Read and increment context count atomic variable under a lock
to avoid race condition between read and increment. This is
necessary to make sure no process goes beyond the specified
context limit.
Change-Id: I483e2ac169beaff49e19b8ef1b46541f6eb740b0
Signed-off-by: Deepak Kumar <dkumar@codeaurora.org>
Added sysfs entries to show kgsl memory usage statistics.
gpumem_mapped: kgsl memory mapped in the process address space.
gpumem_unmapped: kgsl allocated memory but not mapped in process.
imported_mem: graphics memory not allocated by the kgsl.
Below is the sysfs path for new entries:
/sys/class/kgsl/kgsl/proc/<pid>/
Change-Id: I08c2014d28dc0ca1e2b54ebf966d00143b303b54
Signed-off-by: Amit Kushwaha <kushwaha@codeaurora.org>
Unload zap shader during device hibernation and reload it
during resume otherwise scm calls during post hibernation
GPU initialization will fail as there is possibility that
TZ driver is not aware of the hibernation.
Change-Id: I1f62fb97cbc8e6c3e3536d4d5260a543ca15b685
Signed-off-by: Suprith Malligere Shankaregowda <supgow@codeaurora.org>
Signed-off-by: Thomas Yun <wyun@codeaurora.org>
This patch ensures device resumes successfully after
XO shutdown without any KGSL error.
Change-Id: I9eb8e281bc62793dc7521ba72aaeecf946860851
Signed-off-by: Suprith Malligere Shankaregowda <supgow@codeaurora.org>
Enforce a limit of 200 contexts per process. This is done so
that a process cannot use up all the available contexts and
prevent other processes from starving.
Change-Id: I6eda123128381a213315561cd42335e0944f1f55
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
Signed-off-by: Archana Obannagari <aobann@codeaurora.org>
Currently submit ioctl getting blocked till the commands
gets added to ringbuffer incase inflight count is less
than context burst count. If the submit command happens
in GPU slumber state, it will add the GPU wakeup time to
submit IOCTL. This will add latency in preparing next frame
in CPU side. Defer commands submission to dispatcher worker,
if the GPU is in slumber state.
CRs-Fixed: 2055107
Change-Id: I099ba721e02bbcd8ccadb1bc518c7c1ef4fb7e21
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
Robust context attempts to perform a rendering that takes too long
whether due to an infinite loop in a shader or even just a rendering
operation that takes too long on the given hardware. This type of
attempts can result into GPU faults. Robust context expect driver
to replay IB instead skip IB and if it fails on replay context has
to be invalidated.
KGSL_CONTEXT_INVALIDATE_ON_FAULT flag allows draw context to execute
only replay policy on GPU fault recovery instead of going to default
recovery policy. User space has to set this flag during the context
creation.
Change-Id: If42dc5afc7d5ed1226b73ae5abfa2648d7acf2c3
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
Memory retention is needed only for NAP state but not for SLUMBER state.
Disables memory retention for core clock before entering SLUMBER to save
power.
Change-Id: I64a5ecec6fc90d662da8d9d793860e56b0c6473f
Signed-off-by: Deepak Kumar <dkumar@codeaurora.org>
Following changes been made to improve soft fault detection,
which will fix un clocked register access in dispatcher_do_fault()
and incorrect declaration of GPU soft fault.
i) Stop fault timer before entering to NAP state
ii) Don’t start fault timer if the dispatcher inflight count is zero
iii) Add ringbuffer empty check in _isidle()
iv) Add device state check in dispatcher_do_fault()
CRs-Fixed: 2012731
Change-Id: I5ce498029f389eeeb428b4ac7fb07afd84d5764c
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
A context may be detached without submitting any commands
to GPU ringbuffer. This may cause us to wait on a timestamp
that will never be retired. So return immediately from
adreno_drawctxt_wait_rb() if context has not submitted any
commands to GPU ringbuffer.
