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 E493DC77B75 for ; Wed, 10 May 2023 00:50:03 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 835BB10E1F1; Wed, 10 May 2023 00:50:03 +0000 (UTC) Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by gabe.freedesktop.org (Postfix) with ESMTPS id 4275E10E1F1 for ; Wed, 10 May 2023 00:50:02 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1683679802; x=1715215802; h=date:from:to:cc:subject:message-id:references: in-reply-to:mime-version; bh=mpmPHxHEvYtAVRREYfwMjof4+NK6CRcRhVmwZ0oIN/M=; b=f94kkVuiKdxG6VcsmagrDDLmf1SDxFaaY+wm6/vAzUDoJGin4aMxmwSL womBIiEZ4jQdqoA7aO6gdTDB0DqzAxc4O41OPqF5zhQMyJ3m9d1bnwaYw THm6DdJ6Cw69z7sGudSB0rTkay9lNzw/X7USYvM3MmV+0y7wa+kKlhSEI sISwU9IC/Y/EXdX8G6CwXKfFKZxH9FN1y2iVa7WZFkH3EIw9s9UiTqcyP fLwsesGkCouSzMgFlCrnzuNORQsTCP7icCuydP555rIdwR071YE8RJbJA UoG4Tl4O98j5GnQP8ufjvE/N5fCIrf9uGRJEIcwulT3LP3qs9Bf3oAok2 g==; X-IronPort-AV: E=McAfee;i="6600,9927,10705"; a="352272712" X-IronPort-AV: E=Sophos;i="5.99,263,1677571200"; d="scan'208";a="352272712" Received: from fmsmga004.fm.intel.com ([10.253.24.48]) by orsmga102.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 May 2023 17:50:01 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10705"; a="768730571" X-IronPort-AV: E=Sophos;i="5.99,263,1677571200"; d="scan'208";a="768730571" Received: from fmsmsx602.amr.corp.intel.com ([10.18.126.82]) by fmsmga004.fm.intel.com with ESMTP; 09 May 2023 17:50:01 -0700 Received: from fmsmsx612.amr.corp.intel.com (10.18.126.92) by fmsmsx602.amr.corp.intel.com (10.18.126.82) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.23; Tue, 9 May 2023 17:50:01 -0700 Received: from fmsmsx610.amr.corp.intel.com (10.18.126.90) by fmsmsx612.amr.corp.intel.com (10.18.126.92) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.23; Tue, 9 May 2023 17:50:00 -0700 Received: from fmsedg602.ED.cps.intel.com (10.1.192.136) by fmsmsx610.amr.corp.intel.com (10.18.126.90) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.23 via Frontend Transport; Tue, 9 May 2023 17:50:00 -0700 Received: from NAM10-DM6-obe.outbound.protection.outlook.com (104.47.58.103) by edgegateway.intel.com (192.55.55.71) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.1.2507.23; Tue, 9 May 2023 17:50:00 -0700 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=OLd86tlF7sYEygESh3wBbS5JghAaBLkzCnc99RKSE4Fj4OThfwxQht9BzND8gMAb6VRjwp/SP0uhKgz68iP6EBx8/GQSMD4Xk9ZQ0XuLY+RgIm4hS5sBVpIe5GfhAz44PhgkvHbsZaeSsByEbrXReJw94+MXz3Wrj1+WadPdcb+zRdYxbYQzgTHBd1CoGmjh1y8moPmASxUg86psm5qGTaE1iLDZbGXmFsGMb/BRQSNhR0BL67GqaABfL791ss7ghulvwbfzHWSKi3BXLwOoywSS/eFnxiCLqvu3WvI1WMf9RjtEfDZ06GT5uz2I3CpQJPffXFug0OvHx2IUu3F9Ng== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=fo/f6kjZKanTkcSrD0EELh8+qME5cx9M+KVYSPIriyQ=; b=dRLg8foA4DeX038i2O7yZEzOVvRk9X6CMq1VrBMftNm3eW+vIdsiX+SeaMYMAoHUzFw5x6ql3n/wbqi+iudvMExX2Weo5Ku6fvUwymP2/LNb71tNvpwmS6BedYyTYB+D7gIAi2TpX4uoYK8bf1NXXvMikAuZ/y+usucLzZ0J2tAx30+pa/uVN0medvYNGCB2qkUeSCSL8NwGckRfXDvCJhjtfw3MyhH18IVFK7H+574lOdN2BTKM35Kj9G8l/m3rrJRG2rdiQag/TXOmY3mpswM+pK7VTN568m7BI0DLuT2r4uPu9duYmQWnBJqM8i9IubZTOkaDalER9xlnlbIYoA== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=intel.com; dmarc=pass