diff options
Diffstat (limited to 'drivers/gpu/drm/etnaviv')
| -rw-r--r-- | drivers/gpu/drm/etnaviv/common.xml.h | 59 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_buffer.c | 219 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_drv.c | 1 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_drv.h | 5 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_dump.c | 6 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gem.c | 86 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gem.h | 18 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c | 10 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c | 36 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 251 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_gpu.h | 13 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/etnaviv_mmu.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/state_3d.xml.h | 9 | ||||
| -rw-r--r-- | drivers/gpu/drm/etnaviv/state_hi.xml.h | 26 |
14 files changed, 493 insertions, 248 deletions
diff --git a/drivers/gpu/drm/etnaviv/common.xml.h b/drivers/gpu/drm/etnaviv/common.xml.h index 9e585d51fb78..e881482b5971 100644 --- a/drivers/gpu/drm/etnaviv/common.xml.h +++ b/drivers/gpu/drm/etnaviv/common.xml.h @@ -8,8 +8,8 @@ http://0x04.net/cgit/index.cgi/rules-ng-ng git clone git://0x04.net/rules-ng-ng The rules-ng-ng source files this header was generated from are: -- state_vg.xml ( 5973 bytes, from 2015-03-25 11:26:01) -- common.xml ( 18437 bytes, from 2015-03-25 11:27:41) +- state_hi.xml ( 24309 bytes, from 2015-12-12 09:02:53) +- common.xml ( 18379 bytes, from 2015-12-12 09:02:53) Copyright (C) 2015 */ @@ -30,15 +30,19 @@ Copyright (C) 2015 #define ENDIAN_MODE_NO_SWAP 0x00000000 #define ENDIAN_MODE_SWAP_16 0x00000001 #define ENDIAN_MODE_SWAP_32 0x00000002 +#define chipModel_GC200 0x00000200 #define chipModel_GC300 0x00000300 #define chipModel_GC320 0x00000320 +#define chipModel_GC328 0x00000328 #define chipModel_GC350 0x00000350 #define chipModel_GC355 0x00000355 #define chipModel_GC400 0x00000400 #define chipModel_GC410 0x00000410 #define chipModel_GC420 0x00000420 +#define chipModel_GC428 0x00000428 #define chipModel_GC450 0x00000450 #define chipModel_GC500 0x00000500 +#define chipModel_GC520 0x00000520 #define chipModel_GC530 0x00000530 #define chipModel_GC600 0x00000600 #define chipModel_GC700 0x00000700 @@ -46,9 +50,16 @@ Copyright (C) 2015 #define chipModel_GC860 0x00000860 #define chipModel_GC880 0x00000880 #define chipModel_GC1000 0x00001000 +#define chipModel_GC1500 0x00001500 #define chipModel_GC2000 0x00002000 #define chipModel_GC2100 0x00002100 +#define chipModel_GC2200 0x00002200 +#define chipModel_GC2500 0x00002500 +#define chipModel_GC3000 0x00003000 #define chipModel_GC4000 0x00004000 +#define chipModel_GC5000 0x00005000 +#define chipModel_GC5200 0x00005200 +#define chipModel_GC6400 0x00006400 #define RGBA_BITS_R 0x00000001 #define RGBA_BITS_G 0x00000002 #define RGBA_BITS_B 0x00000004 @@ -160,7 +171,7 @@ Copyright (C) 2015 #define chipMinorFeatures2_UNK8 0x00000100 #define chipMinorFeatures2_UNK9 0x00000200 #define chipMinorFeatures2_UNK10 0x00000400 -#define chipMinorFeatures2_SAMPLERBASE_16 0x00000800 +#define chipMinorFeatures2_HALTI1 0x00000800 #define chipMinorFeatures2_UNK12 0x00001000 #define chipMinorFeatures2_UNK13 0x00002000 #define chipMinorFeatures2_UNK14 0x00004000 @@ -189,7 +200,7 @@ Copyright (C) 2015 #define chipMinorFeatures3_UNK5 0x00000020 #define chipMinorFeatures3_UNK6 0x00000040 #define chipMinorFeatures3_UNK7 0x00000080 -#define chipMinorFeatures3_UNK8 0x00000100 +#define chipMinorFeatures3_FAST_MSAA 0x00000100 #define chipMinorFeatures3_UNK9 0x00000200 #define chipMinorFeatures3_BUG_FIXES10 0x00000400 #define chipMinorFeatures3_UNK11 0x00000800 @@ -199,7 +210,7 @@ Copyright (C) 2015 #define chipMinorFeatures3_UNK15 0x00008000 #define chipMinorFeatures3_UNK16 0x00010000 #define chipMinorFeatures3_UNK17 0x00020000 -#define chipMinorFeatures3_UNK18 0x00040000 +#define chipMinorFeatures3_ACE 0x00040000 #define chipMinorFeatures3_UNK19 0x00080000 #define chipMinorFeatures3_UNK20 0x00100000 #define chipMinorFeatures3_UNK21 0x00200000 @@ -207,7 +218,7 @@ Copyright (C) 2015 #define chipMinorFeatures3_UNK23 0x00800000 #define chipMinorFeatures3_UNK24 0x01000000 #define chipMinorFeatures3_UNK25 0x02000000 -#define chipMinorFeatures3_UNK26 0x04000000 +#define chipMinorFeatures3_NEW_HZ 0x04000000 #define chipMinorFeatures3_UNK27 0x08000000 #define chipMinorFeatures3_UNK28 0x10000000 #define chipMinorFeatures3_UNK29 0x20000000 @@ -229,9 +240,9 @@ Copyright (C) 2015 #define chipMinorFeatures4_UNK13 0x00002000 #define chipMinorFeatures4_UNK14 0x00004000 #define chipMinorFeatures4_UNK15 0x00008000 -#define chipMinorFeatures4_UNK16 0x00010000 +#define chipMinorFeatures4_HALTI2 0x00010000 #define chipMinorFeatures4_UNK17 0x00020000 -#define chipMinorFeatures4_UNK18 0x00040000 +#define chipMinorFeatures4_SMALL_MSAA 0x00040000 #define chipMinorFeatures4_UNK19 0x00080000 #define chipMinorFeatures4_UNK20 0x00100000 #define chipMinorFeatures4_UNK21 0x00200000 @@ -245,5 +256,37 @@ Copyright (C) 2015 #define chipMinorFeatures4_UNK29 0x20000000 #define chipMinorFeatures4_UNK30 0x40000000 #define chipMinorFeatures4_UNK31 0x80000000 +#define chipMinorFeatures5_UNK0 0x00000001 +#define chipMinorFeatures5_UNK1 0x00000002 +#define chipMinorFeatures5_UNK2 0x00000004 +#define chipMinorFeatures5_UNK3 0x00000008 +#define chipMinorFeatures5_UNK4 0x00000010 +#define chipMinorFeatures5_UNK5 0x00000020 +#define chipMinorFeatures5_UNK6 0x00000040 +#define chipMinorFeatures5_UNK7 0x00000080 +#define chipMinorFeatures5_UNK8 0x00000100 +#define chipMinorFeatures5_HALTI3 0x00000200 +#define chipMinorFeatures5_UNK10 0x00000400 +#define chipMinorFeatures5_UNK11 0x00000800 +#define chipMinorFeatures5_UNK12 0x00001000 +#define chipMinorFeatures5_UNK13 0x00002000 +#define chipMinorFeatures5_UNK14 0x00004000 +#define chipMinorFeatures5_UNK15 0x00008000 +#define chipMinorFeatures5_UNK16 0x00010000 +#define chipMinorFeatures5_UNK17 0x00020000 +#define chipMinorFeatures5_UNK18 0x00040000 +#define chipMinorFeatures5_UNK19 0x00080000 +#define chipMinorFeatures5_UNK20 0x00100000 +#define chipMinorFeatures5_UNK21 0x00200000 +#define chipMinorFeatures5_UNK22 0x00400000 +#define chipMinorFeatures5_UNK23 0x00800000 +#define chipMinorFeatures5_UNK24 0x01000000 +#define chipMinorFeatures5_UNK25 0x02000000 +#define chipMinorFeatures5_UNK26 0x04000000 +#define chipMinorFeatures5_UNK27 0x08000000 +#define chipMinorFeatures5_UNK28 0x10000000 +#define chipMinorFeatures5_UNK29 0x20000000 +#define chipMinorFeatures5_UNK30 0x40000000 +#define chipMinorFeatures5_UNK31 0x80000000 #endif /* COMMON_XML */ diff --git a/drivers/gpu/drm/etnaviv/etnaviv_buffer.c b/drivers/gpu/drm/etnaviv/etnaviv_buffer.c index 332c55ebba6d..d8d556457427 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_buffer.