From patchwork Fri Jan 24 13:01:07 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 11350205 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 29FCD13A4 for ; Fri, 24 Jan 2020 13:01:20 +0000 (UTC) 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 mail.kernel.org (Postfix) with ESMTPS id 11F2D2077C for ; Fri, 24 Jan 2020 13:01:20 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 11F2D2077C Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=chris-wilson.co.uk Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=intel-gfx-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 05A446E373; Fri, 24 Jan 2020 13:01:17 +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 9877A6E02E; Fri, 24 Jan 2020 13:01:14 +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 19995157-1500050 for multiple; Fri, 24 Jan 2020 13:01:08 +0000 From: Chris Wilson To: dri-devel@lists.freedesktop.org Date: Fri, 24 Jan 2020 13:01:07 +0000 Message-Id: <20200124130107.125404-1-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.25.0 In-Reply-To: <20200124125627.125042-2-chris@chris-wilson.co.uk> References: <20200124125627.125042-2-chris@chris-wilson.co.uk> MIME-Version: 1.0 Subject: [Intel-gfx] [PATCH] drm: Avoid drm_global_mutex for simple inc/dec of dev->open_count X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: intel-gfx@lists.freedesktop.org, =?utf-8?q?Thomas_Hellstr=C3=B6m?= Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" Since drm_global_mutex is a true global mutex across devices, we don't want to acquire it unless absolutely necessary. For maintaining the device local open_count, we can use atomic operations on the counter itself, except when making the transition to/from 0. Here, we tackle the easy portion of delaying acquiring the drm_global_mutex for the final release by using atomic_dec_and_mutex_lock(), leaving the global serialisation across the device opens. Signed-off-by: Chris Wilson Cc: Thomas Hellström (VMware) Reviewed-by: Thomas Hellström --- atomic_dec_and_mutex_lock needs pairing with mutex_unlock (you fool) --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- drivers/gpu/drm/drm_file.c | 16 ++++++++-------- drivers/gpu/drm/i915/i915_switcheroo.c | 2 +- drivers/gpu/drm/nouveau/nouveau_vga.c | 2 +- drivers/gpu/drm/radeon/radeon_device.c | 2 +- include/drm/drm_device.h | 2 +- 6 files changed, 13 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 53d882000101..c3c0356dfa61 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1136,7 +1136,7 @@ static bool amdgpu_switcheroo_can_switch(struct pci_dev *pdev) * locking inversion with the driver load path. And the access here is * completely racy anyway. So don't bother with locking for now. */ - return dev->open_count == 0; + return atomic_read(&dev->open_count) == 0; } static const struct vga_switcheroo_client_ops amdgpu_switcheroo_ops = { diff --git a/drivers/gpu/drm/drm_file.c b/drivers/gpu/drm/drm_file.c index e25306c49cc6..1075b3a8b5b1 100644 --- a/drivers/gpu/drm/drm_file.c +++ b/drivers/gpu/drm/drm_file.c @@ -220,7 +220,7 @@ void drm_file_free(struct drm_file *file) DRM_DEBUG("pid = %d, device = 0x%lx, open_count = %d\n", task_pid_nr(current), (long)old_encode_dev(file->minor->kdev->devt), - READ_ONCE(dev->open_count)); + atomic_read(&dev->open_count)); if (drm_core_check_feature(dev, DRIVER_LEGACY) && dev->driver->preclose) @@ -379,7 +379,7 @@ int drm_open(struct inode *inode, struct file *filp) return PTR_ERR(minor); dev = minor->dev; - if (!dev->open_count++) + if (!atomic_fetch_inc(&dev->open_count)) need_setup = 1; /* share address_space across all char-devs of a single device */ @@ -398,7 +398,7 @@ int drm_open(struct inode *inode, struct file *filp) return 0; err_undo: - dev->open_count--; + atomic_dec(&dev->open_count); drm_minor_release(minor); return retcode; } @@ -440,11 +440,11 @@ int drm_release(struct inode *inode, struct file *filp) mutex_lock(&drm_global_mutex); - DRM_DEBUG("open_count = %d\n", dev->open_count); + DRM_DEBUG("open_count = %d\n", atomic_read(&dev->open_count)); drm_close_helper(filp); - if (!--dev->open_count) + if (atomic_dec_and_test(&dev->open_count)) drm_lastclose(dev); mutex_unlock(&drm_global_mutex); @@ -478,10 +478,10 @@ int drm_release_noglobal(struct inode *inode, struct file *filp) drm_close_helper(filp); - mutex_lock(&drm_global_mutex); - if (!--dev->open_count) + if (atomic_dec_and_mutex_lock(&dev->open_count, &drm_global_mutex)) { drm_lastclose(dev); - mutex_unlock(&drm_global_mutex); + mutex_unlock(&drm_global_mutex); + } drm_minor_release(minor); diff --git a/drivers/gpu/drm/i915/i915_switcheroo.c b/drivers/gpu/drm/i915/i915_switcheroo.c index 39c79e1c5b52..ed69b5d4a375 100644 --- a/drivers/gpu/drm/i915/i915_switcheroo.c +++ b/drivers/gpu/drm/i915/i915_switcheroo.c @@ -43,7 +43,7 @@ static bool i915_switcheroo_can_switch(struct pci_dev *pdev) * locking inversion with the driver load path. And the access here is * completely racy anyway. So don't bother with locking for now. */ - return i915 && i915->drm.open_count == 0; + return i915 && atomic_read(&i915->drm.open_count) == 0; } static const struct vga_switcheroo_client_ops i915_switcheroo_ops = { diff --git a/drivers/gpu/drm/nouveau/nouveau_vga.c b/drivers/gpu/drm/nouveau/nouveau_vga.c index d865d8aeac3c..c85dd8afa3c3 100644 --- a/drivers/gpu/drm/nouveau/nouveau_vga.c +++ b/drivers/gpu/drm/nouveau/nouveau_vga.c @@ -72,7 +72,7 @@ nouveau_switcheroo_can_switch(struct pci_dev *pdev) * locking inversion with the driver load path. And the access here is * completely racy anyway. So don't bother with locking for now. */ - return dev->open_count == 0; + return atomic_read(&dev->open_count) == 0; } static const struct vga_switcheroo_client_ops diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index a522e092038b..266e3cbbd09b 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -1263,7 +1263,7 @@ static bool radeon_switcheroo_can_switch(struct pci_dev *pdev) * locking inversion with the driver load path. And the access here is * completely racy anyway. So don't bother with locking for now. */ - return dev->open_count == 0; + return atomic_read(&dev->open_count) == 0; } static const struct vga_switcheroo_client_ops radeon_switcheroo_ops = { diff --git a/include/drm/drm_device.h b/include/drm/drm_device.h index 1acfc3bbd3fb..bb60a949f416 100644 --- a/include/drm/drm_device.h +++ b/include/drm/drm_device.h @@ -144,7 +144,7 @@ struct drm_device { * Usage counter for outstanding files open, * protected by drm_global_mutex */ - int open_count; + atomic_t open_count; /** @filelist_mutex: Protects @filelist. */ struct mutex filelist_mutex;