action=none header.from=intel.com; dkim=pass header.d=intel.com; arc=none Authentication-Results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=intel.com; Received: from PH7PR11MB6522.namprd11.prod.outlook.com (2603:10b6:510:212::12) by PH7PR11MB6356.namprd11.prod.outlook.com (2603:10b6:510:1fc::10) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.6363.22; Wed, 10 May 2023 00:49:57 +0000 Received: from PH7PR11MB6522.namprd11.prod.outlook.com ([fe80::ff06:a115:e4eb:680e]) by PH7PR11MB6522.namprd11.prod.outlook.com ([fe80::ff06:a115:e4eb:680e%4]) with mapi id 15.20.6363.032; Wed, 10 May 2023 00:49:57 +0000 Date: Wed, 10 May 2023 00:49:21 +0000 From: Matthew Brost To: Rodrigo Vivi Message-ID: References: <20230502001727.3211096-1-matthew.brost@intel.com> <20230502001727.3211096-12-matthew.brost@intel.com> Content-Type: text/plain; charset="us-ascii" Content-Disposition: inline In-Reply-To: X-ClientProxiedBy: BY5PR04CA0026.namprd04.prod.outlook.com (2603:10b6:a03:1d0::36) To PH7PR11MB6522.namprd11.prod.outlook.com (2603:10b6:510:212::12) MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: PH7PR11MB6522:EE_|PH7PR11MB6356:EE_ X-MS-Office365-Filtering-Correlation-Id: a77ec642-02b3-4ae6-a688-08db50f079e3 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:PH7PR11MB6522.namprd11.prod.outlook.com; PTR:; CAT:NONE; SFS:(13230028)(396003)(346002)(366004)(136003)(39860400002)(376002)(451199021)(4326008)(478600001)(86362001)(66476007)(66556008)(6486002)(6636002)(6666004)(316002)(5660300002)(44832011)(8936002)(6862004)(2906002)(41300700001)(8676002)(186003)(38100700002)(82960400001)(6512007)(6506007)(26005)(83380400001)(66946007)(30864003)(66899021)(21314003)(67856001); DIR:OUT; SFP:1102; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: =?us-ascii?Q?ot72+myRZmyRoPTLzPdRcYqsN5OFPfvQ9yjT9wBN2Shxnwm3Ogo3vkPj57qC?= =?us-ascii?Q?N9d/fUajwAAywJYrczpC/EZWlDPQKGkLsUInlU6xPKTMtZiSRkSVzFxTAw+l?= =?us-ascii?Q?wl54VTQ3AMj4NpHlDswWVBdXYAy7l7wHEkoZPmwDtl9cuphIOypvhMNQ9CsH?= =?us-ascii?Q?tvzmIac0DS2mXuInZzmd46FP32r+TtxUg4jYgOsZKQ3e7MQ/Aq0a9T+MbvsG?= =?us-ascii?Q?CnOmOO4wEO01nQ7Hm1WDy2qTr1BdLpXT5sL5xihV7ddcfBAfrgZt+B7C4Hyg?= =?us-ascii?Q?c/rj22VK5d2iOnKrAH4s7ZGzIzrLR9jGhZcOEPMpunj6SvwOVCp+9LTiAMAI?= =?us-ascii?Q?nKDk0mNcatC3o250Jqx1cmlDzjDhxkhtgs5Eoh+8Q3Qz1mD4WgQWCimoSG0K?= =?us-ascii?Q?ZvcNjy7nd39bvtkiC92tSphYxtT4sce99CozuFi1F4KiHEVGLsyffdCh+dlL?= =?us-ascii?Q?H9ZuVUpI9geRxBHrzymkBgdj0ELNTJyLMJ08ENNfZUkw9KNqNtDBe8DUQaVW?= =?us-ascii?Q?SQVswrYkDLrCTrQ4yJ7aqPiT8GDiJ2u4il/7g6W1Iao0hB1QGs9xKa+9cwvr?= =?us-ascii?Q?asVDcEIAIF5cz+WjBWcezeeJofp7WXlxL1Ux43UXtYb8TTIy5EiexBvWNdsE?= =?us-ascii?Q?2ZY76TQ29IfdBbjgj6V1Yfu7HvxNqZaarNYQVBZxx2rDlCa9jPzeffVr+n4K?= =?us-ascii?Q?DNDDXtuIHjoawKQ6/9PVwnlF+3xmVEGQ1HVmu1xeVpEl2iS6HcIo92ca/O09?= =?us-ascii?Q?kUb4HBcOlnJbWjIOzWmk7ml/TFNpU6QGQO9/ap6NmRCGI1dabQAkQMvB8ZZY?