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_buffer.c @@ -21,6 +21,7 @@ #include "common.xml.h" #include "state.xml.h" +#include "state_3d.xml.h" #include "cmdstream.xml.h" /* @@ -85,10 +86,17 @@ static inline void CMD_STALL(struct etnaviv_cmdbuf *buffer, OUT(buffer, VIV_FE_STALL_TOKEN_FROM(from) | VIV_FE_STALL_TOKEN_TO(to)); } -static void etnaviv_cmd_select_pipe(struct etnaviv_cmdbuf *buffer, u8 pipe) +static inline void CMD_SEM(struct etnaviv_cmdbuf *buffer, u32 from, u32 to) { - u32 flush; - u32 stall; + CMD_LOAD_STATE(buffer, VIVS_GL_SEMAPHORE_TOKEN, + VIVS_GL_SEMAPHORE_TOKEN_FROM(from) | + VIVS_GL_SEMAPHORE_TOKEN_TO(to)); +} + +static void etnaviv_cmd_select_pipe(struct etnaviv_gpu *gpu, + struct etnaviv_cmdbuf *buffer, u8 pipe) +{ + u32 flush = 0; /* * This assumes that if we're switching to 2D, we're switching @@ -96,17 +104,13 @@ static void etnaviv_cmd_select_pipe(struct etnaviv_cmdbuf *buffer, u8 pipe) * the 2D core, we need to flush the 3D depth and color caches, * otherwise we need to flush the 2D pixel engine cache. */ - if (pipe == ETNA_PIPE_2D) - flush = VIVS_GL_FLUSH_CACHE_DEPTH | VIVS_GL_FLUSH_CACHE_COLOR; - else + if (gpu->exec_state == ETNA_PIPE_2D) flush = VIVS_GL_FLUSH_CACHE_PE2D; - - stall = VIVS_GL_SEMAPHORE_TOKEN_FROM(SYNC_RECIPIENT_FE) | - VIVS_GL_SEMAPHORE_TOKEN_TO(SYNC_RECIPIENT_PE); + else if (gpu->exec_state == ETNA_PIPE_3D) + flush = VIVS_GL_FLUSH_CACHE_DEPTH | VIVS_GL_FLUSH_CACHE_COLOR; CMD_LOAD_STATE(buffer, VIVS_GL_FLUSH_CACHE, flush); - CMD_LOAD_STATE(buffer, VIVS_GL_SEMAPHORE_TOKEN, stall); - + CMD_SEM(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); CMD_STALL(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); CMD_LOAD_STATE(buffer, VIVS_GL_PIPE_SELECT, @@ -131,6 +135,36 @@ static void etnaviv_buffer_dump(struct etnaviv_gpu *gpu, ptr, len * 4, 0); } +/* + * Safely replace the WAIT of a waitlink with a new command and argument. + * The GPU may be executing this WAIT while we're modifying it, so we have + * to write it in a specific order to avoid the GPU branching to somewhere + * else. 'wl_offset' is the offset to the first byte of the WAIT command. + */ +static void etnaviv_buffer_replace_wait(struct etnaviv_cmdbuf *buffer, + unsigned int wl_offset, u32 cmd, u32 arg) +{ + u32 *lw = buffer->vaddr + wl_offset; + + lw[1] = arg; + mb(); + lw[0] = cmd; + mb(); +} + +/* + * Ensure that there is space in the command buffer to contiguously write + * 'cmd_dwords' 64-bit words into the buffer, wrapping if necessary. + */ +static u32 etnaviv_buffer_reserve(struct etnaviv_gpu *gpu, + struct etnaviv_cmdbuf *buffer, unsigned int cmd_dwords) +{ + if (buffer->user_size + cmd_dwords * sizeof(u64) > buffer->size) + buffer->user_size = 0; + + return gpu_va(gpu, buffer) + buffer->user_size; +} + u16 etnaviv_buffer_init(struct etnaviv_gpu *gpu) { struct etnaviv_cmdbuf *buffer = gpu->buffer; @@ -147,81 +181,79 @@ u16 etnaviv_buffer_init(struct etnaviv_gpu *gpu) void etnaviv_buffer_end(struct etnaviv_gpu *gpu) { struct etnaviv_cmdbuf *buffer = gpu->buffer; + unsigned int waitlink_offset = buffer->user_size - 16; + u32 link_target, flush = 0; - /* Replace the last WAIT with an END */ - buffer->user_size -= 16; - - CMD_END(buffer); - mb(); + if (gpu->exec_state == ETNA_PIPE_2D) + flush = VIVS_GL_FLUSH_CACHE_PE2D; + else if (gpu->exec_state == ETNA_PIPE_3D) + flush = VIVS_GL_FLUSH_CACHE_DEPTH | + VIVS_GL_FLUSH_CACHE_COLOR | + VIVS_GL_FLUSH_CACHE_TEXTURE | + VIVS_GL_FLUSH_CACHE_TEXTUREVS | + VIVS_GL_FLUSH_CACHE_SHADER_L2; + + if (flush) { + unsigned int dwords = 7; + + link_target = etnaviv_buffer_reserve(gpu, buffer, dwords); + + CMD_SEM(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); + CMD_STALL(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); + CMD_LOAD_STATE(buffer, VIVS_GL_FLUSH_CACHE, flush); + if (gpu->exec_state == ETNA_PIPE_3D) + CMD_LOAD_STATE(buffer, VIVS_TS_FLUSH_CACHE, + VIVS_TS_FLUSH_CACHE_FLUSH); + CMD_SEM(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); + CMD_STALL(buffer, SYNC_RECIPIENT_FE, SYNC_RECIPIENT_PE); + CMD_END(buffer); + + etnaviv_buffer_replace_wait(buffer, waitlink_offset, + VIV_FE_LINK_HEADER_OP_LINK | + VIV_FE_LINK_HEADER_PREFETCH(dwords), + link_target); + } else { + /* Replace the last link-wait with an "END" command */ + etnaviv_buffer_replace_wait(buffer, waitlink_offset, + VIV_FE_END_HEADER_OP_END, 0); + } } +/* Append a command buffer to the ring buffer. */ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, unsigned int event, struct etnaviv_cmdbuf *cmdbuf) { struct etnaviv_cmdbuf *buffer = gpu->buffer; - u32 *lw = buffer->vaddr + buffer->user_size - 16; - u32 back, link_target, link_size, reserve_size, extra_size = 0; + unsigned int waitlink_offset = buffer->user_size - 16; + u32 return_target, return_dwords; + u32 link_target, link_dwords; if (drm_debug & DRM_UT_DRIVER) etnaviv_buffer_dump(gpu, buffer, 0, 0x50); + link_target = gpu_va(gpu, cmdbuf); + link_dwords = cmdbuf->size / 8; + /* - * If we need to flush the MMU prior to submitting this buffer, we - * will need to append a mmu flush load state, followed by a new + * If we need maintanence prior to submitting this buffer, we will + * need to append a mmu flush load state, followed by a new * link to this buffer - a total of four additional words. */ if (gpu->mmu->need_flush || gpu->switch_context) { + u32 target, extra_dwords; + /* link command */ - extra_size += 2; + extra_dwords = 1; + /* flush command */ if (gpu->mmu->need_flush) - extra_size += 2; + extra_dwords += 1; + /* pipe switch commands */ if (gpu->switch_context) - extra_size += 8; - } + extra_dwords += 4; - reserve_size = (6 + extra_size) * 4; - - /* - * if we are going to completely overflow the buffer, we need to wrap. - */ - if (buffer->user_size + reserve_size > buffer->size) - buffer->user_size = 0; - - /* save offset back into main buffer */ - back = buffer->user_size + reserve_size - 6 * 