this nv50_devinit_preinit().
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

nvkm_device_tegra_new initializes the irq member of the Tegra device
to -1 in order to signal that it is uninitialized. However,
nvkm_device_tegra_fini tests it against 0 to check whether an IRQ has
been allocated or not. This leads to free_irq being called on -1 during
device initialization.

Fix this by using 0 as the uninitialized value everywhere.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

nvkm_device_fini is never called if a failure occurs in
nvkm_device_init, even when unloading the module. This can lead to a
resources leak (one example is the Tegra interrupt which would never be
freed in that case). Fix this by calling nvkm_device_fini in
nvkm_device_init's failure path.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Use the nvkm_firmware_* functions when loading external firmware to
avoid duplicate code.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Add two functions nvkm_firmware_get() and nvkm_firmware_put() to load a
firmware file and free its resources, respectively. Since firmware files
are becoming a necessity for new GPUs, and their location has been
standardized to nvidia/chip/, this will prevent duplicate and
error-prone name-generation code.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Patch "ltc/gm107: use nvkm_mask to set cbc_ctrl1" sets the 3rd bit
of the CTRL1 register instead of writing it entirely in
gm107_ltc_cbc_clear(). As a counterpart, gm107_ltc_cbc_wait() must also
be modified to wait on that single bit only, otherwise a timeout may
occur if some other bit of that register is set. This happened at least
on GM206 when running glmark2-drm.

While we are at it, use the more compact nvkm_wait_msec() to wait for
the bit to clear.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

This pull request fixes the major VC4 HDMI modesetting bugs found when
the first wave of users showed up in Raspbian.

* tag 'drm-vc4-fixes-2016-03-03' of github.com:anholt/linux:
  drm/vc4: Initialize scaler DISPBKGND on modeset.
  drm/vc4: Fix setting of vertical timings in the CRTC.
  drm/vc4: Fix the name of the VSYNCD_EVEN register.
  drm/vc4: Add another reg to HDMI debug dumping.
  drm/vc4: Bring HDMI up from power off if necessary.
  drm/vc4: Fix a framebuffer reference leak on async flip interrupt.

Merge branch 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos into drm-next

* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: add DRM_EXYNOS_GEM_MAP ioctl

Nouveau wanted this to avoid some worse conflicts when I merge that.

The commit d931589c ("drm/exynos: remove DRM_EXYNOS_GEM_MAP_OFFSET
ioctl") removed it same with the ioctl that this patch adds. The reason
that removed DRM_EXYNOS_GEM_MAP_OFFSET was we could use
DRM_IOCTL_MODE_MAP_DUMB. Both did exactly same thing.

Now we again will revive it as DRM_EXYNOS_GEM_MAP because of render
node. DRM_IOCTL_MODE_MAP_DUMB isn't permitted in render node.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

Two i915 regression fixes.

* tag 'drm-intel-fixes-2016-03-11' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Actually retry with bit-banging after GMBUS timeout
  drm/i915: Fix bogus dig_port_map[] assignment for pre-HSW

Pull ARM SoC fixes from Olof Johansson:
 "Two more fixes for 4.5:

   - One is a fix for OMAP that is urgently needed to avoid DRA7xx chips
     from premature aging, by always keeping the Ethernet clock enabled.

   - The other solves a I/O memory layout issue on Armada, where SROM
     and PCI memory windows were conflicting in some configurations"

* tag 'armsoc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: mvebu: fix overlap of Crypto SRAM with PCIe memory window
  ARM: dts: dra7: do not gate cpsw clock due to errata i877
  ARM: OMAP2+: hwmod: Introduce ti,no-idle dt property

Pull media fix from Mauro Carvalho Chehab:
 "One last time fix: It adds a code that prevents some media tools like
  media-ctl to hide some entities that have their IDs out of the range
  expected by those apps"

* tag 'media/v4.5-5' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] media-device: map new functions into old types for legacy API

When the Crypto SRAM mappings were added to the Device Tree files
describing the Armada XP boards in commit c466d997 ("ARM: mvebu:
define crypto SRAM ranges for all armada-xp boards"), the fact that
those mappings were overlaping with the PCIe memory aperture was
overlooked. Due to this, we currently have for all Armada XP platforms
a situation that looks like this:

Memory mapping on Armada XP boards with internal registers at
0xf1000000:

 - 0x00000000 -> 0xf0000000	3.75G 	RAM
 - 0xf0000000 -> 0xf1000000	16M	NOR flashes (AXP GP / AXP DB)
 - 0xf1000000 -> 0xf1100000	1M	internal registers
 - 0xf8000000 -> 0xffe0000	126M	PCIe memory aperture
 - 0xf8100000 -> 0xf8110000	64KB	Crypto SRAM #0	=> OVERLAPS WITH PCIE !
 - 0xf8110000 -> 0xf8120000	64KB	Crypto SRAM #1	=> OVERLAPS WITH PCIE !
 - 0xffe00000 -> 0xfff00000	1M	PCIe I/O aperture
 - 0xfff0000  -> 0xffffffff	1M	BootROM

