<div dir="ltr"><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Wed, Aug 23, 2023 at 1:17â€¯PM Danilo Krummrich <<a href="mailto:dakr@redhat.com">dakr@redhat.com</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Currently, NO_PREFETCH is passed implicitly through drm_nouveau_gem_pushbuf_push::length and drm_nouveau_exec_push::va_len. Since this is a direct representation of how the HW is programmed it isn't really future proof for a uAPI. Hence, fix this up for the new uAPI and split up the va_len field of struct drm_nouveau_exec_push, such that we keep 32bit for va_len and 32bit for flags. For drm_nouveau_gem_pushbuf_push::length at least provide NOUVEAU_GEM_PUSHBUF_NO_PREFETCH to indicate the bit shift. While at it, fix up nv50_dma_push() as well, such that the caller doesn't need to encode the NO_PREFETCH flag into the length parameter. Signed-off-by: Danilo Krummrich <<a href="mailto:dakr@redhat.com" target="_blank">dakr@redhat.com</a>> </blockquote><div> </div><div>Still</div><div> </div><div>Reviewed-by: Faith Ekstrand <<a href="mailto:faith.ekstrand@collabora.com">faith.ekstrand@collabora.com</a>></div><div>Â </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> --- Changes in v2: Â - dma: rename prefetch to no_prefetch in nv50_dma_push() (Faith) Â - exec: print error message when pushbuf size larger max pushbuf size (Faith) --- Â drivers/gpu/drm/nouveau/nouveau_dma.cÂ |Â 7 +++++-- Â drivers/gpu/drm/nouveau/nouveau_dma.hÂ |Â 8 ++++++-- Â drivers/gpu/drm/nouveau/nouveau_exec.c | 19 ++++++++++++++++--- Â drivers/gpu/drm/nouveau/nouveau_gem.cÂ |Â 6 ++++-- Â include/uapi/drm/nouveau_drm.hÂ Â Â Â Â |Â 8 +++++++- Â 5 files changed, 38 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.c b/drivers/gpu/drm/nouveau/nouveau_dma.c index b90cac6d5772..b01c029f3a90 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.c +++ b/drivers/gpu/drm/nouveau/nouveau_dma.c @@ -69,16 +69,19 @@ READ_GET(struct nouveau_channel *chan, uint64_t *prev_get, int *timeout) Â } Â void -nv50_dma_push(struct nouveau_channel *chan, u64 offset, int length) +nv50_dma_push(struct nouveau_channel *chan, u64 offset, u32 length, +Â Â Â Â Â Â Â bool no_prefetch) Â { Â Â Â Â struct nvif_user *user = &chan->drm->client.device.user; Â Â Â Â struct nouveau_bo *pb = chan->push.buffer; Â Â Â Â int ip = (chan->dma.ib_put * 2) + chan->dma.ib_base; Â Â Â Â BUG_ON(chan->dma.ib_free < 1); +Â Â Â Â WARN_ON(length > NV50_DMA_PUSH_MAX_LENGTH); Â Â Â Â nouveau_bo_wr32(pb, ip++, lower_32_bits(offset)); -Â Â Â Â nouveau_bo_wr32(pb, ip++, upper_32_bits(offset) | length << 8); +Â Â Â Â nouveau_bo_wr32(pb, ip++, upper_32_bits(offset) | length << 8 | +Â Â Â Â Â Â Â Â Â Â Â Â (no_prefetch ? (1 << 31) : 0)); Â Â Â Â chan->dma.ib_put = (chan->dma.ib_put + 1) & chan->dma.ib_max; diff --git a/drivers/gpu/drm/nouveau/nouveau_dma.h b/drivers/gpu/drm/nouveau/nouveau_dma.h index 035a709c7be1..1744d95b233e 100644 --- a/drivers/gpu/drm/nouveau/nouveau_dma.h +++ b/drivers/gpu/drm/nouveau/nouveau_dma.h @@ -31,7 +31,8 @@ Â #include "nouveau_chan.h" Â int nouveau_dma_wait(struct nouveau_channel *, int slots, int size); -void nv50_dma_push(struct nouveau_channel *, u64 addr, int length); +void nv50_dma_push(struct nouveau_channel *, u64 addr, u32 length, +Â Â Â Â Â Â Â Â Â bool no_prefetch); Â /* Â * There's a hw race condition where you can't jump to your PUT offset, @@ -45,6 +46,9 @@ void nv50_dma_push(struct nouveau_channel *, u64 addr, int length); Â */ Â #define NOUVEAU_DMA_SKIPS (128 / 4) +/* Maximum push buffer size. */ +#define NV50_DMA_PUSH_MAX_LENGTH 0x7fffff + Â /* Object handles - for stuff that's doesn't use handle == oclass. */ Â enum { Â Â Â Â NvDmaFBÂ Â Â Â Â = 0x80000002, @@ -89,7 +93,7 @@ FIRE_RING(struct nouveau_channel *chan) Â Â Â Â if (chan->dma.ib_max) { Â Â Â Â Â Â Â Â nv50_dma_push(chan, chan->push.addr + (chan->dma.put << 2), -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â (chan->dma.cur - chan->dma.put) << 2); +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â (chan->dma.cur - chan->dma.put) << 2, false); Â Â Â Â } else { Â Â Â Â Â Â Â Â WRITE_PUT(chan->dma.cur); Â Â Â Â } diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.c b/drivers/gpu/drm/nouveau/nouveau_exec.c index 0f927adda4ed..a90c4cd8cbb2 100644 --- a/drivers/gpu/drm/nouveau/nouveau_exec.c +++ b/drivers/gpu/drm/nouveau/nouveau_exec.c @@ -164,8 +164,10 @@ nouveau_exec_job_run(struct nouveau_job *job) Â Â Â Â } Â Â Â Â for (i = 0; i < exec_job->push.count; i++) { -Â Â Â Â Â Â Â Â nv50_dma_push(chan, exec_job->push.s[i].va, -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â exec_job->push.s[i].va_len); +Â Â Â Â Â Â Â Â struct drm_nouveau_exec_push *p = &exec_job->push.s[i]; +Â Â Â Â Â Â Â Â bool no_prefetch = p->flags & DRM_NOUVEAU_EXEC_PUSH_NO_PREFETCH; + +Â Â Â Â Â Â Â Â nv50_dma_push(chan, p->va, p->va_len, no_prefetch); Â Â Â Â } Â Â Â Â ret = nouveau_fence_emit(fence, chan); @@ -223,7 +225,18 @@ nouveau_exec_job_init(struct nouveau_exec_job **pjob, Â { Â Â Â Â struct nouveau_exec_job *job; Â Â Â Â struct nouveau_job_args args = {}; -Â Â Â Â int ret; +Â Â Â Â int i, ret; + +Â Â Â Â for (i = 0; i < __args->push.count; i++) { +Â Â Â Â Â Â Â Â struct drm_nouveau_exec_push *p = &__args->push.s[i]; + +Â Â Â Â Â Â Â Â if (unlikely(p->va_len > NV50_DMA_PUSH_MAX_LENGTH)) { +Â Â Â Â Â Â Â Â Â Â Â Â NV_PRINTK(err, nouveau_cli(__args->file_priv), +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â "pushbuf size exceeds limit: 0x%x max 0x%x\n", +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â p->va_len, NV50_DMA_PUSH_MAX_LENGTH); +Â Â Â Â Â Â Â Â Â Â Â Â return -EINVAL; +Â Â Â Â Â Â Â Â } +Â Â Â Â } Â Â Â Â job = *pjob = kzalloc(sizeof(*job), GFP_KERNEL); Â Â Â Â if (!job) diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c index f39360870c70..c0b10d8d3d03 100644 --- a/drivers/gpu/drm/nouveau/nouveau_gem.c +++ b/drivers/gpu/drm/nouveau/nouveau_gem.c @@ -856,9 +856,11 @@ nouveau_gem_ioctl_pushbuf(struct drm_device *dev, void *data, Â Â Â Â Â Â Â Â for (i = 0; i < req->nr_push; i++) { Â Â Â Â Â Â Â Â Â Â Â Â struct nouveau_vma *vma = (void *)(unsigned long) Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â bo[push[i].bo_index].user_priv; +Â Â Â Â Â Â Â Â Â Â Â Â u64 addr = vma->addr + push[i].offset; +Â Â Â Â Â Â Â Â Â Â Â Â u32 length = push[i].length & ~NOUVEAU_GEM_PUSHBUF_NO_PREFETCH; +Â Â Â Â Â Â Â Â Â Â Â Â bool no_prefetch = push[i].length & NOUVEAU_GEM_PUSHBUF_NO_PREFETCH; -Â Â Â Â Â Â Â Â Â Â Â Â nv50_dma_push(chan, vma->addr + push[i].offset, -Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â push[i].length); +Â Â Â Â Â Â Â Â Â Â Â Â nv50_dma_push(chan, addr, length, no_prefetch); Â Â Â Â Â Â Â Â } Â Â Â Â } else Â Â Â Â if (drm-><a href="http://client.device.info" target="_blank">client.device.info</a>.chipset >= 0x25) { diff --git a/include/uapi/drm/nouveau_drm.h b/include/uapi/drm/nouveau_drm.h index b1ad9d5ffce8..8f16724b5d05 100644 --- a/include/uapi/drm/nouveau_drm.h +++ b/include/uapi/drm/nouveau_drm.h @@ -138,6 +138,7 @@ struct drm_nouveau_gem_pushbuf_push { Â Â Â Â __u32 pad; Â Â Â Â __u64 offset; Â Â Â Â __u64 length; +#define NOUVEAU_GEM_PUSHBUF_NO_PREFETCH (1 << 23) Â }; Â struct drm_nouveau_gem_pushbuf { @@ -338,7 +339,12 @@ struct drm_nouveau_exec_push { Â Â Â Â /** Â Â Â Â Â * @va_len: the length of the push buffer mapping Â Â Â Â Â */ -Â Â Â Â __u64 va_len; +Â Â Â Â __u32 va_len; +Â Â Â Â /** +Â Â Â Â * flags: the flags for this push buffer mapping +Â Â Â Â */ +Â Â Â Â __u32 flags; +#define DRM_NOUVEAU_EXEC_PUSH_NO_PREFETCH 0x1 Â }; Â /** base-commit: b4e9fa933551e51459c634dc4396171dc65284a6 -- 2.41.0 </blockquote></div></div>