From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 05465C83F1B for ; Mon, 14 Jul 2025 05:23:20 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 720AC10E20F; Mon, 14 Jul 2025 05:23:19 +0000 (UTC) Received: from us-smtp-delivery-44.mimecast.com (us-smtp-delivery-44.mimecast.com [205.139.111.44]) by gabe.freedesktop.org (Postfix) with ESMTPS id 8EFAD10E20F for ; Mon, 14 Jul 2025 05:23:18 +0000 (UTC) Received: from mx-prod-mc-01.mail-002.prod.us-west-2.aws.redhat.com (ec2-54-186-198-63.us-west-2.compute.amazonaws.com [54.186.198.63]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_256_GCM_SHA384) id us-mta-184-aDS3qAOSNSGK5Rr84Iljaw-1; Mon, 14 Jul 2025 01:23:14 -0400 X-MC-Unique: aDS3qAOSNSGK5Rr84Iljaw-1 X-Mimecast-MFC-AGG-ID: aDS3qAOSNSGK5Rr84Iljaw_1752470592 Received: from mx-prod-int-05.mail-002.prod.us-west-2.aws.redhat.com (mx-prod-int-05.mail-002.prod.us-west-2.aws.redhat.com [10.30.177.17]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256) (No client certificate requested) by mx-prod-mc-01.mail-002.prod.us-west-2.aws.redhat.com (Postfix) with ESMTPS id 4F3CB195608F; Mon, 14 Jul 2025 05:23:12 +0000 (UTC) Received: from dreadlord.redhat.com (unknown [10.67.32.31]) by mx-prod-int-05.mail-002.prod.us-west-2.aws.redhat.com (Postfix) with ESMTP id A41A91977000; Mon, 14 Jul 2025 05:23:06 +0000 (UTC) From: Dave Airlie To: dri-devel@lists.freedesktop.org, linux-mm@kvack.org, Johannes Weiner , Christian Koenig Cc: Dave Chinner , Kairui Song , Dave Airlie , Matthew Brost , Andrew Morton Subject: [PATCH 02/18] drm/ttm: use gpu mm stats to track gpu memory allocations. (v3) Date: Mon, 14 Jul 2025 15:18:17 +1000 Message-ID: <20250714052243.1149732-3-airlied@gmail.com> In-Reply-To: <20250714052243.1149732-1-airlied@gmail.com> References: <20250714052243.1149732-1-airlied@gmail.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 3.0 on 10.30.177.17 X-Mimecast-Spam-Score: 0 X-Mimecast-MFC-PROC-ID: mGXohN5prTV9uMHBl3oVrNRcWGBQom3_whNPufKBWQ0_1752470592 X-Mimecast-Originator: gmail.com Content-Transfer-Encoding: quoted-printable content-type: text/plain; charset=WINDOWS-1252; x-default=true X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" From: Dave Airlie This uses the newly introduced per-node gpu tracking stats, to track GPU memory allocated via TTM and reclaimable memory in the TTM page pools. These stats will be useful later for system information and later when mem cgroups are integrated. Cc: Christian Koenig Cc: Matthew Brost Cc: Johannes Weiner Cc: linux-mm@kvack.org Cc: Andrew Morton Signed-off-by: Dave Airlie --- v2: add reclaim parameters and adjust the right counters. v3: drop the nid helper and get it from page. --- drivers/gpu/drm/ttm/ttm_pool.c | 25 +++++++++++++++++++------ 1 file changed, 19 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/ttm/ttm_pool.c b/drivers/gpu/drm/ttm/ttm_pool.