Softpin allows userspace to take greater control of GPU virtual address
space and eliminates the need of relocations. It can also be used to
mirror addresses between GPU and CPU (shared virtual memory).
Calls to drm_intel_bo_emit_reloc are still required to build the list of
drm_i915_gem_exec_objects at exec time, but no entries in relocs are
created. Self-relocs don't make any sense for softpinned objects and can
indicate a programming errors, thus are forbidden. Softpinned objects
are marked by asterisk in debug dumps.
Cc: Thomas Daniel <thomas.daniel@intel.com>
Cc: Kristian Høgsberg <krh@bitplanet.net>
Cc: Zou Nanhai <nanhai.zou@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Gen8+ supports 48-bit virtual addresses, but some objects must always be
allocated inside the 32-bit address range.
In specific, any resource used with flat/heapless (0x00000000-0xfffff000)
General State Heap (GSH) or Instruction State Heap (ISH) must be in a
32-bit range, because the General State Offset and Instruction State Offset
are limited to 32-bits.
The i915 driver has been modified to provide a flag to set when the 4GB
limit is not necessary in a given bo (EXEC_OBJECT_SUPPORTS_48B_ADDRESS).
48-bit range will only be used when explicitly requested.
Callers to the existing drm_intel_bo_emit_reloc function should set the
use_48b_address_range flag beforehand, in order to use full ppgtt range.
v2: Make set/clear functions nops on pre-gen8 platforms, and use them
internally in emit_reloc functions (Ben)
s/48BADDRESS/48B_ADDRESS/ (Dave)
v3: Keep set/clear functions internal, no-one needs to use them directly.
v4: Don't set 48bit-support flag in emit reloc, check for ppgtt type
before enabling set/clear function, print full offsets in debug
statements, using port of lower_32_bits and upper_32_bits from linux
kernel (Michał)
References: http://lists.freedesktop.org/archives/intel-gfx/2015-July/072612.html
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
...to minimise misuse of bo_gem.
If the variable is declared at the top of the function and then used
for two (or more) different contexts this can cause confusion and errors.
Just introduce a wrapper, which can be used in a once off situations.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by Chris Wilson <chris@chris-wilson.co.uk>
Just like we do for the original exec()
v2: move bo_gem declaration to the top of the function.
Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by Chris Wilson <chris@chris-wilson.co.uk>
v2: keep the bo_gem declaration in exec2() within the loop (Chris)
Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by Chris Wilson <chris@chris-wilson.co.uk>
It is possible to hit a race condition in create_from_prime, when trying
to import a BO that's currently being freed. In case of prime sharing
we'll succesfully get a handle, but fail on get_tiling call, potentially
confusing the caller (and requiring different locking scheme than with
sharing using flink). Wrap fd_to_handle with struct_mutex to force
a more consistent behaviour between prime/flink, convert fprintf to DBG
when handling errors.
(From Chris:
The race is that the kernel returns us the same file-private handle as
the first thread, but that first thread is about to call gem_close
(thereby removing the handle from the file completely) and does so
between us acquiring the handle and taking the mutex. If we take
the mutex, then we acquire the refcnt on the bo prior to the first
thread completing its unref (and so preventing the early close). Or we
acquire the handle after the earlier close, in which case we are the new
owner.
)
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Testcase: igt/drm_import_export/import-close-race-prime
Signed-off-by: Rafał Sapała <rafal.a.sapala@intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
We now have a separate tool for this in intel-gpu-tools and we don't
need to clutter up libdrm with this feature. We leave the entry points
in there to avoid breaking API/ABI.
Install intel-gpu-tools, then run (for example)
$ intel_aubdump --output=trace.aub glxgears -geometry 500x500
See the intel_aubdump man page for more details.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
In case of YF/YS tiled buffers libdrm need not know about the tiling
format because these buffers don't have hardware support to be tiled
or detiled through a fenced region. But, libdrm still need to know
about buffer alignment restrictions because kernel uses it when
resolving the relocation.
Mesa uses drm_intel_gem_bo_alloc_for_render() to allocate Yf/Ys buffers.
So, use the passed alignment value in this function to initialize the
align variable in drm_intel_bo. Note that we continue ignoring the
alignment value passed to drm_intel_gem_bo_alloc() to follow the
previous behavior.
V2: Add a condition to avoid allocation from cache. (Ben)
V3: Make no changes in cache allocation strategy. Just update the alignment.