= =?us-ascii?Q?NTfz7sMYM+3AReb+oHtyiZZFECUqa3z/spY2+PAYpj/wfSyAzpgIh37TdfAh?= =?us-ascii?Q?TIdf1HEIlwpz+67Vga+/KnIpVdJ6ADEZQE9LLIuyjEySzzvfEkU1C5akeoLc?= =?us-ascii?Q?qPfsI1PdFcyMbiBPGnSsgnid9AmKdsFhCoGFytfeAVsA9LFLSo60qnW0X7Q+?= =?us-ascii?Q?s4NKpnaL6bt2Qp/kZZss0PWYzLUqMiA145+cq9yWH239IvEliYavDJO6azjO?= =?us-ascii?Q?6afiVtOrguuDWn/KB7yKvpHLaX9rkYFk/3z0o/FwjO/3Ssf7GoQESaVM75xa?= =?us-ascii?Q?8fEM5ms3l5ypGb5r+isFvrraS5BW4Sz3Bpca+ykZJbOEjBqX5V7P2OPoKbCy?= =?us-ascii?Q?EHKNERbnISkuBcAQcaoDO+hpbc489JN/L2345yRZCavfrXHp1uI/+EBNywbA?= =?us-ascii?Q?WGBRkoTb01MiVmYbd3o0NVEyoPM/KIYLSag5TiyzDQiLklfEfwn/jxJLS7lH?= =?us-ascii?Q?gr2nHrdC42DK0+PRfxwr0rxNxEn92rl0V+ADV36TTWoQQod3gvS5vsn83R2e?= =?us-ascii?Q?f4a6WEiGY5QcBurtNev4AT03S46RYj8/S8D9+KLqr0/Y/3vR7hxBY/tRq0bc?= =?us-ascii?Q?eQzfMWvsrtDzsyMODKvhjqsB/rdBvRcEpDk9CkODQsjCXb2yvu/7V/xyGxs9?= =?us-ascii?Q?lGe2Z3Oma/7ljkdWh4/rwXA9hjedDOxRv9M1d9JezVzRVZiHhwvCPPfIlEjl?= =?us-ascii?Q?C5v1aME2Xq6/so/I4Wd06fowPFuNA0HS8JWsDNkMIBIUwaJth5lBcY6i2+Aj?= =?us-ascii?Q?ZQ=3D=3D?= X-MS-Exchange-CrossTenant-Network-Message-Id: a77ec642-02b3-4ae6-a688-08db50f079e3 X-MS-Exchange-CrossTenant-AuthSource: PH7PR11MB6522.namprd11.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 10 May 2023 00:49:57.3292 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 46c98d88-e344-4ed4-8496-4ed7712e255d X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: ol8LrFlQjwX235r96T3+gOdVF2TW+b8rfmmdtSWv3ZO/OTCjX+OitdldM4WCnYmjOUoZ9WUUglqlfDo/u61vtA== X-MS-Exchange-Transport-CrossTenantHeadersStamped: PH7PR11MB6356 X-OriginatorOrg: intel.com Subject: Re: [Intel-xe] [PATCH v2 11/31] drm/xe/guc: Use doorbells for submission if possible X-BeenThere: intel-xe@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel Xe graphics driver List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: intel-xe@lists.freedesktop.org, Faith Ekstrand Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" On Mon, May 08, 2023 at 05:42:05PM -0400, Rodrigo Vivi wrote: > On Mon, May 01, 2023 at 05:17:07PM -0700, Matthew Brost wrote: > > We have 256 doorbells (on most platforms) that we can allocate to bypass > > using the H2G channel for submission. This will avoid contention on the > > CT mutex. > > > > Signed-off-by: Matthew Brost > > Suggested-by: Faith Ekstrand > > --- > > drivers/gpu/drm/xe/regs/xe_guc_regs.h | 1 + > > drivers/gpu/drm/xe/xe_guc.c | 6 + > > drivers/gpu/drm/xe/xe_guc_engine_types.h | 7 + > > drivers/gpu/drm/xe/xe_guc_submit.c | 295 ++++++++++++++++++++++- > > drivers/gpu/drm/xe/xe_guc_submit.h | 1 + > > drivers/gpu/drm/xe/xe_guc_types.h | 4 + > > drivers/gpu/drm/xe/xe_trace.h | 5 + > > 7 files changed, 315 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/gpu/drm/xe/regs/xe_guc_regs.h b/drivers/gpu/drm/xe/regs/xe_guc_regs.h > > index 37e0ac550931..11b117293a62 100644 > > --- a/drivers/gpu/drm/xe/regs/xe_guc_regs.h > > +++ b/drivers/gpu/drm/xe/regs/xe_guc_regs.h > > @@ -109,6 +109,7 @@ struct guc_doorbell_info { > > > > #define DIST_DBS_POPULATED