From: Christoph Bumiller Subject: Re: [PATCH] nv50: remove vtxbuf stateobject after a referenced vtxbuf is mapped Date: Mon, 21 Dec 2009 11:25:08 +0100 Message-ID: <4B2F4D04.1000101@student.tuwien.ac.at> References: <1261323978-5330-1-git-send-email-madman2003@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1261323978-5330-1-git-send-email-madman2003-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: nouveau-bounces-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org Errors-To: nouveau-bounces-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org To: Maarten Maathuis Cc: nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org List-Id: nouveau.vger.kernel.org On 12/20/2009 04:46 PM, Maarten Maathuis wrote: > - This avoids problematic "reloc'ed while mapped" messages and > some associated corruption as well. > - Also add one nouveau_bo_unmap() in the vbo code that wasn't present. I didn't think nouveau_bo_unmap was necessary if the map failed. As far as I can see, map will be NULL, and unmap will do nothing. I don't know what our API doc says about that though ... But the DDX seems to not unmap after failure. > > Signed-off-by: Maarten Maathuis > --- > src/gallium/drivers/nouveau/nouveau_screen.c | 21 +++++++++++++++++++++ > src/gallium/drivers/nouveau/nouveau_screen.h | 3 +++ > src/gallium/drivers/nouveau/nouveau_stateobj.h | 13 +++++++++++++ > src/gallium/drivers/nv50/nv50_screen.c | 23 +++++++++++++++++++++++ > src/gallium/drivers/nv50/nv50_screen.h | 2 ++ > src/gallium/drivers/nv50/nv50_state_validate.c | 2 ++ > src/gallium/drivers/nv50/nv50_vbo.c | 4 +++- > 7 files changed, 67 insertions(+), 1 deletions(-) > > diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c > index e4cf91c..c85057a 100644 > --- a/src/gallium/drivers/nouveau/nouveau_screen.c > +++ b/src/gallium/drivers/nouveau/nouveau_screen.c > @@ -127,8 +127,18 @@ nouveau_screen_bo_map(struct pipe_screen *pscreen, struct pipe_buffer *pb, > unsigned usage) ... > +static int > +nv50_pre_pipebuffer_map(struct pipe_screen *pscreen, struct pipe_buffer *pb, > + unsigned usage) > +{ > + struct nv50_screen *screen = nv50_screen(pscreen); > + struct nv50_context *ctx = screen->cur_ctx; > + > + if (!(pb->usage & PIPE_BUFFER_USAGE_VERTEX)) > + return 0; > + > + /* Our vtxbuf got mapped, it can no longer be considered part of current > + * state, remove it to avoid emitting reloc markers. > + */ > + if (ctx && ctx->state.vtxbuf && so_bo_is_reloc(ctx->state.vtxbuf, > + nouveau_bo(pb))) { > + so_ref(NULL, &ctx->state.vtxbuf); > + ctx->state.dirty |= NV50_NEW_ARRAYS; Hm ... I know I suggested it but I'm not completely sure it works now. ctx->state.dirty will be checked in nv50_state_emit as far as I can see, and then state stateobj better not be NULL. (at least if I do this regardless of so_bo_is_reloc, I get a segfault) Since we're operating on the current context here, we should probably set ctx->(no state.)dirty which is used in nv50_state_validate (no segfault on emit then). > + } > + > + return 0; > +} > + > struct pipe_screen * > nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) > { > @@ -201,6 +223,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) > pscreen->get_param = nv50_screen_get_param; > pscreen->get_paramf = nv50_screen_get_paramf; > pscreen->is_format_supported = nv50_screen_is_format_supported; > + screen->base.pre_pipebuffer_map_callback = nv50_pre_pipebuffer_map; > > nv50_screen_init_miptree_functions(pscreen); > nv50_transfer_init_screen_functions(pscreen); > diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h > index 61e24a5..a038a4e 100644 > --- a/src/gallium/drivers/nv50/nv50_screen.h > +++ b/src/gallium/drivers/nv50/nv50_screen.h > @@ -2,6 +2,7 @@ > #define __NV50_SCREEN_H__ > > #include "nouveau/nouveau_screen.h" > +#include "nv50_context.h" > > struct nv50_screen { > struct nouveau_screen base; > @@ -9,6 +10,7 @@ struct nv50_screen { > struct nouveau_winsys *nvws; > > unsigned cur_pctx; > + struct nv50_context *cur_ctx; > Using struct pipe_context **nouveau_winsys::pcxt would work too ... > struct nouveau_grobj *tesla; > struct nouveau_grobj *eng2d; > diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c > index 871e809..8cf4ff4 100644 > --- a/src/gallium/drivers/nv50/nv50_state_validate.c > +++ b/src/gallium/drivers/nv50/nv50_state_validate.c > @@ -185,6 +185,8 @@ nv50_state_emit(struct nv50_context *nv50) > struct nv50_screen *screen = nv50->screen; > struct nouveau_channel *chan = screen->base.channel; > > + screen->cur_ctx = nv50; > + > if (nv50->pctx_id != screen->cur_pctx) { > if (nv50->state.fb) > nv50->state.dirty |= NV50_NEW_FRAMEBUFFER; > diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c > index db54380..ce6e4eb 100644 > --- a/src/gallium/drivers/nv50/nv50_vbo.c > +++ b/src/gallium/drivers/nv50/nv50_vbo.c > @@ -325,8 +325,10 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, > return FALSE; > > ret = nouveau_bo_map(bo, NOUVEAU_BO_RD); > - if (ret) > + if (ret) { > + nouveau_bo_unmap(bo); > return FALSE; > + } > v = (float *)(bo->map + (vb->buffer_offset + ve->src_offset)); > > so = *pso;