Update the aperture size estimate including the alignment. (Ben, Chris)
V4: Move aperture size adjustments inside drm_intel_bo_gem_set_in_aperture_size()
Don't split sentences across the one-line header and the changelog. (Chris)
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
In order to use userptr, the kernel tracks the owner's mm with a
mmu_notifier. Setting that is very expensive - it involves taking all
mm_locks and a stop_machine(). This tracking lives only for as long as
the client is using userptr objects - so if the client allocates then
frees a userptr in a loop, we will be executing that heavyweight setup
everytime. To ammoritize this cost, just leak the test bo and the single
backing page we use for detecting userptr.
v2: Free the object and memory when bufmgr is destroyed.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Some compilers (like the Oracle Studio), require that the function
declaration must be annotated with the same visibility attribute as the
definition. As annotating functions with drm_public is no longer
required just remove the macro.
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: Michel Dänzer <michel.daenzer@amd.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Thierry Reding <treding@nvidia.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Running __mmu_notifier_register() is surprisingly expensive, so let's
not do that unless we have to.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Update kernel interface with new I915_GETPARAM ioctl entries for
subslice total and EU total. Add a wrapping function for each
parameter. Userspace drivers need these values when constructing
GPGPU commands. This kernel query method is intended to replace
the PCI ID-based tables that userspace drivers currently maintain.
The kernel driver can employ fuse register reads as needed to
ensure the most accurate determination of GT config attributes.
This first became important with Cherryview in which the config
could differ between devices with the same PCI ID.
The kernel detection of these values is device-specific. Userspace
drivers should continue to maintain ID-based tables for older
devices which return ENODEV when using this query.
v2: remove unnecessary include of <stdbool.h> and increment the
I915_GETPARAM indices to match updated kernel patch.
For: VIZ-4636
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
This reverts commit 080b4929b7.
Chris noticed that "negative values wait forever" is indeed intended
behaviour and the issue is just that we didn't have a testcase (fixed
now) and that a regression slipped through (fixed and on track for all
stable kernels).
So lets undo the documentation change for consistency, since working
around kernel regressions isn't good. Practical impact is nil anyway.
v2: Add a note to docs that some kernels have been broken.
v3: Remove the random garbage included by accident.
Cc: Kristian Høgsberg <krh@bitplanet.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
The kernel doesn't actually wait indefinately when passed a negative,
timeout, it returns immediately. Document this and suggest using INT64_MAX
for indefinite waits.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We really have to do this to avoid surprises when extending the ABI
later on. Especially when growing the structures.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
On recent emulator GTT entry setup for aub dump needs mem type as
GTT_ENTRY instead of NONLOCAL. NONLOCAL would write data in main
memory space which is wrong with new memory layout. GTT_ENTRY write
would setup GTT memory pool and other required internal buffers. With
this I can run aub dump on latest release without crash.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
drm_intel_gem_bo_free() crashes because the list bo_gem->vma_list is not
yet initialised, but the error path tries to free it.
See also https://bugs.freedesktop.org/show_bug.cgi?id=75844
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Thomas Meyer <thomas@m3y3r.de>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
When handling the error on GEM_CLOSE, we weren't freeing the allocated
page. Plug that.
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
The changes make sure that members of the bufmgr_gem and bo_gem
name lists are sychronized between threads
when using the create from prime and create from name methods.
Signed-off-by: Rafal Sapala <rafal.a.sapala@intel.com>
Testcase: igt/drm_import_export
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Allow userptr objects to be created and used via libdrm_intel.
At the moment tiling and mapping to GTT aperture is not supported
due hardware limitations across different generations and uncertainty
about its usefulness.
v2: Improved error handling in feature detection per review comments.
v3: Rebase on top of the drm_public addition, minor whitespace addition.
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com> (v3)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> (v1,v2)
When using Mesa and LibVA in the same process, one would like to be
able bind buffers from the output of the decoder to a GL texture
through an EGLImage.
LibVA can reuse buffers allocated by Gbm through a file descriptor. It
will then wrap it into a drm_intel_bo with
drm_intel_bo_gem_create_from_prime().
The problem at the moment is that both library get a different
drm_intel_bufmgr object when they call drm_intel_bufmgr_gem_init()
even though they're using the same drm file descriptor. As a result,
instead of manipulating the same buffer object for a given file
descriptor, they get 2 different drm_intel_bo objects and 2 different
refcounts, leading one of the library to get errors from the kernel on
invalid BO when one of the 2 library is done with a shared buffer.
