Age | Commit message (Collapse) | Author |
|
|
|
This works - no need for the message any more.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
Tidy up a few lines, and remove debug lines which would slow us down.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This just alters some formatting and removes old code.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
Conflicts:
drivers/mfd/glamo/glamo-cmdq.c
drivers/mfd/glamo/glamo-cmdq.h
drivers/mfd/glamo/glamo-drm-drv.c
drivers/mfd/glamo/glamo-drm-private.h
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
Doesn't work yet...
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
That wasn't the right way to go about this. And it didn't compile.
This reverts commit d8e4b9ff4dc8bf57e53fa174f977c8fb00ec4e90.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This makes the GEM "info" ioctl produce an object handle which can be used to refer
to the front buffer (i.e. screen pixmap) in Xorg or otherwise. This is unfortunately
necessary since this memory is allocated in the kernel by the framebuffer driver, and
is not available to DRM until we move to KMS.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
The rest_size was treated like the count of words instead of
bytes. This code also switches to memcpy_toio.
Signed-off-by: Andreas Pokorny <andreas.pokorny@gmail.com>
|
|
Removed the isr and everything interrupt related. IRQ is not
needed for basic operation on the glamo cmdq. This code directly
updates the write pointer of the queue without any waits.
Signed-off-by: Andreas Pokorny <andreas.pokorny@gmail.com>
|
|
|
|
|
|
The code is still not functional, only the first set of commands get
executed.
Signed-off-by: Andreas Pokorny <andreas.pokorny@gmail.com>
|
|
|
|
This is the first step towards the irq based update of the write
pointer. With this revision the previous wait, and direct update is
deactivated.
Signed-off-by: Andreas Pokorny <andreas.pokorny@gmail.com>
|
|
This reuinites the "framebuffer" and "work" VRAM areas. With KMS, both roles will be done
by one pool of memory.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This allows command sequences to make use of buffer objects by providing
their GEM handles.
It also reorganises the code a bit, and adds a lot of missing security stuff.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This removes a lot of debugging. We can't afford to splurge out a
page of printk()s on every command submission.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This just moves the engine enabling calls around a bit.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This adds [glamo-drm] before the debugging messages.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This still doesn't quite work...
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This allows glamo_cmdq_wait() to timeout.
We will probably need to substitute this for something better later on, but for the
time being I wanted to avoid a deadlock if the command queue never cleared.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This fixes the licence text at the top of glamo-cmdq.c, noting that code from
Xorg has been incorporated.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|
|
This adds handling of Glamo's ring buffer at the kernel level, to be accessed
via ioctl DRM_GLAMO_CMDBUF.
Signed-off-by: Thomas White <taw@bitwiz.org.uk>
|