From patchwork Fri Apr 27 13:08:11 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?q?Michel_D=C3=A4nzer?= X-Patchwork-Id: 10368515 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id A7DD0601D3 for ; Fri, 27 Apr 2018 13:08:19 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 95504293CB for ; Fri, 27 Apr 2018 13:08:19 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 86D8E293D6; Fri, 27 Apr 2018 13:08:19 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-5.2 required=2.0 tests=BAYES_00, MAILING_LIST_MULTI, RCVD_IN_DNSWL_MED autolearn=ham version=3.3.1 Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher DHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id 6C326293CB for ; Fri, 27 Apr 2018 13:08:18 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id E60DC6E8EB; Fri, 27 Apr 2018 13:08:15 +0000 (UTC) X-Original-To: dri-devel@lists.freedesktop.org Delivered-To: dri-devel@lists.freedesktop.org Received: from netline-mail3.netline.ch (mail.netline.ch [148.251.143.178]) by gabe.freedesktop.org (Postfix) with ESMTP id AB1A66E8EB; Fri, 27 Apr 2018 13:08:14 +0000 (UTC) Received: from localhost (localhost [127.0.0.1]) by netline-mail3.netline.ch (Postfix) with ESMTP id E7B932A604D; Fri, 27 Apr 2018 15:08:13 +0200 (CEST) X-Virus-Scanned: Debian amavisd-new at netline-mail3.netline.ch Received: from netline-mail3.netline.ch ([127.0.0.1]) by localhost (netline-mail3.netline.ch [127.0.0.1]) (amavisd-new, port 10024) with LMTP id kdWomtqXoXI0; Fri, 27 Apr 2018 15:08:12 +0200 (CEST) Received: from kaveri (145.233.60.188.dynamic.wline.res.cust.swisscom.ch [188.60.233.145]) by netline-mail3.netline.ch (Postfix) with ESMTPSA id BBCB92A604C; Fri, 27 Apr 2018 15:08:12 +0200 (CEST) Received: from daenzer by kaveri with local (Exim 4.91) (envelope-from ) id 1fC36h-0001zz-NG; Fri, 27 Apr 2018 15:08:11 +0200 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= To: =?UTF-8?q?Christian=20K=C3=B6nig?= Subject: [PATCH v2 1/2] drm/ttm: Only allocate huge pages with new flag TTM_PAGE_FLAG_TRANSHUGE Date: Fri, 27 Apr 2018 15:08:11 +0200 Message-Id: <20180427130811.7642-1-michel@daenzer.net> X-Mailer: git-send-email 2.17.0 In-Reply-To: <20180426150618.13470-1-michel@daenzer.net> References: <20180426150618.13470-1-michel@daenzer.net> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" X-Virus-Scanned: ClamAV using ClamSMTP From: Michel Dänzer Previously, TTM would always (with CONFIG_TRANSPARENT_HUGEPAGE enabled) try to allocate huge pages. However, not all drivers can take advantage of huge pages, but they would incur the overhead for allocating and freeing them anyway. Now, drivers which can take advantage of huge pages need to set the new flag TTM_PAGE_FLAG_TRANSHUGE to get them. Drivers not setting this flag no longer incur any overhead for allocating or freeing huge pages. v2: * Also guard swapping of consecutive pages in ttm_get_pages * Reword commit log, hopefully clearer now Cc: stable@vger.kernel.org Signed-off-by: Michel Dänzer --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +- drivers/gpu/drm/ttm/ttm_page_alloc.c | 35 +++++++++++++++++------- drivers/gpu/drm/ttm/ttm_page_alloc_dma.c | 8 ++++-- include/drm/ttm/ttm_tt.h | 1 + 4 files changed, 32 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index dfd22db13fb1..e03e9e361e2a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -988,7 +988,7 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo, return NULL; } gtt->ttm.ttm.func = &amdgpu_backend_func; - if (ttm_sg_tt_init(>t->ttm, bo, page_flags)) { + if (ttm_sg_tt_init(>t->ttm, bo, page_flags | TTM_PAGE_FLAG_TRANSHUGE)) { kfree(gtt); return NULL; } diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c index f0481b7b60c5..476d668e1cbd 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c @@ -760,7 +760,7 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags, { struct ttm_page_pool *pool = ttm_get_pool(flags, false, cstate); #ifdef CONFIG_TRANSPARENT_HUGEPAGE - struct ttm_page_pool *huge = ttm_get_pool(flags, true, cstate); + struct ttm_page_pool *huge = NULL; #endif unsigned long irq_flags; unsigned i; @@ -780,7 +780,8 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags, } #ifdef CONFIG_TRANSPARENT_HUGEPAGE - if (!