Change-Id: If8b3f8df92ec9b54a1a83d2f6704d4d15eb1b979
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
kgsl_ioctl_gpu_sparse_command() is added to for user to
specify list of binds/unbinds for a memory entry
and syncpoints they depend on. If user specifies both
create a sync object for syncpoints and bind object for
binds/unbinds and add them to dispatcher draw queue. Sync
object should be inserted before the bind object in the
draw queue. Once the bind object reaches the head of
draw queue the corresponding binds/unbinds are performed.
kgsl_ioctl_gpu_sparse_command() only accepts commands from
context created with flag KGSL_CONTEXT_SPARSE, commands
from all other context types will return an error.
Change-Id: Ib0a2361f854ae01d0d8090cdd48cfa96308daf93
Signed-off-by: Tarun Karra <tkarra@codeaurora.org>
By default A5xx GPUs use CP crash dumper to get GPU
snapshot in case of any fault.
At times it is required to disable crash dumper
in case of any abnormalities, add support to do so.
Change-Id: Iea6497778bcd711e769f0e509103bd3bd0fd8574
Signed-off-by: Rajesh Kemisetti <rajeshk@codeaurora.org>
Currently dispatcher accepts kgsl_cmdbatch object. This object
is a superset of all the types of objects dispatcher accepts.
Split kgsl_cmdbatch object to SYNC and IB/MARKER objects and
structure the code to make it easier for new type of objects
to be added to the dispatcher queue.
CRs-Fixed: 1054354
Change-Id: I2d482d1081ce6fdb7925243c88ce00ea6b864efe
Signed-off-by: Tarun Karra <tkarra@codeaurora.org>
Rename all cmdbatch to drawobj. This forms a platform
for future changes where cmdbatch is split into different
types of drawobjs.
CRs-Fixed: 1054353
Change-Id: Ib84bee679e859db34e0d1f8a0ac70319eabddf53
Signed-off-by: Tarun Karra <tkarra@codeaurora.org>
Add new sysfs nodes which satisfy a generic format requested
by customer. Also add a new node to track GPU temperature.
Create links to these nodes at a generic location:
/sys/kernel/gpu/
CRs-Fixed: 1064728
Change-Id: I414a07ff4f9ee14b8f882d15644b06a73d5fcf76
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
Few gpu faults need system level information like voltages,
interface clock etc. This information can't be extracted
through the snapshot dump. Add a facility to force panic
on gpu fault, which will help to extract additional
information from the ramdumps.
To trigger BUG_ON for a gpu fault:
echo 1 > /sys/class/kgsl/kgsl-3d0/snapshot/force_panic
Change-Id: I020b5518fcaedd4fecc572b580a989bf117f6ceb
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
Add a new linux header to support 64 bit pagetable address in Linux OS
section.
Add a new linux context header to support v2+ Linux version in Linux
context section.
Remove multiple register reads on CP IB registers and have a single
CP register read to get the active IB information.
Add Active IBs to obj_list in case static snapshot blob is not having
enough memory, so that it get dumped into dynamic blob.
Add a message in case any active IB is not getting dumped into snapshot
binary.
Change-Id: Icee01f6ba95e469acac9eed6bf2fdbc83947f5a8
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
Call clk_set_flag() to turn off both memory core and periphery for
bimc_gfx_clk clock and memory for gfx_3d.
CRs-Fixed: 1046649
Change-Id: I941f91eeba01f4e7aa5427056bc57875e7edf197
Signed-off-by: Oleg Perelet <operelet@codeaurora.org>
DEEP-NAP and SLEEP states are not used in targets of previous
two generations. They are neither saving GPU power, nor saving
system power. Remove to reduce maintenance overhead.
CRs-Fixed: 1053516
Change-Id: If2fc2701548f90bb7ea9559a87752e13a7b0f736
Signed-off-by: George Shen <sqiao@codeaurora.org>
Scheduling issues were occurring with the GPU event worker after
b7be807 (msm: kgsl: Unbind the kgsl-event workqueue) was merged.
In certain conditions, it seems that the kgsl-event workqueue
was conflicting with the KGSL worker and slowing it down.