4; - link_target = gpu_va(gpu, buffer) + buffer->user_size; - link_size = 6; - - /* Skip over any extra instructions */ - link_target += extra_size * sizeof(u32); - - if (drm_debug & DRM_UT_DRIVER) - pr_info("stream link to 0x%08x @ 0x%08x %p\n", - link_target, gpu_va(gpu, cmdbuf), cmdbuf->vaddr); - - /* jump back from cmd to main buffer */ - CMD_LINK(cmdbuf, link_size, link_target); - - link_target = gpu_va(gpu, cmdbuf); - link_size = cmdbuf->size / 8; - - - - if (drm_debug & DRM_UT_DRIVER) { - print_hex_dump(KERN_INFO, "cmd ", DUMP_PREFIX_OFFSET, 16, 4, - cmdbuf->vaddr, cmdbuf->size, 0); - - pr_info("link op: %p\n", lw); - pr_info("link addr: %p\n", lw + 1); - pr_info("addr: 0x%08x\n", link_target); - pr_info("back: 0x%08x\n", gpu_va(gpu, buffer) + back); - pr_info("event: %d\n", event); - } - - if (gpu->mmu->need_flush || gpu->switch_context) { - u32 new_target = gpu_va(gpu, buffer) + buffer->user_size; + target = etnaviv_buffer_reserve(gpu, buffer, extra_dwords); if (gpu->mmu->need_flush) { /* Add the MMU flush */ @@ -236,32 +268,59 @@ void etnaviv_buffer_queue(struct etnaviv_gpu *gpu, unsigned int event, } if (gpu->switch_context) { - etnaviv_cmd_select_pipe(buffer, cmdbuf->exec_state); + etnaviv_cmd_select_pipe(gpu, buffer, cmdbuf->exec_state); + gpu->exec_state = cmdbuf->exec_state; gpu->switch_context = false; } - /* And the link to the first buffer */ - CMD_LINK(buffer, link_size, link_target); + /* And the link to the submitted buffer */ + CMD_LINK(buffer, link_dwords, link_target); /* Update the link target to point to above instructions */ - link_target = new_target; - link_size = extra_size; + link_target = target; + link_dwords = extra_dwords; } - /* trigger event */ + /* + * Append a LINK to the submitted command buffer to return to + * the ring buffer. return_target is the ring target address. + * We need three dwords: event, wait, link. + */ + return_dwords = 3; + return_target = etnaviv_buffer_reserve(gpu, buffer, return_dwords); + CMD_LINK(cmdbuf, return_dwords, return_target); + + /* + * Append event, wait and link pointing back to the wait + * command to the ring buffer. + */ CMD_LOAD_STATE(buffer, VIVS_GL_EVENT, VIVS_GL_EVENT_EVENT_ID(event) | VIVS_GL_EVENT_FROM_PE); - - /* append WAIT/LINK to main buffer */ CMD_WAIT(buffer); - CMD_LINK(buffer, 2, gpu_va(gpu, buffer) + (buffer->user_size - 4)); + CMD_LINK(buffer, 2, return_target + 8); - /* Change WAIT into a LINK command; write the address first. */ - *(lw + 1) = link_target; - mb(); - *(lw) = VIV_FE_LINK_HEADER_OP_LINK | - VIV_FE_LINK_HEADER_PREFETCH(link_size); - mb(); + if (drm_debug & DRM_UT_DRIVER) + pr_info("stream link to 0x%08x @ 0x%08x %p\n", + return_target, gpu_va(gpu, cmdbuf), cmdbuf->vaddr); + + if (drm_debug & DRM_UT_DRIVER) { + print_hex_dump(KERN_INFO, "cmd ", DUMP_PREFIX_OFFSET, 16, 4, + cmdbuf->vaddr, cmdbuf->size, 0); + + pr_info("link op: %p\n", buffer->vaddr + waitlink_offset); + pr_info("addr: 0x%08x\n", link_target); + pr_info("back: 0x%08x\n", return_target); + pr_info("event: %d\n", event); + } + + /* + * Kick off the submitted command by replacing the previous + * WAIT with a link to the address in the ring buffer. + */ + etnaviv_buffer_replace_wait(buffer, waitlink_offset, + VIV_FE_LINK_HEADER_OP_LINK | + VIV_FE_LINK_HEADER_PREFETCH(link_dwords), + link_target); if (drm_debug & DRM_UT_DRIVER) etnaviv_buffer_dump(gpu, buffer, 0, 0x50); diff --git a/drivers/gpu/drm/etnaviv/etnaviv_drv.c b/drivers/gpu/drm/etnaviv/etnaviv_drv.c index 5c89ebb52fd2..e8858985f01e 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_drv.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_drv.c @@ -668,7 +668,6 @@ static struct platform_driver etnaviv_platform_driver = { .probe = etnaviv_pdev_probe, .remove = etnaviv_pdev_remove, .driver = { - .owner = THIS_MODULE, .name = "etnaviv", .of_match_table = dt_match, }, diff --git a/drivers/gpu/drm/etnaviv/etnaviv_drv.h b/drivers/gpu/drm/etnaviv/etnaviv_drv.h index d6bd438bd5be..115c5bc6d7c8 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_drv.h +++ b/drivers/gpu/drm/etnaviv/etnaviv_drv.h @@ -75,9 +75,6 @@ int etnaviv_ioctl_gem_submit(struct drm_device *dev, void *data, int etnaviv_gem_mmap(struct file *filp, struct vm_area_struct *vma); int etnaviv_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf); int etnaviv_gem_mmap_offset(struct drm_gem_object *obj, u64 *offset); -int etnaviv_gem_get_iova(struct etnaviv_gpu *gpu, - struct drm_gem_object *obj, u32 *iova); -void etnaviv_gem_put_iova(struct etnaviv_gpu *gpu, struct drm_gem_object *obj); struct sg_table *etnaviv_gem_prime_get_sg_table(struct drm_gem_object *obj); void *etnaviv_gem_prime_vmap(struct drm_gem_object *obj); void etnaviv_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); @@ -85,7 +82,7 @@ struct drm_gem_object *etnaviv_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sg); int etnaviv_gem_prime_pin(struct drm_gem_object *obj); void etnaviv_gem_prime_unpin(struct drm_gem_object *obj); -void *etnaviv_gem_vaddr(struct drm_gem_object *obj); +void *etnaviv_gem_vmap(struct drm_gem_object *obj); int etnaviv_gem_cpu_prep(struct drm_gem_object *obj, u32 op, struct timespec *timeout); int etnaviv_gem_cpu_fini(struct drm_gem_object *obj); diff --git a/drivers/gpu/drm/etnaviv/etnaviv_dump.c b/drivers/gpu/drm/etnaviv/etnaviv_dump.c index bf8fa859e8be..4a29eeadbf1e 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_dump.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_dump.c @@ -201,7 +201,9 @@ void etnaviv_core_dump(struct etnaviv_gpu *gpu) obj = vram->object; + mutex_lock(&obj->lock); pages = etnaviv_gem_get_pages(obj); + mutex_unlock(&obj->lock); if (pages) { int j; @@ -213,8 +215,8 @@ void etnaviv_core_dump(struct etnaviv_gpu *gpu) iter.hdr->iova = cpu_to_le64(vram->iova); - vaddr = etnaviv_gem_vaddr(&obj->base); - if (vaddr && !IS_ERR(vaddr)) + vaddr = etnaviv_gem_vmap(&obj->base); + if (vaddr) memcpy(iter.data, vaddr, obj->base.size); etnaviv_core_dump_header(&iter, ETDUMP_BUF_BO, iter.data + diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.c b/drivers/gpu/drm/etnaviv/etnaviv_gem.c index 9f77c3b94cc6..281c6eca20a8 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.c @@ -260,8 +260,32 @@ etnaviv_gem_get_vram_mapping(struct etnaviv_gem_object *obj, return NULL; } -int etnaviv_gem_get_iova(struct etnaviv_gpu *gpu, - struct drm_gem_object *obj, u32 *iova) +void etnaviv_gem_mapping_reference(struct