From patchwork Sun Jun 10 19:43:20 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 10456735 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 6B77F601F7 for ; Sun, 10 Jun 2018 19:44:11 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 56C8B26E79 for ; Sun, 10 Jun 2018 19:44:11 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 4A962274D0; Sun, 10 Jun 2018 19:44:11 +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 A525826E79 for ; Sun, 10 Jun 2018 19:44:10 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id C0DAB6E212; Sun, 10 Jun 2018 19:44:09 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from fireflyinternet.com (mail.fireflyinternet.com [109.228.58.192]) by gabe.freedesktop.org (Postfix) with ESMTPS id ACAFC6E204 for ; Sun, 10 Jun 2018 19:44:06 +0000 (UTC) X-Default-Received-SPF: pass (skip=forwardok (res=PASS)) x-ip-name=78.156.65.138; Received: from haswell.alporthouse.com (unverified [78.156.65.138]) by fireflyinternet.com (Firefly Internet (M1)) with ESMTP id 12004299-1500050 for multiple; Sun, 10 Jun 2018 20:43:59 +0100 Received: by haswell.alporthouse.com (sSMTP sendmail emulation); Sun, 10 Jun 2018 20:43:59 +0100 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Sun, 10 Jun 2018 20:43:20 +0100 Message-Id: <20180610194325.13467-13-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20180610194325.13467-1-chris@chris-wilson.co.uk> References: <20180610194325.13467-1-chris@chris-wilson.co.uk> X-Originating-IP: 78.156.65.138 X-Country: code=GB country="United Kingdom" ip=78.156.65.138 Subject: [Intel-gfx] [PATCH 12/17] drm/i915/gtt: Free unused page tables on unbind the context X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" X-Virus-Scanned: ClamAV using ClamSMTP As we cannot reliably change used page tables while the context is active, the earliest opportunity we have to recover excess pages is when the context becomes idle. So whenever we unbind the context (it must be idle, and indeed being evicted) free the unused ptes. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Cc: Matthew Auld Reviewed-by: Matthew Auld Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_gtt.c | 44 +++++++++++++++++++++++++---- drivers/gpu/drm/i915/i915_gem_gtt.h | 1 + 2 files changed, 40 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 3dc17793afac..8883e75cb594 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1753,20 +1753,28 @@ static void gen6_ppgtt_enable(struct drm_i915_private *dev_priv) static void gen6_ppgtt_clear_range(struct i915_address_space *vm, u64 start, u64 length) { - struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); + struct gen6_hw_ppgtt *ppgtt = to_gen6_ppgtt(i915_vm_to_ppgtt(vm)); unsigned int first_entry = start >> PAGE_SHIFT; unsigned int pde = first_entry / GEN6_PTES; unsigned int pte = first_entry % GEN6_PTES; unsigned int num_entries = length >> PAGE_SHIFT; - gen6_pte_t scratch_pte = + const gen6_pte_t scratch_pte = vm->pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC, 0); while (num_entries) { - struct i915_page_table *pt = ppgtt->pd.page_table[pde++]; - unsigned int end = min(pte + num_entries, GEN6_PTES); + struct i915_page_table *pt = ppgtt->base.pd.page_table[pde++]; + const unsigned int end = min(pte + num_entries, GEN6_PTES); + const unsigned int count = end - pte; gen6_pte_t *vaddr; - num_entries -= end - pte; + GEM_BUG_ON(pt == vm->scratch_pt); + + num_entries -= count; + + GEM_BUG_ON(count > pt->used_ptes); + pt->used_ptes -= count; + if (!pt->used_ptes) + ppgtt->scan_for_unused_pt = true; /* * Note that the hw doesn't support removing PDE on the fly @@ -1798,6 +1806,8 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm, struct sgt_dma iter = sgt_dma(vma); gen6_pte_t *vaddr; + GEM_BUG_ON(ppgtt->pd.page_table[act_pt] == vm->scratch_pt); + vaddr = kmap_atomic_px(ppgtt->pd.page_table[act_pt]); do { vaddr[act_pte] = pte_encode | GEN6_PTE_ADDR_ENCODE(iter.dma); @@ -1833,6 +1843,8 @@ static int gen6_alloc_va_range(struct i915_address_space *vm, bool flush = false; gen6_for_each_pde(pt, &ppgtt->base.pd, start, length, pde) { + const unsigned int count = gen6_pte_count(start, length); + if (pt == vm->scratch_pt) { pt = alloc_pt(vm); if (IS_ERR(pt)) @@ -1846,7 +1858,11 @@ static int gen6_alloc_va_range(struct i915_address_space *vm, gen6_write_pde(ppgtt, pde, pt); flush = true; } + + GEM_BUG_ON(pt->used_ptes); } + + pt->used_ptes += count; } if (flush) { @@ -1948,6 +1964,24 @@ static int pd_vma_bind(struct i915_vma *vma, static void pd_vma_unbind(struct i915_vma *vma) { + struct gen6_hw_ppgtt *ppgtt = vma->private; + struct i915_page_table * const scratch_pt = ppgtt->base.vm.scratch_pt; + struct i915_page_table *pt; + unsigned int pde; + + if (!ppgtt->scan_for_unused_pt) + return; + + /* Free all no longer used page tables */ + gen6_for_all_pdes(pt, &ppgtt->base.pd, pde) { + if (pt->used_ptes || pt == scratch_pt) + continue; + + free_pt(&ppgtt->base.vm, pt); + ppgtt->base.pd.page_table[pde] = scratch_pt; + } + + ppgtt->scan_for_unused_pt = false; } static const struct i915_vma_ops pd_vma_ops = { diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h index d7b7b4afe060..46c95d188580 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.h +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h @@ -414,6 +414,7 @@ struct gen6_hw_ppgtt { gen6_pte_t __iomem *pd_addr; unsigned int pin_count; + bool scan_for_unused_pt; }; #define __to_gen6_ppgtt(base) container_of(base, struct gen6_hw_ppgtt, base)