This patch modifies drm_intel_bufmgr_gem_init() so, given a file
descriptor, it will look for an already existing drm_intel_bufmgr
using the same file descriptor and return that object.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
No exports changed for this driver.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The existing 'offset' field is unfortunately typed as 'unsigned long',
which is unfortunately only 4 bytes with a 32-bit userspace.
Traditionally, the hardware has only supported 32-bit virtual addresses,
so even though the kernel uses a __u64, the value would always fit.
However, Broadwell supports 48-bit addressing. So with a 64-bit kernel,
the card virtual address may be too large to fit in the 'offset' field.
Ideally, we would change the type of 'offset' to be a uint64_t---but
this would break the libdrm ABI. Instead, we create a new 'offset64'
field to hold the full 64-bit value from the kernel, and store the
32-bit truncation in the existing 'offset' field, for compatibility.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
I've seen a number of apps spending unreasonable amounts of time in
drm_intel_bo_busy during the buffer mapping process.
We can't track idleness in general, in the case of buffers shared
across processes. But this should significantly reduce our overhead
for checking for busy on things like VBOs.
Improves (unoptimized) glamor x11perf -f8text by 0.243334% +/-
0.161498% (n=1549), which has formerly been spending about .5% of its
time hitting the kernel for drm_intel_gem_bo_busy().
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The previous code would just use the potentially unallocated variable,
which is probably okay most of the time, but not very nice to the user
of the library.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
If the application sends us a file descriptor pointing at a prime
buffer that we've already got, we have to re-use the same bo_gem
structure or chaos will result.
Track the set of all known prime objects and look to see if the kernel
has returned one of those for a new file descriptor.
Also checks for prime buffers in the flink case.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The ioctl expects that certain fields will be zeroed, so we should allow
the helper function to actually work in non-Valgrind builds.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reported-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.
This ioctl is in the drm-intel-next tree as b635991.
v2: Update based on Mika's kernel work.
v3: Fix compile failures from last-minute typos. Sigh.
v4: Import the actual changes from the kernel i915_drm.h. Only comments
on some fields of drm_i915_reset_stats differed. There are still some
deltas between the kernel i915_drm.h and the one in libdrm, but those
can be resolved in other patches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> [v3]
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
This reverts commit 6335e1d28c.
No taxation without representation, in other words no userspace without kernel
stuff being in a stable location, either drm-next but I'll accept drm-intel-next
for intel specific stuff.
I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.
v2: Update based on Mika's kernel work.
v3: Fix compile failures from last-minute typos. Sigh.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
The command now takes a 48bits address and is thus 1 dword longer.
v2 (Ben): commit message: s/byte/dword (Eric)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Since our aub file dumping's GTT handling is totally fake, we always put
everything in the low 4GB anyway and shouldn't ever need to set
AddressHigh to anything other than 0.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
[ben: slight commit message change]
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
The various create and open functions set the buffer size, but
drm_intel_bo_gem_create_from_prime() is an exception. In the 3.12 kernel
we can now use lseek on the prime fd to determine the size of the bo.
Use that and override the userprovided size. If the kernel doesn't
support this, we get an error and fall back to the user provided size.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Mark the address ranges as accessible with VALGRIND_MAKE_MEM_DEFINED.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
No need to prepare the .aub header and dump in that case, it'll be
done with the next call with true.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We didn't set the ring flag for BLT batches, so they got run on the
render ring. Shenanigans ensued, especially when we sent commands that
were only valid on the BLT ring.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
As we clear the relocs from the bo, we also need to clear the
contribution of the reloc_target_bo from the fence count. Otherwise they
are leaked and prevent any further relocations being added to the bo.
Eric Engestrom
Michał Winiarski
Michel Thierry
Emil Velikov
Rafał Sapała
Kristian Høgsberg Kristensen
Anuj Phogat
Tvrtko Ursulin
Chris Wilson
Jeff McGee
Daniel Vetter
Kristian Høgsberg
Zhenyu Wang
Thomas Meyer
Damien Lespiau
Lionel Landwerlin
Maarten Lankhorst
Kenneth Graunke
Eric Anholt
Ben Widawsky
Keith Packard
Ian Romanick
Dave Airlie
Ben Widawsky
Chia-I Wu
Xiang, Haihao