etnaviv_vram_mapping *mapping) +{ + struct etnaviv_gem_object *etnaviv_obj = mapping->object; + + drm_gem_object_reference(&etnaviv_obj->base); + + mutex_lock(&etnaviv_obj->lock); + WARN_ON(mapping->use == 0); + mapping->use += 1; + mutex_unlock(&etnaviv_obj->lock); +} + +void etnaviv_gem_mapping_unreference(struct etnaviv_vram_mapping *mapping) +{ + struct etnaviv_gem_object *etnaviv_obj = mapping->object; + + mutex_lock(&etnaviv_obj->lock); + WARN_ON(mapping->use == 0); + mapping->use -= 1; + mutex_unlock(&etnaviv_obj->lock); + + drm_gem_object_unreference_unlocked(&etnaviv_obj->base); +} + +struct etnaviv_vram_mapping *etnaviv_gem_mapping_get( + struct drm_gem_object *obj, struct etnaviv_gpu *gpu) { struct etnaviv_gem_object *etnaviv_obj = to_etnaviv_bo(obj); struct etnaviv_vram_mapping *mapping; @@ -329,47 +353,45 @@ int etnaviv_gem_get_iova(struct etnaviv_gpu *gpu, out: mutex_unlock(&etnaviv_obj->lock); - if (!ret) { - /* Take a reference on the object */ - drm_gem_object_reference(obj); - *iova = mapping->iova; - } + if (ret) + return ERR_PTR(ret); - return ret; + /* Take a reference on the object */ + drm_gem_object_reference(obj); + return mapping; } -void etnaviv_gem_put_iova(struct etnaviv_gpu *gpu, struct drm_gem_object *obj) +void *etnaviv_gem_vmap(struct drm_gem_object *obj) { struct etnaviv_gem_object *etnaviv_obj = to_etnaviv_bo(obj); - struct etnaviv_vram_mapping *mapping; - mutex_lock(&etnaviv_obj->lock); - mapping = etnaviv_gem_get_vram_mapping(etnaviv_obj, gpu->mmu); + if (etnaviv_obj->vaddr) + return etnaviv_obj->vaddr; - WARN_ON(mapping->use == 0); - mapping->use -= 1; + mutex_lock(&etnaviv_obj->lock); + /* + * Need to check again, as we might have raced with another thread + * while waiting for the mutex. + */ + if (!etnaviv_obj->vaddr) + etnaviv_obj->vaddr = etnaviv_obj->ops->vmap(etnaviv_obj); mutex_unlock(&etnaviv_obj->lock); - drm_gem_object_unreference_unlocked(obj); + return etnaviv_obj->vaddr; } -void *etnaviv_gem_vaddr(struct drm_gem_object *obj) +static void *etnaviv_gem_vmap_impl(struct etnaviv_gem_object *obj) { - struct etnaviv_gem_object *etnaviv_obj = to_etnaviv_bo(obj); - - mutex_lock(&etnaviv_obj->lock); - if (!etnaviv_obj->vaddr) { - struct page **pages = etnaviv_gem_get_pages(etnaviv_obj); + struct page **pages; - if (IS_ERR(pages)) - return ERR_CAST(pages); + lockdep_assert_held(&obj->lock); - etnaviv_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT, - VM_MAP, pgprot_writecombine(PAGE_KERNEL)); - } - mutex_unlock(&etnaviv_obj->lock); + pages = etnaviv_gem_get_pages(obj); + if (IS_ERR(pages)) + return NULL; - return etnaviv_obj->vaddr; + return vmap(pages, obj->base.size >> PAGE_SHIFT, + VM_MAP, pgprot_writecombine(PAGE_KERNEL)); } static inline enum dma_data_direction etnaviv_op_to_dma_dir(u32 op) @@ -522,6 +544,7 @@ static void etnaviv_gem_shmem_release(struct etnaviv_gem_object *etnaviv_obj) static const struct etnaviv_gem_ops etnaviv_gem_shmem_ops = { .get_pages = etnaviv_gem_shmem_get_pages, .release = etnaviv_gem_shmem_release, + .vmap = etnaviv_gem_vmap_impl, }; void etnaviv_gem_free_object(struct drm_gem_object *obj) @@ -738,9 +761,9 @@ static struct page **etnaviv_gem_userptr_do_get_pages( down_read(&mm->mmap_sem); while (pinned < npages) { - ret = get_user_pages(task, mm, ptr, npages - pinned, - !etnaviv_obj->userptr.ro, 0, - pvec + pinned, NULL); + ret = get_user_pages_remote(task, mm, ptr, npages - pinned, + !etnaviv_obj->userptr.ro, 0, + pvec + pinned, NULL); if (ret < 0) break; @@ -866,6 +889,7 @@ static void etnaviv_gem_userptr_release(struct etnaviv_gem_object *etnaviv_obj) static const struct etnaviv_gem_ops etnaviv_gem_userptr_ops = { .get_pages = etnaviv_gem_userptr_get_pages, .release = etnaviv_gem_userptr_release, + .vmap = etnaviv_gem_vmap_impl, }; int etnaviv_gem_new_userptr(struct drm_device *dev, struct drm_file *file, diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.h b/drivers/gpu/drm/etnaviv/etnaviv_gem.h index a300b4b3d545..02665d8c10ee 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem.h +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.h @@ -78,6 +78,7 @@ struct etnaviv_gem_object *to_etnaviv_bo(struct drm_gem_object *obj) struct etnaviv_gem_ops { int (*get_pages)(struct etnaviv_gem_object *); void (*release)(struct etnaviv_gem_object *); + void *(*vmap)(struct etnaviv_gem_object *); }; static inline bool is_active(struct etnaviv_gem_object *etnaviv_obj) @@ -87,6 +88,12 @@ static inline bool is_active(struct etnaviv_gem_object *etnaviv_obj) #define MAX_CMDS 4 +struct etnaviv_gem_submit_bo { + u32 flags; + struct etnaviv_gem_object *obj; + struct etnaviv_vram_mapping *mapping; +}; + /* Created per submit-ioctl, to track bo's and cmdstream bufs, etc, * associated with the cmdstream submission for synchronization (and * make it easier to unwind when things go wrong, etc). This only @@ -98,11 +105,7 @@ struct etnaviv_gem_submit { struct ww_acquire_ctx ticket; u32 fence; unsigned int nr_bos; - struct { - u32 flags; - struct etnaviv_gem_object *obj; - u32 iova; - } bos[0]; + struct etnaviv_gem_submit_bo bos[0]; }; int etnaviv_gem_wait_bo(struct etnaviv_gpu *gpu, struct drm_gem_object *obj, @@ -114,4 +117,9 @@ int etnaviv_gem_obj_add(struct drm_device *dev, struct drm_gem_object *obj); struct page **etnaviv_gem_get_pages(struct etnaviv_gem_object *obj); void etnaviv_gem_put_pages(struct etnaviv_gem_object *obj); +struct etnaviv_vram_mapping *etnaviv_gem_mapping_get( + struct drm_gem_object *obj, struct etnaviv_gpu *gpu); +void etnaviv_gem_mapping_reference(struct etnaviv_vram_mapping *mapping); +void etnaviv_gem_mapping_unreference(struct etnaviv_vram_mapping *mapping); + #endif /* __ETNAVIV_GEM_H__ */ diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c b/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c index e94db4f95770..4e67395f5fa1 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem_prime.c @@ -31,7 +31,7 @@ struct sg_table *etnaviv_gem_prime_get_sg_table(struct drm_gem_object *obj) void *etnaviv_gem_prime_vmap(struct drm_gem_object *obj) { - return etnaviv_gem_vaddr(obj); + return etnaviv_gem_vmap(obj); } void etnaviv_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) @@ -77,9 +77,17 @@ static void etnaviv_gem_prime_release(struct etnaviv_gem_object *etnaviv_obj) drm_prime_gem_destroy(&etnaviv_obj->base, etnaviv_obj->sgt); } +static void *etnaviv_gem_prime_vmap_impl(struct