It turns out that everywhere we schedule the event worker
and the dispatcher worker at the same time. Since the worker
is singlethread, the event worker and the dispatcher run
synchronously anyway, so it makes sense to run the event processor
from within the dispatcher and save the extra schedule.
CRs-Fixed: 1043509
Change-Id: Ic0dedbad67eb04d41afb6add4477f146dfff9784
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
The memstore shared between the CPU and GPU is old but can not be
messed with. Rather than stealing values from it where available,
add a new block of shared memory that is exclusive to the driver
and GPU. This block can be used more freely than the old
memstore block.
Program the GPU to write the RPTR out to an address the CPU can read rather
than having the CPU read a GPU register directly. There are some very
small but very real conditions where different blocks on the GPU have
outdated values for the RPTR. When scheduling preemption the value read
from the register could not reflect the actual value of the RPTR in the CP.
This can cause the save/restore from preemption to give back incorrect RPTR
values causing much confusion between the GPU and CPU.
Remove the ringbuffers copy of the read pointer shadow.
Now that the GPU will update a shared memory address with the
value of the read pointer, there is no need to poll the register
to get the value and then keep a local copy of it.
CRs-Fixed: 987082
Change-Id: Ic44759d1a5c6e48b2f0f566ea8c153f01cf68279
Signed-off-by: Carter Cooper <ccooper@codeaurora.org>
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
Remove unused #defines, structs and members that are no longer used.
CRs-Fixed: 971156
Change-Id: Ibdf6fef6f3f700f3c5315c228c0473e47fb62163
Signed-off-by: Carter Cooper <ccooper@codeaurora.org>
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
The active context tracker that we bolted on the side of the
dispatcher was designed mainly to see if there was 1 OR more than 1
active context, not much more than that. Since it is apparent that
we'll need to track up to 4 contexts and possibly more later the
algorithm needs to change.
The new algorithm puts all active contexts on a linked list - every
time a context is used it is popped from the list and put on the
tip with an updated time. To count the number of active contexts
walk the list until you get a context with an active time older than
you are looking for. You also can do other magic on the context,
like see if it matches up with a given command queue.
Change-Id: Ic0dedbad6be9fd1925121ee54e0000c42b089f44
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: Prakash Kamliya <pkamliya@codeaurora.org>
Global pagetable entries are exclusively for IOMMU and per-process
pagetables. Move all the code out of the generic driver and into
the IOMMU driver and clean up a bunch of stuff along the way.
Change-Id: Ic0dedbadbb368bb2a289ba4393f729d7e6066a17
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
Make the various timeout values HZ agnostic by using the proper
macros and values instead.
Change-Id: I708cd491f593782f0172cd7d2cca058cd41044a5
Signed-off-by: Suman Tatiraju <sumant@codeaurora.org>
struct kgsl_mmu is a static member of struct kgsl_device so we can
use the usual container_of trick to get the device from a mmu
pointer rather than carry around an unneeded back reference.
Change-Id: Ic0dedbad7ff22e598b03d980dfbb738374ed5a7a
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
If there are concurrent sysfs reads of snapshot binary
there can be a race condition where the snapshot data
is prematurely free'd by one reader while the other reader
is still reading it. Fix this by proper refcounting using
an atomic.
CRs-Fixed: 902816
Change-Id: I7a156c3a22f5475df0394ae30328d0fd6140f3da
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
Make several changes to build the GPU driver for 4.4:
- Rename CONFIG_MSM to CONFIG_QCOM where applicable
- Add msm_kgsl.h to the Kbuild exports
- Remove linux/coresight_of.h (as it has been merged into
coresight.h) and remove the .owner member of the
coresight_desc struct.
- Use the new location for the sync.h file (in staging)
- Remove an unused sync function
- Move oneshot_sync.h inside of #ifdef wrappers
Signed-off-by: Jordan Crouse <jcrouse@codeauorora.org>
Snapshot of the Qualcom Adreno GPU driver (KGSL) as of msm-3.18 commit
commit e70ad0cd5efd ("Promotion of kernel.lnx.3.18-151201.").
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