The overlap means that when PCIe devices are added, depending on their
memory window needs, they might or might not be mapped into the
physical address space. Indeed, they will not be mapped if the area
allocated in the PCIe memory aperture by the PCI core overlaps with
one of the Crypto SRAM. Typically, a Intel IGB PCIe NIC that needs 8MB
of PCIe memory will see its PCIe memory window allocated from
0xf80000000 for 8MB, which overlaps with the Crypto SRAM windows. Due
to this, the PCIe window is not created, and any attempt to access the
PCIe window makes the kernel explode:

[    3.302213] igb: Copyright (c) 2007-2014 Intel Corporation.
[    3.307841] pci 0000:00:09.0: enabling device (0140 -> 0143)
[    3.313539] mvebu_mbus: cannot add window '4:f8', conflicts with another window
[    3.320870] mvebu-pcie soc:pcie-controller: Could not create MBus window at [mem 0xf8000000-0xf87fffff]: -22
[    3.330811] Unhandled fault: external abort on non-linefetch (0x1008) at 0xf08c0018

This problem does not occur on Armada 370 boards, because we use the
following memory mapping (for boards that have internal registers at
0xf1000000):

 - 0x00000000 -> 0xf0000000	3.75G 	RAM
 - 0xf0000000 -> 0xf1000000	16M	NOR flashes (AXP GP / AXP DB)
 - 0xf1000000 -> 0xf1100000	1M	internal registers
 - 0xf1100000 -> 0xf1110000	64KB	Crypto SRAM #0 => OK !
 - 0xf8000000 -> 0xffe0000	126M	PCIe memory
 - 0xffe00000 -> 0xfff00000	1M	PCIe I/O
 - 0xfff0000  -> 0xffffffff	1M	BootROM

Obviously, the solution is to align the location of the Crypto SRAM
mappings of Armada XP to be similar with the ones on Armada 370, i.e
have them between the "internal registers" area and the beginning of
the PCIe aperture.

However, we have a special case with the OpenBlocks AX3-4 platform,
which has a 128 MB NOR flash. Currently, this NOR flash is mapped from
0xf0000000 to 0xf8000000. This is possible because on OpenBlocks
AX3-4, the internal registers are not at 0xf1000000. And this explains
why the Crypto SRAM mappings were not configured at the same place on
Armada XP.

Hence, the solution is two-fold:

 (1) Move the NOR flash mapping on Armada XP OpenBlocks AX3-4 from
     0xe8000000 to 0xf0000000. This frees the 0xf0000000 ->
     0xf80000000 space.

 (2) Move the Crypto SRAM mappings on Armada XP to be similar to
     Armada 370 (except of course that Armada XP has two Crypto SRAM
     and not one).

After this patch, the memory mapping on Armada XP boards with
registers at 0xf1 is:

 - 0x00000000 -> 0xf0000000	3.75G 	RAM
 - 0xf0000000 -> 0xf1000000	16M	NOR flashes (AXP GP / AXP DB)
 - 0xf1000000 -> 0xf1100000	1M	internal registers
 - 0xf1100000 -> 0xf1110000	64KB	Crypto SRAM #0
 - 0xf1110000 -> 0xf1120000	64KB	Crypto SRAM #1
 - 0xf8000000 -> 0xffe0000	126M	PCIe memory
 - 0xffe00000 -> 0xfff00000	1M	PCIe I/O
 - 0xfff0000  -> 0xffffffff	1M	BootROM

And the memory mapping for the special case of the OpenBlocks AX3-4
(internal registers at 0xd0000000, NOR of 128 MB):

 - 0x00000000 -> 0xc0000000	3G 	RAM
 - 0xd0000000 -> 0xd1000000	1M	internal registers
 - 0xe800000  -> 0xf0000000	128M	NOR flash
 - 0xf1100000 -> 0xf1110000	64KB	Crypto SRAM #0
 - 0xf1110000 -> 0xf1120000	64KB	Crypto SRAM #1
 - 0xf8000000 -> 0xffe0000	126M	PCIe memory
 - 0xffe00000 -> 0xfff00000	1M	PCIe I/O
 - 0xfff0000  -> 0xffffffff	1M	BootROM

Fixes: c466d997 ("ARM: mvebu: define crypto SRAM ranges for all armada-xp boards")
Reported-by: Phil Sutter <phil@nwl.cc>
Cc: Phil Sutter <phil@nwl.cc>
Cc: <stable@vger.kernel.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Olof Johansson <olof@lixom.net>

Pull dmaengine fixes from Vinod Koul:
 "Two fixes showed up in last few days, and they should be included in
  4.5.  Summary:

  Two more late fixes to drivers, nothing major here:

   - A memory leak fix in fsdma unmap the dma descriptors on freeup

   - A fix in xdmac driver for residue calculation of dma descriptor"

* tag 'dmaengine-fix-4.5' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: at_xdmac: fix residue computation
  dmaengine: fsldma: fix memory leak

Pull power management and ACPI fixes from Rafael Wysocki:
 "Two more fixes for issues introduced recently, one in the generic
  device properties framework and one in ACPICA.

