- make createGraphicsAllocationFromSharedHandle and
freeGraphicsMemory more threadsafe - there was race condition
where one thread called PRIME_FD_TO_HANDLE ioctl when the other just
closed the last handle with GEM_CLOSE ioctl. This led to situation
where newly acquired handle from PRIME_FD_TO_HANDLE was already invalid
and used in exec.
- adding mutex for the time of accessing shared handles container
and ioctls (GEM_CLOSE and PRIME_FD_TO_HANDLE) resolves racing problem
Change-Id: I8417a036a1429be8f1ba97e63ebdda2707960564
- This is first phase of moving wddm specific internals to Wddm
Interface helper
- Instead of having 2 different wddm classes driver will route interface
specific functions to dedicated helper
- Helper will be initialized when interface version will be known,
therefore we would not need to initialize wddm multiple times
Change-Id: Ic71788ccb2f8a71bf2f3f3c2a04117f16417d85e
- OsInterface is now a part of execution environment
- it is created when getDevice is being called
- move ownership of wddm from Wddm Memory manager to OSInterface
- reuse osInterface on Linux in Command Stream Receiver
- currently OsInterface is not reused upon command stream receiver creation
on Windows this will change in further commits.
- make enumAdapters non static function
Change-Id: I10f36c01e6729f48653e3b5c11cbc32e811ce754
This commit addresses crash issue in AUB standalone mode
when queue profiling is enabled on Linux platform.
Change-Id: Ie2d02093d3628efac3a8925890f22afd6a49469d
- This is constant setting and will not change in device lifetime.
- This is a part of device removal from memory manager.
Change-Id: Ib80559bac67d65c1bcc488fb033be7343839a108
- resources are dumped in make non resident call
- in order to dump correct data we need to be sure that GPU is done processing
- waiting needs to be unconditional to handle all cases
- remove not needed parameter to makeSurfacePackNonResident
Change-Id: Ib2b065d486cd3a5d86e599c51b24f3c958c3a10b
- createGraphicsAllocationFromSharedHandle was not threadsafe,
instead of reusing BufferObject for a single handle when
multiple threads were creating memory objects from a single VASurface,
new BO could be created and placed in container with BOs for reuse.
This was leading to errors in ioctl calls.
- add lock for following set of operations:
1. find BufferObject with a given handle in container
2. create shared BO when not found
3. add shared BO to container
prevents creating multiple BOs for a single handle
- replace recursive mutex with regular mutex as mutex shouldn't
be locked recursively
Change-Id: I0937e2abf3bf1c672c6d77422d46e441f7216a68
- do not always expect failures in tests with failure injections
there is retry mechanism for some cases
Change-Id: If7589d2dacc41216d2f3b08f861209bbab179615
- new ExtendableEnum struct that serves as enum but can
be extended with values
- decrease dependencies on graphics_allocation.h header -
use forward class declaration when possible
- memoryPool indicates what kind of memory is allocated
for a given GraphicsAllocation
Change-Id: I7a707c28dc4544cc73abc5f0ed5263ba5be17452
- replace createGraphicsAllocationWithRequiredBitness with more general
methodallocateGraphicsMemoryInPreferredPool based on passed
AllocationData
- proper flags for allocation selected based on AllocationType
- remove allocateGraphicsMemory(size_t size, size_t alignment)
and use allocateGraphicsMemory(size_t size) instead where default
alignment is sufficient, otherwise use full options version:
allocateGraphicsMemory(size_t size, size_t alignment,
bool forcePin, bool uncacheable)
Change-Id: I2da891f372ee181253cb840568a61b33c0d71fc9
- change GraphicsAllocatoin::AllocationType to scoped enumeration
so that ALLOCATION_TYPE_ prefix in every enum value can be removed
- all accesses are typed (example AllocationType::IMAGE)
- Rename allocationType to AllocationUsage to eliminate confusion
with multiple AllocationType enums / types
Change-Id: I16003297ecfcb0aaa5779ad00706c5d983914bbe
- makeCoherent should be called after TBX finished processing
- this is when tagAddress is updated with taskCount
makeCoherent is called from makeNonResident which is invoked just
after flush and may happen before TBX server finished processing
leading to invalid data to be read back to CPU accessible memory
- this fix adds waiting for taskCount to blocking calls for TBX CSR
before calling makeNonResident on surfaces to guarantee correct data
from TBX server is ready.
Change-Id: I498a5454e0826eec2a5413a08880af40268550e1
previously we couldn't add new variable to capabilityTable structure
becasue then we get too large function argument passed by
value
Change-Id: I162967ba195ec18fc781d127012f82e8876913a6
- move AUBDumpToggleCaptureOnOff from HKLM to HKCU without admin rights
- fix AUB subcapture in toggle mode to call make resident in standalone mode
Change-Id: Ia6971921f33bb9ca63112790af870217da8a1585
Hoppe, Mateusz