Adds support for "state object" cmdstream buffers which can be
constructed once, and re-used many times. This enables the use
for CP_SET_DRAW_STATE packets on newer hardware, to lower the
CPU overhead.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Splitting code-motion out from next patch. Once we add stateobj rb's
this loop could add new entries to bos table, so it needs to be before
we set req.bos/req.nr_bos.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Add new API for reusable "state objects" which can be re-used multiple
times. Backend implementation for msm will follow. (Probably not
needed to support this for any device that uses kgsl backend, since this
is mostly useful for a5xx+.)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This will prevent any more missing `#include "config.h"` bug, at the
cost of having to recompile some files that didn't need to be when
changing build options.
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Needed for clover/OpenCL. Fortunately the kernel interface is already
in place.
Include a stub _put_iova() so mesa can tell us when it no longer needs
the buffer to be pinned. There is no kernel interface for this (yet),
but at least if we want to unpin buffers we won't need mesa changes.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
In case of a kernel that is new enough to support multiple submit-
queues, but with an adreno generation which doesn't support multiple
prioritized ringbuffers, we'd attempt to open a submit-queue with
prio=1 (medium), which is rejected by the kernel.
This could happen either w/ an older mesa (which uses fd_pipe_new())
or a newer mesa which defaults to prio=1 if no pipe context priority
flags are set.
The simple answer to fix both cases is to clamp the requested priority
according to the number of rings. This might not do exactly what you
want, if we hypothetically had 2 rings (it would result in requested
medium priority being high priority instead of low priority). But the
number of rings (for hw gen's that support this) is purely a software
construct, so the easy answer there is to have the kernel advertise at
least 3 rings if it supports more than one. There isn't really any
reason to do otherwise.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
With a new-enough kernel to support prioritized submit-queues, we can
expose priority level support to mesa. Open a submit queue associated
with the fd_pipe and pass it's id back to SUBMIT ioctl.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
`ring` cannot be non-null, so the label reduces to a simple return.
Then, there is no point initialising `ring` just to overwrite it before
anyone reads it.
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Something that valgrind spotted:
==8441== Invalid read of size 4
==8441== at 0x5DEE168: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:506)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4910F: fd5_emit_tile_gmem2mem (fd5_gmem.c:477)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
==8441== by 0x4AC8A67: MakeContextCurrent (glxcurrent.c:214)
==8441== Address 0x6f5eb1c is 204 bytes inside a block of size 240 free'd
==8441== at 0x4868F44: realloc (vg_replace_malloc.c:785)
==8441== by 0x5DEE143: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:502)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4910F: fd5_emit_tile_gmem2mem (fd5_gmem.c:477)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
==8441== Block was alloc'd at
==8441== at 0x4868F44: realloc (vg_replace_malloc.c:785)
==8441== by 0x5DEE08B: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:481)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4909F: fd5_emit_tile_gmem2mem (fd5_gmem.c:465)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
a5xx and later are 64bit devices.. make reloc's handle that. A new
public symbol is introduced to avoid silent problems with new mesa and
old libdrm (since on 64b reloc consumes two dwords).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Note: cache the last ring the bo was emitted on, to avoid excess
hashtable lookups. We do this by tracking ring seqno to avoid
problems with dangling pointers.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
First step towards supporting a single logical ringbuffer mapping to
multiple physical cmd buffers, which will enable dynamically growing
ringbuffers.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Not actually needed. It just needs to ensure that there is a
corresponding entry in the submit's cmds table.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Since they get vmap'd on the kernel side, they are a bit more costly.
Don't let them mingle with the riffraff.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If user has emit'd reloc's, and then resets or deletes the ring, we want
to drop the ref's that the ring holds to the bo's to avoid a leak.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
gallium needs to know if the kernel is new enough to support explicit
fencing, dynamically grown ringbuffers, etc.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
With a new enough drm/msm, we can let the kernel know about buffers that
are in the bo cache, so the kernel can free them under memory pressure.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
User should only see these with LIBGL_DEBUG=verbose. But in case you
are hitting issues like "handle X at index Y already on submit list"
errors from the kernel, this gives some useful visibility for debug.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It should be a less common case, but it is possible for a single bo to
be on multiple rings, for example when sharing a buffer across multiple
pipe_context's created from same pipe_screen.
So rather than completely fall over in this case, fallback to slow-path
of looping over all bo's in the ring's bo-table (but retain the fast-
path of constant-lookup for the first ring the buffer is on).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Group the parts related to building out submit ioctl into their own
sub-struct. Split out from next commit since it is just boring churn.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Abeit quite unlikely to get hit by this bug here, let just fix it.
v2: Correct conditional (do not call ioctl(DRM_IOCTL_PRIME_HANDLE_TO_FD)
when we already have the fd).
v3: Fix kgsl_pipe.c, suggested by Thierry.
Cc: freedreno@lists.freedesktop.org
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Thierry Reding <thierry.reding@gmail.com>
They are less and easier to track than the public ones. The macro
drm_public will be going away by the end of the series.
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
4c2766b (drm_mmap/drm_unmap) brought this error for every .c file that
was not #including config.h:
In file included from private.h:4:0,
from abi16.c:29:
../libdrm.h: In function 'drm_munmap':
../libdrm.h:81:4: error: size of unnamed array is negative
Signed-off-by: Rob Clark <robdclark@gmail.com>
Need to update timestamp on all ring's associated with a submit (ie.
both the binning pass and main ring). Also, make sure nr_reloc's
in particular gets cleared if the rb is reset.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Allow IB to different ringbuffer in addition to just different part of
same ringbuffer. In particular, we need to add bo's to the parent (ie.
one passed to flush) bo table, since the bo table applies to all the
cmd buffers in submit ioctl.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Otherwise build will fail, as drm/drm.h is not available.
Cc: Rob Clark <robclark@freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Rob Clark
Eric Engestrom
Eric Engestrom
Rob Clark
Emil Velikov