The blocking gather copy allocation is a major performance downside of the
Host1x firewall, it may take hundreds milliseconds which is unacceptable
for the real-time graphics operations. Let's try a non-blocking allocation
first as a least invasive solution, it makes opentegra (Xorg driver)
performance indistinguishable with/without the firewall.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Check waits in the firewall in a way it is done for relocations.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Several channels could be made to write the same unit concurrently via
the SETCLASS opcode, trusting userspace is a bad idea. It should be
possible to drop the per-client channel reservation and add a per-unit
locking by inserting MLOCK's to the command stream to re-allow the
SETCLASS opcode, but it will be much more work. Let's forbid the
unit-unrelated class changes for now.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The RESTART opcode terminates the gather and restarts the CDMA fetching
from a specified word << 2 relative to the CDMA start address. That
shouldn't be allowed to be done by userspace.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Incorrectly shifted relocation address will cause a lower memory
corruption and likely a hang on a write or a read of an arbitrary data
in case of IOMMU absence. As of now, there is no known use for the
address shifting and adding a proper shifts / sizes validation is a much
more work. Let's forbid shifts in the firewall till a proper validation
is implemented.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Perform gathers coping before patching them, so that original gathers are
left untouched. That's not as bad as leaking kernel addresses, but still
doesn't feel right.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
In case of relocations / waitchecks patching failure the jobs pins stay
referenced till DRM file get closed, wasting memory. Add the missed
unpinning.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The commands stream is prepended by the jobs class on the CDMA
submission, so that explicitly setting a module class in the commands
stream isn't necessary. The firewall initializes its class to 0 and the
command stream that doesn't explicitly specify the class effectively
bypasses the firewall.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Reviewed-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add support for the Host1x unit to be located behind
an IOMMU. This is required when gather buffers may be
allocated non-contiguously in physical memory, as can
be the case when TegraDRM is also using the IOMMU.
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Currently job pinning is optimized to handle only the first buffer
using a certain host1x_bo object and all subsequent buffers using
the same host1x_bo are considered done.
In most cases this is correct, however, in case the same host1x_bo
is used in multiple gathers inside the same job, we skip also
storing the device address (physical or iova) to this buffer.
This patch reworks the host1x_job_pin() to store the device address
to all gathers.
Signed-off-by: Andrew Chew <achew@nvidia.com>
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Insert a number of blank lines in places where they increase readability
of the code. Also collapse various variable declarations to shorten some
functions and finally rewrite some code for readability.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Fix a couple of occurrences where no blank line was used to separate
variable declarations from code or where block comments were wrongly
formatted.
Signed-off-by: Thierry Reding <treding@nvidia.com>
IDs can never be negative so use unsigned int. In some instances an
explicitly sized type (such as u32) was used for no particular reason,
so turn those into unsigned int as well for consistency.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Pull drm updates from Dave Airlie:
"This is the main drm pull request for 4.6 kernel.
Overall the coolest thing here for me is the nouveau maxwell signed
firmware support from NVidia, it's taken a long while to extract this
from them.
I also wish the ARM vendors just designed one set of display IP, ARM
display block proliferation is definitely increasing.
Core:
- drm_event cleanups
- Internal API cleanup making mode_fixup optional.
- Apple GMUX vga switcheroo support.
- DP AUX testing interface
Panel:
- Refactoring of DSI core for use over more transports.
New driver:
- ARM hdlcd driver
i915:
- FBC/PSR (framebuffer compression, panel self refresh) enabled by default.
- Ongoing atomic display support work
- Ongoing runtime PM work
- Pixel clock limit checks
- VBT DSI description support
- GEM fixes
- GuC firmware scheduler enhancements
amdkfd:
- Deferred probing fixes to avoid make file or link ordering.
amdgpu/radeon:
- ACP support for i2s audio support.
- Command Submission/GPU scheduler/GPUVM optimisations
- Initial GPU reset support for amdgpu
vmwgfx:
- Support for DX10 gen mipmaps
- Pageflipping and other fixes.
exynos:
- Exynos5420 SoC support for FIMD
- Exynos5422 SoC support for MIPI-DSI
nouveau:
- GM20x secure boot support - adds acceleration for Maxwell GPUs.