= c index baf27c70a419..ee2344089d47 100644 --- a/drivers/gpu/drm/ttm/ttm_pool.c +++ b/drivers/gpu/drm/ttm/ttm_pool.c @@ -150,8 +150,10 @@ static struct page *ttm_pool_alloc_page(struct ttm_poo= l *pool, gfp_t gfp_flags, =20 =09if (!pool->use_dma_alloc) { =09=09p =3D alloc_pages_node(pool->nid, gfp_flags, order); -=09=09if (p) +=09=09if (p) { =09=09=09p->private =3D order; +=09=09=09mod_node_page_state(NODE_DATA(page_to_nid(p)), NR_GPU_ACTIVE, (1 = << order)); +=09=09} =09=09return p; =09} =20 @@ -186,7 +188,7 @@ static struct page *ttm_pool_alloc_page(struct ttm_pool= *pool, gfp_t gfp_flags, =20 /* Reset the caching and pages of size 1 << order */ static void ttm_pool_free_page(struct ttm_pool *pool, enum ttm_caching cac= hing, -=09=09=09 unsigned int order, struct page *p) +=09=09=09 unsigned int order, struct page *p, bool reclaim) { =09unsigned long attr =3D DMA_ATTR_FORCE_CONTIGUOUS; =09struct ttm_pool_dma *dma; @@ -201,6 +203,9 @@ static void ttm_pool_free_page(struct ttm_pool *pool, e= num ttm_caching caching, #endif =20 =09if (!pool || !pool->use_dma_alloc) { +=09=09mod_node_page_state(NODE_DATA(page_to_nid(p)), +=09=09=09=09 reclaim ? NR_GPU_RECLAIM : NR_GPU_ACTIVE, +=09=09=09=09 -(1 << order)); =09=09__free_pages(p, order); =09=09return; =09} @@ -276,6 +281,7 @@ static void ttm_pool_unmap(struct ttm_pool *pool, dma_a= ddr_t dma_addr, static void ttm_pool_type_give(struct ttm_pool_type *pt, struct page *p) { =09unsigned int i, num_pages =3D 1 << pt->order; +=09int nid =3D page_to_nid(p); =20 =09for (i =3D 0; i < num_pages; ++i) { =09=09if (PageHighMem(p)) @@ -288,17 +294,24 @@ static void ttm_pool_type_give(struct ttm_pool_type *= pt, struct page *p) =09list_add(&p->lru, &pt->pages); =09spin_unlock(&pt->lock); =09atomic_long_add(1 << pt->order, &allocated_pages); + +=09mod_node_page_state(NODE_DATA(nid), NR_GPU_ACTIVE, -num_pages); +=09mod_node_page_state(NODE_DATA(nid), NR_GPU_RECLAIM, num_pages); } =20 /* Take pages from a specific pool_type, return NULL when nothing availabl= e */ static struct page *ttm_pool_type_take(struct ttm_pool_type *pt) { =09struct page *p; +=09int nid; =20 =09spin_lock(&pt->lock); =09p =3D list_first_entry_or_null(&pt->pages, typeof(*p), lru); =09if (p) { +=09=09nid =3D page_to_nid(p); =09=09atomic_long_sub(1 << pt->order, &allocated_pages); +=09=09mod_node_page_state(NODE_DATA(nid), NR_GPU_ACTIVE, (1 << pt->order))= ; +=09=09mod_node_page_state(NODE_DATA(nid), NR_GPU_RECLAIM, -(1 << pt->order= )); =09=09list_del(&p->lru); =09} =09spin_unlock(&pt->lock); @@ -331,7 +344,7 @@ static void ttm_pool_type_fini(struct ttm_pool_type *pt= ) =09spin_unlock(&shrinker_lock); =20 =09while ((p =3D ttm_pool_type_take(pt))) -=09=09ttm_pool_free_page(pt->pool, pt->caching, pt->order, p); +=09=09ttm_pool_free_page(pt->pool, pt->caching, pt->order, p, true); } =20 /* Return the pool_type to use for the given caching and order */ @@ -383,7 +396,7 @@ static unsigned int ttm_pool_shrink(void) =20 =09p =3D ttm_pool_type_take(pt); =09if (p) { -=09=09ttm_pool_free_page(pt->pool, pt->caching, pt->order, p); +=09=09ttm_pool_free_page(pt->pool, pt->caching, pt->order, p, true); =09=09num_pages =3D 1 << pt->order; =09} else { =09=09num_pages =3D 0; @@ -475,7 +488,7 @@ static pgoff_t ttm_pool_unmap_and_free(struct ttm_pool = *pool, struct page *page, =09if (pt) =09=09ttm_pool_type_give(pt, page); =09else -=09=09ttm_pool_free_page(pool, caching, order, page); +=09=09ttm_pool_free_page(pool, caching, order, page, false); =20 =09return nr; } @@ -780,7 +793,7 @@ static int __ttm_pool_alloc(struct ttm_pool *pool, stru= ct ttm_tt *tt, =09return 0; =20 error_free_page: -=09ttm_pool_free_page(pool, page_caching, order, p); +=09ttm_pool_free_page(pool, page_caching, order, p, false); =20 error_free_all: =09if (tt->restore) --=20 2.49.0