  Specifics:

   - Revert a recent ACPICA commit that has been reverted upstream,
     because it caused problems to happen on user systems and the
     problem it attempted to address will not be relevant any more after
     upcoming ACPI specification changes (Bob Moore).

   - Fix crash in the generic device properties framework introduced by
     a recent change that forgot to check pointers against error values
     in addition to checking them against NULL (Heikki Krogerus)"

* tag 'pm+acpi-4.5-final' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  device property: fwnode->secondary may contain ERR_PTR(-ENODEV)
  ACPICA: Revert "Parser: Fix for SuperName method invocation"

Pull xfs fixes from Dave Chinner:
 "This is a fix for a regression introduced in 4.5-rc1 by the new torn
  log write detection code.  The regression only affects people moving a
  clean filesystem between machines/kernels of different architecture
  (such as changing between 32 bit and 64 bit kernels), but this is the
  recommended (and only!) safe way to migrate a filesystem between
  architectures so we really need to ensure it works.

  The changes are larger than I'd prefer right at the end of the release
  cycle, but the majority of the change is just factoring code to enable
  the detection of a clean log at the correct time to avoid this issue.

  Changes:

   - Only perform torn log write detection on dirty logs.  This prevents
     failures being detected due to a clean filesystem being moved
     between machines or kernels of different architectures (e.g.  32 ->
     64 bit, BE -> LE, etc).  This fixes a regression introduced by the
     torn log write detection in 4.5-rc1"

* tag 'xfs-for-linus-4.5-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs:
  xfs: only run torn log write detection on dirty logs
  xfs: refactor in-core log state update to helper
  xfs: refactor unmount record detection into helper
  xfs: separate log head record discovery from verification

Pull vfs fixes from Al Viro:
 "A couple of fixes: Fix for my dumb braino in ncpfs and a long-standing
  breakage on recovery from failed rename() in jffs2"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  jffs2: reduce the breakage on recovery from halfway failed rename()
  ncpfs: fix a braino in OOM handling in ncp_fill_cache()

* device-properties-fixes:
  device property: fwnode->secondary may contain ERR_PTR(-ENODEV)

* acpica-fixes:
  ACPICA: Revert "Parser: Fix for SuperName method invocation"

After the GMBUS transfer times out, we set force_bit=1 and
return -EAGAIN expecting the i2c core to call the .master_xfer
hook again so that we will retry the same transfer via bit-banging.
This is in case the gmbus hardware is somehow faulty.

Unfortunately we left adapter->retries to 0, meaning the i2c core
didn't actually do the retry. Let's tell the core we want one retry
when we return -EAGAIN.

Note that i2c-algo-bit also uses this retry count for some internal
retries, so we'll end up increasing those a bit as well.

Cc: Jani Nikula <jani.nikula@intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: bffce907 ("drm/i915: abstract i2c bit banging fallback in gmbus xfer")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1457366220-29409-2-git-send-email-ville.syrjala@linux.intel.comReviewed-by: Jani Nikula <jani.nikula@intel.com>
(cherry picked from commit 8b1f165a)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

Pull KVM fixes from Paolo Bonzini:
 "A few simple fixes for ARM, x86, PPC and generic code.

  The x86 MMU fix is a bit larger because the surrounding code needed a
  cleanup, but nothing worrisome"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: MMU: fix reserved bit check for ept=0/CR0.WP=0/CR4.SMEP=1/EFER.NX=0
  KVM: MMU: fix ept=0/pte.u=1/pte.w=0/CR0.WP=0/CR4.SMEP=1/EFER.NX=0 combo
  kvm: cap halt polling at exactly halt_poll_ns
  KVM: s390: correct fprs on SIGP (STOP AND) STORE STATUS
  KVM: VMX: disable PEBS before a guest entry
  KVM: PPC: Book3S HV: Sanitize special-purpose register values on guest exit

Pull arm64 fixes from Will Deacon:
 "I thought we were done for 4.5, but then the 64k-page chaps came
  crawling out of the woodwork.  *sigh*

  The vmemmap fix I sent for -rc7 caused a regression with 64k pages and
  sparsemem and at some point during the release cycle the new hugetlb
  code using contiguous ptes started failing the libhugetlbfs tests with
  64k pages enabled.

  So here are a couple of patches that fix the vmemmap alignment and
  disable the new hugetlb page sizes whilst a proper fix is being
  developed:

   - Temporarily disable huge pages built using contiguous ptes

   - Ensure vmemmap region is sufficiently aligned for sparsemem
     sections"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: hugetlb: partial revert of 66b3923a
  arm64: account for sparsemem section alignment when choosing vmemmap offset