- GM200 support
- GM20B clock driver support
- Power sensors work
etnaviv:
- Correctness fixes for GPU cache flushing
- Better support for i.MX6 systems.
imx-drm:
- VBlank IRQ support
- Fence support
- OF endpoint support
msm:
- HDMI support for 8996 (snapdragon 820)
- Adreno 430 support
- Timestamp queries support
virtio-gpu:
- Fixes for Android support.
rockchip:
- Add support for Innosilicion HDMI
rcar-du:
- Support for 4 crtcs
- R8A7795 support
- RCar Gen 3 support
omapdrm:
- HDMI interlace output support
- dma-buf import support
- Refactoring to remove a lot of legacy code.
tilcdc:
- Rewrite of pageflipping code
- dma-buf support
- pinctrl support
vc4:
- HDMI modesetting bug fixes
- Significant 3D performance improvement.
fsl-dcu (FreeScale):
- Lots of fixes
tegra:
- Two small fixes
sti:
- Atomic support for planes
- Improved HDMI support"
* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (1063 commits)
drm/amdgpu: release_pages requires linux/pagemap.h
drm/sti: restore mode_fixup callback
drm/amdgpu/gfx7: add MTYPE definition
drm/amdgpu: removing BO_VAs shouldn't be interruptible
drm/amd/powerplay: show uvd/vce power gate enablement for tonga.
drm/amd/powerplay: show uvd/vce power gate info for fiji
drm/amdgpu: use sched fence if possible
drm/amdgpu: move ib.fence to job.fence
drm/amdgpu: give a fence param to ib_free
drm/amdgpu: include the right version of gmc header files for iceland
drm/radeon: fix indentation.
drm/amd/powerplay: add uvd/vce dpm enabling flag to fix the performance issue for CZ
drm/amdgpu: switch back to 32bit hw fences v2
drm/amdgpu: remove amdgpu_fence_is_signaled
drm/amdgpu: drop the extra fence range check v2
drm/amdgpu: signal fences directly in amdgpu_fence_process
drm/amdgpu: cleanup amdgpu_fence_wait_empty v2
drm/amdgpu: keep all fences in an RCU protected array v2
drm/amdgpu: add number of hardware submissions to amdgpu_fence_driver_init_ring
drm/amdgpu: RCU protected amd_sched_fence_release
...
The return type "unsigned int" was used by the do_relocs() function
despite the fact that it will eventually return a negative error code.
Use a signed integer instead to accomodate for error codes.
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Rename dma_*_writecombine() to dma_*_wc(), so that the naming
is coherent across the various write-combining APIs. Keep the
old names for compatibility for a while, these can be removed
at a later time. A guard is left to enable backporting of the
rename, and later remove of the old mapping defines seemlessly.
Build tested successfully with allmodconfig.
The following Coccinelle SmPL patch was used for this simple
transformation:
@ rename_dma_alloc_writecombine @
expression dev, size, dma_addr, gfp;
@@
-dma_alloc_writecombine(dev, size, dma_addr, gfp)
+dma_alloc_wc(dev, size, dma_addr, gfp)
@ rename_dma_free_writecombine @
expression dev, size, cpu_addr, dma_addr;
@@
-dma_free_writecombine(dev, size, cpu_addr, dma_addr)
+dma_free_wc(dev, size, cpu_addr, dma_addr)
@ rename_dma_mmap_writecombine @
expression dev, vma, cpu_addr, dma_addr, size;
@@
-dma_mmap_writecombine(dev, vma, cpu_addr, dma_addr, size)
+dma_mmap_wc(dev, vma, cpu_addr, dma_addr, size)
We also keep the old names as compatibility helpers, and
guard against their definition to make backporting easier.
