From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Vetter Subject: [PATCH 13/37] drm/nouveau: protect evo_wait/evo_kick sections with a channel mutex Date: Wed, 12 Dec 2012 14:06:53 +0100 Message-ID: <1355317637-16742-14-git-send-email-daniel.vetter@ffwll.ch> References: <1355317637-16742-1-git-send-email-daniel.vetter@ffwll.ch> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1355317637-16742-1-git-send-email-daniel.vetter-/w4YWyX8dFk@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: xorg-driver-ati-bounces+gcfxda-xorg-driver-ati=m.gmane.org-go0+a7rfsptAfugRpC6u6w@public.gmane.org Errors-To: xorg-driver-ati-bounces+gcfxda-xorg-driver-ati=m.gmane.org-go0+a7rfsptAfugRpC6u6w@public.gmane.org To: DRI Development Cc: Nouveau Dev , Intel Graphics Development , Radeon Dev , Daniel Vetter List-Id: dri-devel@lists.freedesktop.org With per-crtc locks modeset operations can run in parallel, and the cursor code uses the device-global evo master channel for hw frobbing. But the pageflip code can also sync with the master under some circumstances. Hence just wrap things up in a mutex to ensure that pushbuf access doesn't intermingle. The approach here is a bit overkill since the per-crtc channels used to schedule the pageflips could probably be used without this pushbuf locking, but I'm not familiar enough with the nouveau codebase to be sure of that. Signed-off-by: Daniel Vetter --- drivers/gpu/drm/nouveau/nv50_display.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 3587408..5751d63 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -128,6 +128,11 @@ struct nv50_dmac { struct nv50_chan base; dma_addr_t handle; u32 *ptr; + + /* Protects against concurrent pushbuf access to this channel, lock is + * grabbed by evo_wait (if the pushbuf reservation is successful) and + * dropped again by evo_kick. */ + struct mutex lock; }; static void @@ -395,11 +400,13 @@ evo_wait(void *evoc, int nr) struct nv50_dmac *dmac = evoc; u32 put = nv_ro32(dmac->base.user, 0x0000) / 4; + mutex_lock(&dmac->lock); if (put + nr >= (PAGE_SIZE / 4) - 8) { dmac->ptr[put] = 0x20000000; nv_wo32(dmac->base.user, 0x0000, 0x00000000); if (!nv_wait(dmac->base.user, 0x0004, ~0, 0x00000000)) { + mutex_unlock(&dmac->lock); NV_ERROR(dmac->base.user, "channel stalled\n"); return NULL; } @@ -415,6 +422,7 @@ evo_kick(u32 *push, void *evoc) { struct nv50_dmac *dmac = evoc; nv_wo32(dmac->base.user, 0x0000, (push - dmac->ptr) << 2); + mutex_unlock(&dmac->lock); } #define evo_mthd(p,m,s) *((p)++) = (((s) << 18) | (m)) -- 1.7.10.4