etnaviv_gem_object *etnaviv_obj) +{ + lockdep_assert_held(&etnaviv_obj->lock); + + return dma_buf_vmap(etnaviv_obj->base.import_attach->dmabuf); +} + static const struct etnaviv_gem_ops etnaviv_gem_prime_ops = { /* .get_pages should never be called */ .release = etnaviv_gem_prime_release, + .vmap = etnaviv_gem_prime_vmap_impl, }; struct drm_gem_object *etnaviv_gem_prime_import_sg_table(struct drm_device *dev, diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c b/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c index 1aba01a999df..236ada93df53 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c @@ -187,12 +187,10 @@ static void submit_unpin_objects(struct etnaviv_gem_submit *submit) int i; for (i = 0; i < submit->nr_bos; i++) { - struct etnaviv_gem_object *etnaviv_obj = submit->bos[i].obj; - if (submit->bos[i].flags & BO_PINNED) - etnaviv_gem_put_iova(submit->gpu, &etnaviv_obj->base); + etnaviv_gem_mapping_unreference(submit->bos[i].mapping); - submit->bos[i].iova = 0; + submit->bos[i].mapping = NULL; submit->bos[i].flags &= ~BO_PINNED; } } @@ -203,22 +201,24 @@ static int submit_pin_objects(struct etnaviv_gem_submit *submit) for (i = 0; i < submit->nr_bos; i++) { struct etnaviv_gem_object *etnaviv_obj = submit->bos[i].obj; - u32 iova; + struct etnaviv_vram_mapping *mapping; - ret = etnaviv_gem_get_iova(submit->gpu, &etnaviv_obj->base, - &iova); - if (ret) + mapping = etnaviv_gem_mapping_get(&etnaviv_obj->base, + submit->gpu); + if (IS_ERR(mapping)) { + ret = PTR_ERR(mapping); break; + } submit->bos[i].flags |= BO_PINNED; - submit->bos[i].iova = iova; + submit->bos[i].mapping = mapping; } return ret; } static int submit_bo(struct etnaviv_gem_submit *submit, u32 idx, - struct etnaviv_gem_object **obj, u32 *iova) + struct etnaviv_gem_submit_bo **bo) { if (idx >= submit->nr_bos) { DRM_ERROR("invalid buffer index: %u (out of %u)\n", @@ -226,10 +226,7 @@ static int submit_bo(struct etnaviv_gem_submit *submit, u32 idx, return -EINVAL; } - if (obj) - *obj = submit->bos[idx].obj; - if (iova) - *iova = submit->bos[idx].iova; + *bo = &submit->bos[idx]; return 0; } @@ -245,8 +242,8 @@ static int submit_reloc(struct etnaviv_gem_submit *submit, void *stream, for (i = 0; i < nr_relocs; i++) { const struct drm_etnaviv_gem_submit_reloc *r = relocs + i; - struct etnaviv_gem_object *bobj; - u32 iova, off; + struct etnaviv_gem_submit_bo *bo; + u32 off; if (unlikely(r->flags)) { DRM_ERROR("invalid reloc flags\n"); @@ -268,17 +265,16 @@ static int submit_reloc(struct etnaviv_gem_submit *submit, void *stream, return -EINVAL; } - ret = submit_bo(submit, r->reloc_idx, &bobj, &iova); + ret = submit_bo(submit, r->reloc_idx, &bo); if (ret) return ret; - if (r->reloc_offset >= - bobj->base.size - sizeof(*ptr)) { + if (r->reloc_offset >= bo->obj->base.size - sizeof(*ptr)) { DRM_ERROR("relocation %u outside object", i); return -EINVAL; } - ptr[off] = iova + r->reloc_offset; + ptr[off] = bo->mapping->iova + r->reloc_offset; last_offset = off; } diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index 056a72e6ed26..09198d0b5814 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -72,6 +72,14 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu *gpu, u32 param, u64 *value) *value = gpu->identity.minor_features3; break; + case ETNAVIV_PARAM_GPU_FEATURES_5: + *value = gpu->identity.minor_features4; + break; + + case ETNAVIV_PARAM_GPU_FEATURES_6: + *value = gpu->identity.minor_features5; + break; + case ETNAVIV_PARAM_GPU_STREAM_COUNT: *value = gpu->identity.stream_count; break; @@ -112,6 +120,10 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu *gpu, u32 param, u64 *value) *value = gpu->identity.num_constants; break; + case ETNAVIV_PARAM_GPU_NUM_VARYINGS: + *value = gpu->identity.varyings_count; + break; + default: DBG("%s: invalid param: %u", dev_name(gpu->dev), param); return -EINVAL; @@ -120,46 +132,56 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu *gpu, u32 param, u64 *value) return 0; } + +#define etnaviv_is_model_rev(gpu, mod, rev) \ + ((gpu)->identity.model == chipModel_##mod && \ + (gpu)->identity.revision == rev) +#define etnaviv_field(val, field) \ + (((val) & field##__MASK) >> field##__SHIFT) + static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) { if (gpu->identity.minor_features0 & chipMinorFeatures0_MORE_MINOR_FEATURES) { - u32 specs[2]; + u32 specs[4]; + unsigned int streams; specs[0] = gpu_read(gpu, VIVS_HI_CHIP_SPECS); specs[1] = gpu_read(gpu, VIVS_HI_CHIP_SPECS_2); - - gpu->identity.stream_count = - (specs[0] & VIVS_HI_CHIP_SPECS_STREAM_COUNT__MASK) - >> VIVS_HI_CHIP_SPECS_STREAM_COUNT__SHIFT; - gpu->identity.register_max = - (specs[0] & VIVS_HI_CHIP_SPECS_REGISTER_MAX__MASK) - >> VIVS_HI_CHIP_SPECS_REGISTER_MAX__SHIFT; - gpu->identity.thread_count = - (specs[0] & VIVS_HI_CHIP_SPECS_THREAD_COUNT__MASK) - >> VIVS_HI_CHIP_SPECS_THREAD_COUNT__SHIFT; - gpu->identity.vertex_cache_size = - (specs[0] & VIVS_HI_CHIP_SPECS_VERTEX_CACHE_SIZE__MASK) - >> VIVS_HI_CHIP_SPECS_VERTEX_CACHE_SIZE__SHIFT; - gpu->identity.shader_core_count = - (specs[0] & VIVS_HI_CHIP_SPECS_SHADER_CORE_COUNT__MASK) - >> VIVS_HI_CHIP_SPECS_SHADER_CORE_COUNT__SHIFT; - gpu->identity.pixel_pipes = - (specs[0] & VIVS_HI_CHIP_SPECS_PIXEL_PIPES__MASK) - >> VIVS_HI_CHIP_SPECS_PIXEL_PIPES__SHIFT; + specs[2] = gpu_read(gpu, VIVS_HI_CHIP_SPECS_3); + specs[3] = gpu_read(gpu, VIVS_HI_CHIP_SPECS_4); + + gpu->identity.stream_count = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_STREAM_COUNT); + gpu->identity.register_max = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_REGISTER_MAX); + gpu->identity.thread_count = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_THREAD_COUNT); + gpu->identity.vertex_cache_size = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_VERTEX_CACHE_SIZE); + gpu->identity.shader_core_count = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_SHADER_CORE_COUNT); + gpu->identity.pixel_pipes = etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_PIXEL_PIPES); gpu->identity.vertex_output_buffer_size = - (specs[0] & VIVS_HI_CHIP_SPECS_VERTEX_OUTPUT_BUFFER_SIZE__MASK) - >> VIVS_HI_CHIP_SPECS_VERTEX_OUTPUT_BUFFER_SIZE__SHIFT; - - gpu->identity.buffer_size = - (specs[1] & VIVS_HI_CHIP_SPECS_2_BUFFER_SIZE__MASK) - >> VIVS_HI_CHIP_SPECS_2_BUFFER_SIZE__SHIFT; - gpu->identity.instruction_count = - (specs[1] & VIVS_HI_CHIP_SPECS_2_INSTRUCTION_COUNT__MASK) - >> VIVS_HI_CHIP_SPECS_2_INSTRUCTION_COUNT__SHIFT; - gpu->identity.num_constants = - (specs[1] & VIVS_HI_CHIP_SPECS_2_NUM_CONSTANTS__MASK) - >> VIVS_HI_CHIP_SPECS_2_NUM_CONSTANTS__SHIFT; + etnaviv_field(specs[0], + VIVS_HI_CHIP_SPECS_VERTEX_OUTPUT_BUFFER_SIZE); + + gpu->identity.buffer_size = etnaviv_field(specs[1], + VIVS_HI_CHIP_SPECS_2_BUFFER_SIZE); + gpu->identity.instruction_count = etnaviv_field(specs[1], + VIVS_HI_CHIP_SPECS_2_INSTRUCTION_COUNT); + gpu->identity.num_constants = etnaviv_field(specs[1], + VIVS_HI_CHIP_SPECS_2_NUM_CONSTANTS); + + gpu->identity.varyings_count = etnaviv_field(specs[2], + VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT); + + /* This overrides the value from older register if non-zero */ + streams = etnaviv_field(specs[3], + VIVS_HI_CHIP_SPECS_4_STREAM_COUNT); + if (streams) + gpu->identity.stream_count = streams; } /* Fill in the stream count if not specified */ @@ -173,7 +195,7 @@ static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) /* Convert the register max value */ if (gpu->identity.register_max) gpu->identity.register_max = 1 << gpu->identity.register_max; - else if (gpu->identity.model == 0x0400) + else if (gpu->identity.model == chipModel_GC400) gpu->identity.register_max = 32; else gpu->identity.register_max = 64; @@ -181,10 +203,10 @@ static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) /* Convert thread count */ if (gpu->identity.thread_count) gpu->identity.thread_count = 1 << gpu->identity.thread_count; - else if (gpu->identity.model == 0x0400) + else if (gpu->identity.model == chipModel_GC400) gpu->identity.thread_count = 64; - else if (gpu->identity.model == 0x0500 || - gpu->identity.model == 0x0530) + else if (gpu->identity.model == chipModel_GC500 || + gpu->identity.model == chipModel_GC530) gpu->identity.thread_count = 128; else gpu->identity.thread_count = 256; @@ -206,7 +228,7 @@ static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) if (gpu->identity.vertex_output_buffer_size) { gpu->identity.vertex_output_buffer_size = 1 << gpu->identity.vertex_output_buffer_size; - } else if (gpu->identity.model == 0x0400) { + } else if (gpu->identity.model == chipModel_GC400) { if (gpu->identity.revision < 0x4000) gpu->identity.vertex_output_buffer_size = 512; else if (gpu->identity.revision < 0x4200) @@ -219,9 +241,8 @@ static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) switch (gpu->identity.instruction_count) { case 0: - if ((gpu->identity.model == 0x2000 && - gpu->identity.revision == 0x5108) || - gpu->identity.model == 0x880) + if (etnaviv_is_model_rev(gpu, GC2000, 0x5108) || + gpu->identity.model == chipModel_GC880) gpu->identity.instruction_count = 512; else gpu->identity.instruction_count = 256; @@ -242,6 +263,30 @@ static void etnaviv_hw_specs(struct etnaviv_gpu *gpu) if (gpu->identity.num_constants == 0) gpu->identity.num_constants = 168; + + if (gpu->identity.varyings_count == 0) { + if (gpu->identity.minor_features1 & chipMinorFeatures1_HALTI0) + gpu->identity.varyings_count = 12; + else + gpu->identity.varyings_count = 8; + } + + /* + * For some cores, two varyings are consumed for position, so the + * maximum varying count needs to be reduced by one. + */ + if (etnaviv_is_model_rev(gpu, GC5000, 0x5434) || + etnaviv_is_model_rev(gpu, GC4000, 0x5222) || + etnaviv_is_model_rev(gpu, GC4000, 0x5245) || + etnaviv_is_model_rev(gpu, GC4000, 0x5208) || + etnaviv_is_model_rev(gpu, GC3000, 0x5435) || + etnaviv_is_model_rev(gpu, GC2200, 0x5244) || + etnaviv_is_model_rev(gpu, GC2100, 0x5108) || + etnaviv_is_model_rev(gpu, GC2000, 0x5108) || + etnaviv_is_model_rev(gpu, GC1500, 0x5246) || + etnaviv_is_model_rev(gpu, GC880, 0x5107) || + etnaviv_is_model_rev(gpu, GC880, 0x5106)) + gpu->identity.varyings_count -= 1; } static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) @@ -251,12 +296,10 @@ static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) chipIdentity = gpu_read(gpu, VIVS_HI_CHIP_IDENTITY); /* Special case for older graphic cores. */ - if (((chipIdentity & VIVS_HI_CHIP_IDENTITY_FAMILY__MASK) - >> VIVS_HI_CHIP_IDENTITY_FAMILY__SHIFT) == 0x01) { - gpu->identity.model = 0x500; /* gc500 */ - gpu->identity.revision = - (chipIdentity & VIVS_HI_CHIP_IDENTITY_REVISION__MASK) - >> VIVS_HI_CHIP_IDENTITY_REVISION__SHIFT; + if (etnaviv_field(chipIdentity, VIVS_HI_CHIP_IDENTITY_FAMILY) == 0x01) { + gpu->identity.model = chipModel_GC500; + gpu->identity.revision = etnaviv_field(chipIdentity, + VIVS_HI_CHIP_IDENTITY_REVISION); } else { gpu->identity.model = gpu_read(gpu, VIVS_HI_CHIP_MODEL); @@ -269,13 +312,12 @@ static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) * same. Only for GC400 family. */ if ((gpu->identity.model & 0xff00) == 0x0400 && - gpu->identity.model != 0x0420) { + gpu->identity.model != chipModel_GC420) { gpu->identity.model = gpu->identity.model & 0x0400; } /* Another special case */ - if (gpu->identity.model == 0x300 && - gpu->identity.revision == 0x2201) { + if (etnaviv_is_model_rev(gpu, GC300, 0x2201)) { u32 chipDate = gpu_read(gpu, VIVS_HI_CHIP_DATE); u32 chipTime = gpu_read(gpu, VIVS_HI_CHIP_TIME); @@ -295,11 +337,13 @@ static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) gpu->identity.features = gpu_read(gpu, VIVS_HI_CHIP_FEATURE); /* Disable fast clear on GC700. */ - if (gpu->identity.model == 0x700) + if (gpu->identity.model == chipModel_GC700) gpu->identity.features &= ~chipFeatures_FAST_CLEAR; - if ((gpu->identity.model == 0x500 && gpu->identity.revision < 2) || - (gpu->identity.model == 0x300 && gpu->identity.revision < 0x2000)) { + if ((gpu->identity.model == chipModel_GC500 && + gpu->identity.revision < 2) || + (gpu->identity.model == chipModel_GC300 && + gpu->identity.revision < 0x2000)) { /* * GC500 rev 1.x and GC300 rev < 2.0 doesn't have these @@ -309,6 +353,8 @@ static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) gpu->identity.minor_features1 = 0; gpu->identity.minor_features2 = 0; gpu->identity.minor_features3 = 0; + gpu->identity.minor_features4 = 0; + gpu->identity.minor_features5 = 0; } else gpu->identity.minor_features0 = gpu_read(gpu, VIVS_HI_CHIP_MINOR_FEATURE_0); @@ -321,6 +367,10 @@ static void etnaviv_hw_identify(struct etnaviv_gpu *gpu) gpu_read(gpu, VIVS_HI_CHIP_MINOR_FEATURE_2); gpu->identity.minor_features3 = gpu_read(gpu, VIVS_HI_CHIP_MINOR_FEATURE_3); + gpu->identity.minor_features4 = + gpu_read(gpu, VIVS_HI_CHIP_MINOR_FEATURE_4); + gpu->identity.minor_features5 = + gpu_read(gpu, VIVS_HI_CHIP_MINOR_FEATURE_5); } /* GC600 idle register reports zero bits where modules aren't present */ @@ -441,10 +491,9 @@ static void etnaviv_gpu_hw_init(struct etnaviv_gpu *gpu) { u16 prefetch; - if (gpu->identity.model == chipModel_GC320 && - gpu_read(gpu, VIVS_HI_CHIP_TIME) != 0x2062400 && - (gpu->identity.revision == 0x5007 || - gpu->identity.revision == 0x5220)) { + if ((etnaviv_is_model_rev(gpu, GC320, 0x5007) || + etnaviv_is_model_rev(gpu, GC320, 0x5220)) && + gpu_read(gpu, VIVS_HI_CHIP_TIME) != 0x2062400) { u32 mc_memory_debug; mc_memory_debug = gpu_read(gpu, VIVS_MC_DEBUG_MEMORY) & ~0xff; @@ -466,7 +515,7 @@ static void etnaviv_gpu_hw_init(struct etnaviv_gpu *gpu) VIVS_HI_AXI_CONFIG_ARCACHE(2)); /* GC2000 rev 5108 needs a special bus config */ - if (gpu->identity.model == 0x2000 && gpu->identity.revision == 0x5108) { + if (etnaviv_is_model_rev(gpu, GC2000, 0x5108)) { u32 bus_config = gpu_read(gpu, VIVS_MC_BUS_CONFIG); bus_config &= ~(VIVS_MC_BUS_CONFIG_FE_BUS_CONFIG__MASK | VIVS_MC_BUS_CONFIG_TX_BUS_CONFIG__MASK); @@ -511,8 +560,16 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) if (gpu->identity.model == 0) { dev_err(gpu->dev, "Unknown GPU model\n"); - pm_runtime_put_autosuspend(gpu->dev); - return -ENXIO; + ret = -ENXIO; + goto fail; + } + + /* Exclude VG cores with FE2.0 */ + if (gpu->identity.features & chipFeatures_PIPE_VG && + gpu->identity.features & chipFeatures_FE20) { + dev_info(gpu->dev, "Ignoring GPU with VG and FE2.0\n"); + ret = -ENXIO; + goto fail; } ret = etnaviv_hw_reset(gpu); @@ -539,10 +596,9 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) goto fail; } - /* TODO: we will leak here memory - fix it! */ - gpu->mmu = etnaviv_iommu_new(gpu, iommu, version); if (!gpu->mmu) { + iommu_domain_free(iommu); ret = -ENOMEM; goto fail; } @@ -552,7 +608,7 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) if (!gpu->buffer) { ret = -ENOMEM; dev_err(gpu->dev, "could not create command buffer\n"); - goto fail; + goto destroy_iommu; } if (gpu->buffer->paddr - gpu->memory_base > 0x80000000) { ret = -EINVAL; @@ -572,6 +628,7 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) /* Now program the hardware */ mutex_lock(&gpu->lock); etnaviv_gpu_hw_init(gpu); + gpu->exec_state = -1; mutex_unlock(&gpu->lock); pm_runtime_mark_last_busy(gpu->dev); @@ -582,6 +639,9 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) free_buffer: etnaviv_gpu_cmdbuf_free(gpu->buffer); gpu->buffer = NULL; +destroy_iommu: + etnaviv_iommu_destroy(gpu->mmu); + gpu->mmu = NULL; fail: pm_runtime_mark_last_busy(gpu->dev); pm_runtime_put_autosuspend(gpu->dev); @@ -642,6 +702,10 @@ int etnaviv_gpu_debugfs(struct etnaviv_gpu *gpu, struct seq_file *m) gpu->identity.minor_features2); seq_printf(m, "\t minor_features3: 0x%08x\n", gpu->identity.minor_features3); + seq_printf(m, "\t minor_features4: 0x%08x\n", + gpu->identity.minor_features4); + seq_printf(m, "\t minor_features5: 0x%08x\n", + gpu->identity.minor_features5); seq_puts(m, "\tspecs\n"); seq_printf(m, "\t stream_count: %d\n", @@ -664,6 +728,8 @@ int etnaviv_gpu_debugfs(struct etnaviv_gpu *gpu, struct seq_file *m) gpu->identity.instruction_count); seq_printf(m, "\t num_constants: %d\n", gpu->identity.num_constants); + seq_printf(m, "\t varyings_count: %d\n", + gpu->identity.varyings_count); seq_printf(m, "\taxi: 0x%08x\n", axi); seq_printf(m, "\tidle: 0x%08x\n", idle); @@ -806,17 +872,13 @@ static void recover_worker(struct work_struct *work) gpu->event[i].fence = NULL; gpu->event[i].used = false; complete(&gpu->event_free); - /* - * Decrement the PM count for each stuck event. This is safe - * even in atomic context as we use ASYNC RPM here. - */ - pm_runtime_put_autosuspend(gpu->dev); } spin_unlock_irqrestore(&gpu->event_spinlock, flags); gpu->completed_fence = gpu->active_fence; etnaviv_gpu_hw_init(gpu); gpu->switch_context = true; + gpu->exec_state = -1; mutex_unlock(&gpu->lock); pm_runtime_mark_last_busy(gpu->dev); @@ -1041,15 +1103,15 @@ struct etnaviv_cmdbuf *etnaviv_gpu_cmdbuf_new(struct etnaviv_gpu *gpu, u32 size, size_t nr_bos) { struct etnaviv_cmdbuf *cmdbuf; - size_t sz = size_vstruct(nr_bos, sizeof(cmdbuf->bo[0]), + size_t sz = size_vstruct(nr_bos, sizeof(cmdbuf->bo_map[0]), sizeof(*cmdbuf)); cmdbuf = kzalloc(sz, GFP_KERNEL); if (!cmdbuf) return NULL; - cmdbuf->vaddr = dma_alloc_writecombine(gpu->dev, size, &cmdbuf->paddr, - GFP_KERNEL); + cmdbuf->vaddr = dma_alloc_wc(gpu->dev, size, &cmdbuf->paddr, + GFP_KERNEL); if (!cmdbuf->vaddr) { kfree(cmdbuf); return NULL; @@ -1063,8 +1125,8 @@ struct etnaviv_cmdbuf *etnaviv_gpu_cmdbuf_new(struct etnaviv_gpu *gpu, u32 size, void etnaviv_gpu_cmdbuf_free(struct etnaviv_cmdbuf *cmdbuf) { - dma_free_writecombine(cmdbuf->gpu->dev, cmdbuf->size, - cmdbuf->vaddr, cmdbuf->paddr); + dma_free_wc(cmdbuf->gpu->dev, cmdbuf->size, cmdbuf->vaddr, + cmdbuf->paddr); kfree(cmdbuf); } @@ -1085,14 +1147,23 @@ static void retire_worker(struct work_struct *work) fence_put(cmdbuf->fence); for (i = 0; i < cmdbuf->nr_bos; i++) { - struct etnaviv_gem_object *etnaviv_obj = cmdbuf->bo[i]; + struct etnaviv_vram_mapping *mapping = cmdbuf->bo_map[i]; + struct etnaviv_gem_object *etnaviv_obj = mapping->object; atomic_dec(&etnaviv_obj->gpu_active); /* drop the refcount taken in etnaviv_gpu_submit */ - etnaviv_gem_put_iova(gpu, &etnaviv_obj->base); + etnaviv_gem_mapping_unreference(mapping); } etnaviv_gpu_cmdbuf_free(cmdbuf); + /* + * We need to balance the runtime PM count caused by + * each submission. Upon submission, we increment + * the runtime PM counter, and allocate one event. + * So here, we put the runtime PM count for each + * completed event. + */ + pm_runtime_put_autosuspend(gpu->dev); } gpu->retired_fence = fence; @@ -1239,11 +1310,10 @@ int etnaviv_gpu_submit(struct etnaviv_gpu *gpu, for (i = 0; i < submit->nr_bos; i++) { struct etnaviv_gem_object *etnaviv_obj = submit->bos[i].obj; - u32 iova; - /* Each cmdbuf takes a refcount on the iova */ - etnaviv_gem_get_iova(gpu, &etnaviv_obj->base, &iova); - cmdbuf->bo[i] = etnaviv_obj; + /* Each cmdbuf takes a refcount on the mapping */ + etnaviv_gem_mapping_reference(submit->bos[i].mapping); + cmdbuf->bo_map[i] = submit->bos[i].mapping; atomic_inc(&etnaviv_obj->gpu_active); if (submit->bos[i].flags & ETNA_SUBMIT_BO_WRITE) @@ -1313,15 +1383,6 @@ static irqreturn_t irq_handler(int irq, void *data) gpu->completed_fence = fence->seqno; event_free(gpu, event); - - /* - * We need to balance the runtime PM count caused by - * each submission. Upon submission, we increment - * the runtime PM counter, and allocate one event. - * So here, we put the runtime PM count for each - * completed event. - */ - pm_runtime_put_autosuspend(gpu->dev); } /* Retire the buffer objects in a work */ @@ -1416,6 +1477,7 @@ static int etnaviv_gpu_hw_resume(struct etnaviv_gpu *gpu) etnaviv_gpu_hw_init(gpu); gpu->switch_context = true; + gpu->exec_state = -1; mutex_unlock(&gpu->lock); @@ -1504,6 +1566,7 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; struct etnaviv_gpu *gpu; + u32 dma_mask; int err = 0; gpu = devm_kzalloc(dev, sizeof(*gpu), GFP_KERNEL); @@ -1514,12 +1577,16 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) mutex_init(&gpu->lock); /* - * Set the GPU base address to the start of physical memory. This - * ensures that if we have up to 2GB, the v1 MMU can address the - * highest memory. This is important as command buffers may be - * allocated outside of this limit. + * Set the GPU linear window to be at the end of the DMA window, where + * the CMA area is likely to reside. This ensures that we are able to + * map the command buffers while having the linear window overlap as + * much RAM as possible, so we can optimize mappings for other buffers. */ - gpu->memory_base = PHYS_OFFSET; + dma_mask = (u32)dma_get_required_mask(dev); + if (dma_mask < PHYS_OFFSET + SZ_2G) + gpu->memory_base = PHYS_OFFSET; + else + gpu->memory_base = dma_mask - SZ_2G + 1; /* Map registers: */ gpu->mmio = etnaviv_ioremap(pdev, NULL, dev_name(gpu->dev)); diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h index c75d50359ab0..f5321e2f25ff 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h @@ -23,6 +23,7 @@ #include "etnaviv_drv.h" struct etnaviv_gem_submit; +struct etnaviv_vram_mapping; struct etnaviv_chip_identity { /* Chip model. */ @@ -46,6 +47,12 @@ struct etnaviv_chip_identity { /* Supported minor feature 3 fields. */ u32 minor_features3; + /* Supported minor feature 4 fields. */ + u32 minor_features4; + + /* Supported minor feature 5 fields. */ + u32 minor_features5; + /* Number of streams supported. */ u32 stream_count; @@ -75,6 +82,9 @@ struct etnaviv_chip_identity { /* Buffer size */ u32 buffer_size; + + /* Number of varyings */ + u8 varyings_count; }; struct etnaviv_event { @@ -94,6 +104,7 @@ struct etnaviv_gpu { /* 'ring'-buffer: */ struct etnaviv_cmdbuf *buffer; + int exec_state; /* bus base address of memory */ u32 memory_base; @@ -157,7 +168,7 @@ struct etnaviv_cmdbuf { struct list_head node; /* BOs attached to this command buffer */ unsigned int nr_bos; - struct etnaviv_gem_object *bo[0]; + struct etnaviv_vram_mapping *bo_map[0]; }; static inline void gpu_write(struct etnaviv_gpu *gpu, u32 reg, u32 data) diff --git a/drivers/gpu/drm/etnaviv/etnaviv_mmu.c b/drivers/gpu/drm/etnaviv/etnaviv_mmu.c index 6743bc648dc8..29a723fabc17 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_mmu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_mmu.c @@ -193,7 +193,7 @@ int etnaviv_iommu_map_gem(struct etnaviv_iommu *mmu, /* * Unmap the blocks which need to be reaped from the MMU. - * Clear the mmu pointer to prevent the get_iova finding + * Clear the mmu pointer to prevent the mapping_get finding * this mapping. */ list_for_each_entry_safe(m, n, &list, scan_node) { diff --git a/drivers/gpu/drm/etnaviv/state_3d.xml.h b/drivers/gpu/drm/etnaviv/state_3d.xml.h new file mode 100644 index 000000000000..d7146fd13943 --- /dev/null +++ b/drivers/gpu/drm/etnaviv/state_3d.xml.h @@ -0,0 +1,9 @@ +#ifndef STATE_3D_XML +#define STATE_3D_XML + +/* This is a cut-down version of the state_3d.xml.h file */ + +#define VIVS_TS_FLUSH_CACHE 0x00001650 +#define VIVS_TS_FLUSH_CACHE_FLUSH 0x00000001 + +#endif /* STATE_3D_XML */ diff --git a/drivers/gpu/drm/etnaviv/state_hi.xml.h b/drivers/gpu/drm/etnaviv/state_hi.xml.h index 0064f2640396..6a7de5f1454a 100644 --- a/drivers/gpu/drm/etnaviv/state_hi.xml.h +++ b/drivers/gpu/drm/etnaviv/state_hi.xml.h @@ -8,8 +8,8 @@ http://0x04.net/cgit/index.cgi/rules-ng-ng git clone git://0x04.net/rules-ng-ng The rules-ng-ng source files this header was generated from are: -- state_hi.xml ( 23420 bytes, from 2015-03-25 11:47:21) -- common.xml ( 18437 bytes, from 2015-03-25 11:27:41) +- state_hi.xml ( 24309 bytes, from 2015-12-12 09:02:53) +- common.xml ( 18437 bytes, from 2015-12-12 09:02:53) Copyright (C) 2015 */ @@ -182,8 +182,25 @@ Copyright (C) 2015 #define VIVS_HI_CHIP_MINOR_FEATURE_3 0x00000088 +#define VIVS_HI_CHIP_SPECS_3 0x0000008c +#define VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT__MASK 0x000001f0 +#define VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT__SHIFT 4 +#define VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT(x) (((x) << VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT__SHIFT) & VIVS_HI_CHIP_SPECS_3_VARYINGS_COUNT__MASK) +#define VIVS_HI_CHIP_SPECS_3_GPU_CORE_COUNT__MASK 0x00000007 +#define VIVS_HI_CHIP_SPECS_3_GPU_CORE_COUNT__SHIFT 0 +#define VIVS_HI_CHIP_SPECS_3_GPU_CORE_COUNT(x) (((x) << VIVS_HI_CHIP_SPECS_3_GPU_CORE_COUNT__SHIFT) & VIVS_HI_CHIP_SPECS_3_GPU_CORE_COUNT__MASK) + #define VIVS_HI_CHIP_MINOR_FEATURE_4 0x00000094 +#define VIVS_HI_CHIP_SPECS_4 0x0000009c +#define VIVS_HI_CHIP_SPECS_4_STREAM_COUNT__MASK 0x0001f000 +#define VIVS_HI_CHIP_SPECS_4_STREAM_COUNT__SHIFT 12 +#define VIVS_HI_CHIP_SPECS_4_STREAM_COUNT(x) (((x) << VIVS_HI_CHIP_SPECS_4_STREAM_COUNT__SHIFT) & VIVS_HI_CHIP_SPECS_4_STREAM_COUNT__MASK) + +#define VIVS_HI_CHIP_MINOR_FEATURE_5 0x000000a0 + +#define VIVS_HI_CHIP_PRODUCT_ID 0x000000a8 + #define VIVS_PM 0x00000000 #define VIVS_PM_POWER_CONTROLS 0x00000100 @@ -206,6 +223,11 @@ Copyright (C) 2015 #define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_FE 0x00000001 #define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_DE 0x00000002 #define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_PE 0x00000004 +#define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_SH 0x00000008 +#define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_PA 0x00000010 +#define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_SE 0x00000020 +#define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_RA 0x00000040 +#define VIVS_PM_MODULE_STATUS_MODULE_CLOCK_GATED_TX 0x00000080 #define VIVS_PM_PULSE_EATER 0x0000010c |