Generated-by: Coccinelle SmPL
Suggested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Luis R. Rodriguez <mcgrof@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: airlied@linux.ie
Cc: akpm@linux-foundation.org
Cc: benh@kernel.crashing.org
Cc: bhelgaas@google.com
Cc: bp@suse.de
Cc: dan.j.williams@intel.com
Cc: daniel.vetter@ffwll.ch
Cc: dhowells@redhat.com
Cc: julia.lawall@lip6.fr
Cc: konrad.wilk@oracle.com
Cc: linux-fbdev@vger.kernel.org
Cc: linux-pci@vger.kernel.org
Cc: luto@amacapital.net
Cc: mst@redhat.com
Cc: tomi.valkeinen@ti.com
Cc: toshi.kani@hp.com
Cc: vinod.koul@intel.com
Cc: xen-devel@lists.xensource.com
Link: http://lkml.kernel.org/r/1453516462-4844-1-git-send-email-mcgrof@do-not-panic.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Job submission currently relies on the fact that struct drm_tegra_reloc
and struct host1x_reloc are the same size and uses a simple call to the
copy_from_user() function to copy them to kernel space. This causes the
handle to be stored in the buffer object field, which then needs a cast
to a 32 bit integer to resolve it to a proper buffer object pointer and
store it back in the buffer object field.
On 64-bit architectures that will no longer work, since pointers are 64
bits wide whereas handles will remain 32 bits. This causes the sizes of
both structures to because different and copying will no longer work.
Fix this by adding a new function, host1x_reloc_get_user(), that copies
the structures field by field.
While at it, use substructures for the command and target buffers in
struct host1x_reloc for better readability. Also use unsized types to
make it more obvious that this isn't part of userspace ABI.
Signed-off-by: Thierry Reding <treding@nvidia.com>
When patching gathers, we don't need to check against
gathers with lower indices than the current one, as
they are guaranteed to already have been handled.
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Acked-By: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Expose the buffer objects, syncpoint and channel functionality in the
public public header so that drivers can use them.
Signed-off-by: Thierry Reding <treding@nvidia.com>
The same code sequence is used in various places to validate a register
access in the command stream. This can be refactored into a separate
function.
Signed-off-by: Thierry Reding <treding@nvidia.com>
The value stored in this field is a pointer to a command buffer, not an
ID. Avoid some confusion by reflecting that in the field's name.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Arguments on subsequent lines should be aligned with the first argument.
This one occurrence went unnoticed during code review.
Signed-off-by: Thierry Reding <treding@nvidia.com>
The num_relocs count are passed to the kernel per job, not per gather.
For multi-gather jobs, we would previously fail if there were relocs in
other gathers aside from the first one.
Fix this by simply moving the check until all gathers have been
consumed.
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Arto Merilainen <amerilainen@nvidia.com>
Acked-By: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
There is a mistake here so it returns PTR_ERR(NULL) which is success
instead of -ENOMEM.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Tegra is a 32 bit arch. On 32 bit systems then size_t is 32 bits so
"total" will never be higher than UINT_MAX because of integer overflows.
We need cast to u64 first before doing the math.
Also the addition earlier:
unsigned int num_unpins = num_cmdbufs + num_relocs;
That can overflow as well, but I think it's still safe because we check
both "num_cmdbufs" and "num_relocs" again in this test.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The firewall verified gather buffers before copying them. This
allowed a malicious application to rewrite the buffer content by
timing the rewrite carefully.
This patch makes the buffer validation occur after copying the
buffers.
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
The firewall was reinitialised for each gather. Because the filter
was reinitialised, it did not track the class over gather boundaries.
This allowed the user application to set host1x class to one class
in one gather and use that class in another gather without firewall
having knowledge about that.
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
The firewall assumed that the user space always delivers a relocation
table when it is accessing address registers. If userspace did not
deliver a relocation table and tried to access the address registers,
the code performed bad memory accesses.
This patch modifies the firewall to check correctly that the firewall
table is available before accessing it. In addition, check_reloc() is
converted to use boolean return value (true when the reloc is valid,
false when invalid).
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Acked-By: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
The firewall code used a wrong loop condition (pointer to a
structure) while checking INCR opcode. This patch fixes the code to
use correct loop condition (number of words remaining).
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Add support for host1x client modules, and host1x channels to submit
work to the clients.
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: Thierry Reding <thierry.reding@avionic-design.de>
Tested-by: Thierry Reding <thierry.reding@avionic-design.de>
Tested-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
