UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity
OpenCloudOS-Kernel/drivers/gpu/drm/etnaviv/etnaviv_cmdbuf.h

40 lines
1.0 KiB
C
Raw Normal View History

drm/etnaviv: replace license text with SPDX tags This replaces the repetitive GPL-2.0 license text in code and header files with the SPDX tags. Generated hardware headers aren't changed, as any changes there need to be done in the upstream rnndb repository. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2018-05-08 22:20:54 +08:00
/* SPDX-License-Identifier: GPL-2.0 */
drm/etnaviv: move cmdbuf de-/allocation into own file This will get more complex with the following changes, so move it into its own place. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:09:51 +08:00
/*
* Copyright (C) 2017 Etnaviv Project
*/
#ifndef __ETNAVIV_CMDBUF_H__
#define __ETNAVIV_CMDBUF_H__
#include <linux/types.h>
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
struct etnaviv_gpu;
struct etnaviv_cmdbuf_suballoc;
drm/etnaviv: add internal representation of perfmon_request Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com> Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
2017-09-24 21:15:23 +08:00
struct etnaviv_perfmon_request;
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
drm/etnaviv: move cmdbuf de-/allocation into own file This will get more complex with the following changes, so move it into its own place. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:09:51 +08:00
struct etnaviv_cmdbuf {
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
/* suballocator this cmdbuf is allocated from */
struct etnaviv_cmdbuf_suballoc *suballoc;
drm/etnaviv: move cmdbuf de-/allocation into own file This will get more complex with the following changes, so move it into its own place. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:09:51 +08:00
/* user context key, must be unique between all active users */
struct etnaviv_file_private *ctx;
/* cmdbuf properties */
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
int suballoc_offset;
drm/etnaviv: move cmdbuf de-/allocation into own file This will get more complex with the following changes, so move it into its own place. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:09:51 +08:00
void *vaddr;
u32 size;
u32 user_size;
};
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
struct etnaviv_cmdbuf_suballoc *
etnaviv_cmdbuf_suballoc_new(struct etnaviv_gpu * gpu);
void etnaviv_cmdbuf_suballoc_destroy(struct etnaviv_cmdbuf_suballoc *suballoc);
drm/etnaviv: move cmdbuf into submit object Less dynamic allocations and slims down the cmdbuf object to only the required information, as everything else is already available in the submit object. This also simplifies buffer and mappings lifetime management, as they are now exlusively attached to the submit object and not additionally to the cmdbuf. Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
2017-11-24 23:56:37 +08:00
int etnaviv_cmdbuf_init(struct etnaviv_cmdbuf_suballoc *suballoc,
struct etnaviv_cmdbuf *cmdbuf, u32 size);
drm/etnaviv: add cmdbuf suballocator There are 3 big benefits to suballocating a single big DMA buffer for command submission: 1. Avoid hammering CMA. The old way of allocating and freeing a DMA buffer for each submission was hitting some of the real slow pathes in CMA, as this allocator was not designed for a concurrent small buffers load. 2. Less TLB flushes on IOMMUv2. If a new command buffer is mapped into the GPU address space the MMU TLBs need to be flushed. By having one big buffer statically mapped to the GPU, a lot of those flushes can be avoided. 3. No funky workarounds for GC3000. The FE TLB flush on GC3000 isn't reliable. To work around that we tried to lay out the cmdbufs in the GPU address space in a way to avoid this issue. This hasn't always worked if the address space is crowded. A single statically mapped buffer avoids the erratum completely. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:29:57 +08:00
void etnaviv_cmdbuf_free(struct etnaviv_cmdbuf *cmdbuf);
drm/etnaviv: wire up iova handling in new cmdbuf abstraction Don't call the IOMMU directly, but go through the new cmdbuf abstraction. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:52:44 +08:00
u32 etnaviv_cmdbuf_get_va(struct etnaviv_cmdbuf *buf);
drm/etnaviv: get cmdbuf physical address through the cmdbuf abstraction Don't allow IOMMUv2 to peek directly into the cmdbuf, but get the needed PA through a dedicated function. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-17 00:00:08 +08:00
dma_addr_t etnaviv_cmdbuf_get_pa(struct etnaviv_cmdbuf *buf);
drm/etnaviv: wire up iova handling in new cmdbuf abstraction Don't call the IOMMU directly, but go through the new cmdbuf abstraction. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:52:44 +08:00
drm/etnaviv: move cmdbuf de-/allocation into own file This will get more complex with the following changes, so move it into its own place. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
2017-01-16 23:09:51 +08:00
#endif /* __ETNAVIV_CMDBUF_H__ */