The bottom up search of drm_mm hole nodes is
currently broken. It does not correctly look
for the first hole (smallest address) that fits
the allocation from the bottom up. Use an RB tree
to record hole nodes with their address as key,
so that the hole with the smallest address that
satisfies the request can be looked up.
Change-Id: Ie3020cb4e8b6db13a35d0fe5caaa8781cb6ced9f
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
SVM buffer objects share the same virtual address on
both the CPU and GPU. Register for notifications when
SVM address ranges are unmapped on the CPU. When such
a notification is received, unmap the corresponding
SVM objects from the SMMU, after waiting on the most
recent fence that uses them. The notifier struct is
reference counted starting with the creation of the
first SVM bo in the process and is released when the
last SVM bo is freed.
Change-Id: I01f590d21fd1d146f5324539e5041f03653f858a
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
The reference count on a buffer object was being
decremented before it was initialized. Fix the
error path clean-up to put the reference only
after the kref has been initialized. Also, avoid
passing the object by reference when possible.
Change-Id: Ifbb1874deb5db9ae3fb2fbd40a58162998284e96
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Shared virtual memory allows the CPU and GPU to use
the same virtual address for a buffer. This change
implements an ioctl to allow creation of SVM buffer
objects.
Change-Id: I0d929a2e37a9eeef948dc2a37250c1eb9adf6fc7
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
At a higher level, all objects are created with definite size i.e. 0 is
illegal. In forthcoming patches, this assumption is dependent upon in
the drm_mm range manager, i.e. trying to create a drm_mm node with size
0 will have undefined behaviour. Add a couple of WARNs upon creating the
drm_mm node to prevent later bugs.
Change-Id: Ieb81a8ef5a05a32dd37dfc040ea3f78243b59161
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470248788-30873-1-git-send-email-chris@chris-wilson.co.uk
Git-commit: aafdcfd3f9aa3c9f77ae4e9385f21bf9ae120d3e
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
As we always add this to the drm_mm->hole_stack as our first operation,
we do not need to initialise the list node.
Change-Id: If2e9e06141bdb9a9411ef875cb83e41a48e5f1a9
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470236651-678-3-git-send-email-chris@chris-wilson.co.uk
Git-commit: a787900564ffc6eebbcb7086b45e04a0bc3370f0
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
In addition to the last-in/first-out stack for accessing drm_mm nodes,
we occasionally and in the future often want to find a drm_mm_node by an
address. To do so efficiently we need to track the nodes in an interval
tree - lookups for a particular address will then be O(lg(N)), where N
is the number of nodes in the range manager as opposed to O(N).
Insertion however gains an extra O(lg(N)) step for all nodes
irrespective of whether the interval tree is in use. For future i915
patches, eliminating the linear walk is a significant improvement.
v2: Use generic interval-tree template for u64 and faster insertion.
Change-Id: Iddcb7891480ca7d6e0469c8d394fcd1962ed4583
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1470236651-678-1-git-send-email-chris@chris-wilson.co.uk
Git-commit: 202b52b7fbf70858609ec20829c7d69a13ffa351
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Buffer object specific resources like pages, domains, sg list
need not be protected with struct_mutex. They can be protected
with a buffer object level lock. This simplifies locking and
makes it easier to avoid potential recursive locking scenarios
for SVM involving mmap_sem and struct_mutex. This also removes
unnecessary serialization when creating buffer objects, and also
between buffer object creation and GPU command submission.
Change-Id: I40cb437d0186c3d9aac365c9baba0aa4792f0aa1
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Restrict the virtual address size to 39-bits to allow
a maximum of 3 pagetable levels.
Change-Id: I264f23a2e16cc7599bdad8a161854bcf6e24dd4a
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Define a 36-bit address space for TTBR1 which is used for
kernel side GPU buffer objects.
Change-Id: I1c4eaee0fd92236793621c7d3dba1700e56fefd2
Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
Due to an egregious misunderstanding of the specification it
was thought that to do the TTBR1 matching correctly the sign
extension bit needed to match the input address space. This
is incorrect - the TTBR1 range is determined by the size of
the TTBR1 space (in our case, the same as the input address
space). For example if the input address size is 36, the
effective range of the pagetables are:
TTBR0: 0x00000000_00000000 - 0x0000000f_ffffffff
TTBR1: 0xfffffff0_00000000 - 0xffffffff_ffffffff
For its part the sign extension bit needs should be set based on
the upstream bus size. If the device has a UBS of 49 then the
sign extension bit is assumed by design to be 48 otherwise
the driver needs to pick the highest available bit and reduce
the input address space by 1.
Because the client driver shouldn't need to know the upstream
bus size, convention is to use a fully sign extended unsigned
long address for TTBR1 mappings. If the sign extension bit is set
lower than the upstream bus size some implementations assume that
bits above the sign extension bit need to be zero and breaks the
convention. Setting the sign extension bit correctly for the
upstream bus size ensures that sign extension always works.
The hardware will match an address to the TTBR1 if all the bits
between the sign extension bit and the input address size are set
to 1. We emulate this behavior in software when looking up a pagetable
for a software operation.
Change-Id: Ic0dedbad80c72f11bc8a7e6792f0e3c2f58bc674
Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
If flow control functionality is not enabled in traffic controller
userspace modules, WARN_ONCE messages are triggered. Note that qdisc
will still be setup even if these debug messages are logged.
Remove these messages to reduce log spam.
CRs-Fixed: 2053221
Change-Id: I62a2fcb30d19579180b3df16f33953546f94511a
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
The char device /dev/sp_kernel is used for loading SP application and
creating new channel char device, therefore might be temporary busy.
Return EBUSY error to user system call open() so it can retry.
Change-Id: I230577fe6b03ac7afd7175f72f9a3fa80db5a524
Signed-off-by: Amir Samuelov <amirs@codeaurora.org>
This fix checks the validity of dci client's process descriptor
before issuing a signal to it when subsystem restart is performed.
This fix avoids accessing cleaned-up process descriptor's fields.
CRs-Fixed: 2047235
Change-Id: Ic26977dc22c68f0a7007dd963c9273bba2a5dbfe
Signed-off-by: Gopikrishna Mogasati <gmogas@codeaurora.org>
Current handling of wil->platform_ops.suspend can lead to
bad cases, as wil_status_suspending was cleared before
wil_status_resuming was set and bus request voting was restored
too late.
In addition, bus voting in suspend should be done only at the end
of he suspend flow.
Change-Id: I8856d393c1796a2bd8fd5e07b233a5d61efc80c0
Signed-off-by: Maya Erez <merez@codeaurora.org>
Move the bus request restore to be done prior to
sending wmi_resume to the device, as the device can restore
the data path upon receiving this command.
Change-Id: Ib2442fcfea055a855bb6be65f9013bb7ad754767
Signed-off-by: Maya Erez <merez@codeaurora.org>