(flags & TTM_PAGE_FLAG_DMA32)) { + if ((flags & (TTM_PAGE_FLAG_DMA32 | TTM_PAGE_FLAG_TRANSHUGE)) == + TTM_PAGE_FLAG_TRANSHUGE) { for (j = 0; j < HPAGE_PMD_NR; ++j) if (p++ != pages[i + j]) break; @@ -805,6 +806,8 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags, i = 0; #ifdef CONFIG_TRANSPARENT_HUGEPAGE + if (flags & TTM_PAGE_FLAG_TRANSHUGE) + huge = ttm_get_pool(flags, true, cstate); if (huge) { unsigned max_size, n2free; @@ -877,7 +880,7 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, { struct ttm_page_pool *pool = ttm_get_pool(flags, false, cstate); #ifdef CONFIG_TRANSPARENT_HUGEPAGE - struct ttm_page_pool *huge = ttm_get_pool(flags, true, cstate); + struct ttm_page_pool *huge = NULL; #endif struct list_head plist; struct page *p = NULL; @@ -906,7 +909,8 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, i = 0; #ifdef CONFIG_TRANSPARENT_HUGEPAGE - if (!(gfp_flags & GFP_DMA32)) { + if ((flags & (TTM_PAGE_FLAG_DMA32 | TTM_PAGE_FLAG_TRANSHUGE)) == + TTM_PAGE_FLAG_TRANSHUGE) { while (npages >= HPAGE_PMD_NR) { gfp_t huge_flags = gfp_flags; @@ -933,9 +937,13 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, return -ENOMEM; } - /* Swap the pages if we detect consecutive order */ - if (i > first && pages[i - 1] == p - 1) - swap(p, pages[i - 1]); +#ifdef CONFIG_TRANSPARENT_HUGEPAGE + if (flags & TTM_PAGE_FLAG_TRANSHUGE) { + /* Swap the pages if we detect consecutive order */ + if (i > first && pages[i - 1] == p - 1) + swap(p, pages[i - 1]); + } +#endif pages[i++] = p; --npages; @@ -946,6 +954,8 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, count = 0; #ifdef CONFIG_TRANSPARENT_HUGEPAGE + if (flags & TTM_PAGE_FLAG_TRANSHUGE) + huge = ttm_get_pool(flags, true, cstate); if (huge && npages >= HPAGE_PMD_NR) { INIT_LIST_HEAD(&plist); ttm_page_pool_get_pages(huge, &plist, flags, cstate, @@ -969,9 +979,14 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags, list_for_each_entry(p, &plist, lru) { struct page *tmp = p; - /* Swap the pages if we detect consecutive order */ - if (count > first && pages[count - 1] == tmp - 1) - swap(tmp, pages[count - 1]); +#ifdef CONFIG_TRANSPARENT_HUGEPAGE + if (flags & TTM_PAGE_FLAG_TRANSHUGE) { + /* Swap the pages if we detect consecutive order */ + if (count > first && pages[count - 1] == tmp - 1) + swap(tmp, pages[count - 1]); + } +#endif + pages[count++] = tmp; } diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c index 8a25d1974385..291b04213ec5 100644 --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c @@ -949,7 +949,8 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev, type = ttm_to_type(ttm->page_flags, ttm->caching_state); #ifdef CONFIG_TRANSPARENT_HUGEPAGE - if (ttm->page_flags & TTM_PAGE_FLAG_DMA32) + if ((ttm->page_flags & (TTM_PAGE_FLAG_DMA32 | TTM_PAGE_FLAG_TRANSHUGE)) + != TTM_PAGE_FLAG_TRANSHUGE) goto skip_huge; pool = ttm_dma_find_pool(dev, type | IS_HUGE); @@ -1035,7 +1036,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev) { struct ttm_tt *ttm = &ttm_dma->ttm; struct ttm_mem_global *mem_glob = ttm->bdev->glob->mem_glob; - struct dma_pool *pool; + struct dma_pool *pool = NULL; struct dma_page *d_page, *next; enum pool_type type; bool is_cached = false; @@ -1045,7 +1046,8 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev) type = ttm_to_type(ttm->page_flags, ttm->caching_state); #ifdef CONFIG_TRANSPARENT_HUGEPAGE - pool = ttm_dma_find_pool(dev, type | IS_HUGE); + if (ttm->page_flags & TTM_PAGE_FLAG_TRANSHUGE) + pool = ttm_dma_find_pool(dev, type | IS_HUGE); if (pool) { count = 0; list_for_each_entry_safe(d_page, next, &ttm_dma->pages_list, diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h index c0e928abf592..c7d2120f0362 100644 --- a/include/drm/ttm/ttm_tt.h +++ b/include/drm/ttm/ttm_tt.h @@ -41,6 +41,7 @@ struct ttm_operation_ctx; #define TTM_PAGE_FLAG_DMA32 (1 << 7) #define TTM_PAGE_FLAG_SG (1 << 8) #define TTM_PAGE_FLAG_NO_RETRY (1 << 9) +#define TTM_PAGE_FLAG_TRANSHUGE (1 << 10) enum ttm_caching_state { tt_uncached,