XE_REG(0xd08) > > #define DOORBELLS_PER_SQIDI_MASK REG_GENMASK(23, 16) > > +#define DOORBELLS_PER_SQIDI_SHIFT 16 > > #define SQIDIS_DOORBELL_EXIST_MASK REG_GENMASK(15, 0) > > > > #define GUC_BCS_RCS_IER XE_REG(0xC550) > > diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c > > index 89d20faced19..0c87f78a868b 100644 > > --- a/drivers/gpu/drm/xe/xe_guc.c > > +++ b/drivers/gpu/drm/xe/xe_guc.c > > @@ -297,6 +297,12 @@ int xe_guc_init(struct xe_guc *guc) > > */ > > int xe_guc_init_post_hwconfig(struct xe_guc *guc) > > { > > + int ret; > > + > > + ret = xe_guc_submit_init_post_hwconfig(guc); > > + if (ret) > > + return ret; > > + > > return xe_guc_ads_init_post_hwconfig(&guc->ads); > > } > > > > diff --git a/drivers/gpu/drm/xe/xe_guc_engine_types.h b/drivers/gpu/drm/xe/xe_guc_engine_types.h > > index 5d83132034a6..420b7f53e649 100644 > > --- a/drivers/gpu/drm/xe/xe_guc_engine_types.h > > +++ b/drivers/gpu/drm/xe/xe_guc_engine_types.h > > @@ -12,6 +12,7 @@ > > #include > > > > struct dma_fence; > > +struct xe_bo; > > struct xe_engine; > > > > /** > > @@ -37,6 +38,10 @@ struct xe_guc_engine { > > struct work_struct fini_async; > > /** @resume_time: time of last resume */ > > u64 resume_time; > > + /** @doorbell_bo: BO for memory doorbell */ > > + struct xe_bo *doorbell_bo; > > + /** @doorbell_offset: MMIO doorbell offset */ > > + u32 doorbell_offset; > > /** @state: GuC specific state for this xe_engine */ > > atomic_t state; > > /** @wqi_head: work queue item tail */ > > @@ -45,6 +50,8 @@ struct xe_guc_engine { > > u32 wqi_tail; > > /** @id: GuC id for this xe_engine */ > > u16 id; > > + /** @doorbell_id: doorbell id */ > > + u16 doorbell_id; > > /** @suspend_wait: wait queue used to wait on pending suspends */ > > wait_queue_head_t suspend_wait; > > /** @suspend_pending: a suspend of the engine is pending */ > > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c > > index 0a41f5d04f6d..1b6f36b04cd1 100644 > > --- a/drivers/gpu/drm/xe/xe_guc_submit.c > > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c > > @@ -13,7 +13,10 @@ > > > > #include > > > > +#include "regs/xe_guc_regs.h" > > #include "regs/xe_lrc_layout.h" > > + > > +#include "xe_bo.h" > > #include "xe_device.h" > > #include "xe_engine.h" > > #include "xe_force_wake.h" > > @@ -26,12 +29,22 @@ > > #include "xe_lrc.h" > > #include "xe_macros.h" > > #include "xe_map.h" > > +#include "xe_mmio.h" > > #include "xe_mocs.h" > > #include "xe_ring_ops_types.h" > > #include "xe_sched_job.h" > > #include "xe_trace.h" > > #include "xe_vm.h" > > > > +#define HAS_GUC_MMIO_DB(xe) (IS_DGFX(xe) || GRAPHICS_VERx100(xe) >= 1250) > > +#define HAS_GUC_DIST_DB(xe) \ > > + (GRAPHICS_VERx100(xe) >= 1200 && !HAS_GUC_MMIO_DB(xe)) > > + > > +#define GUC_NUM_HW_DOORBELLS 256 > > + > > +#define GUC_MMIO_DB_BAR_OFFSET SZ_4M > > +#define GUC_MMIO_DB_BAR_SIZE SZ_4M > > + > > static struct xe_gt * > > guc_to_gt(struct xe_guc *guc) > > { > > @@ -63,6 +76,7 @@ engine_to_guc(struct xe_engine *e) > > #define ENGINE_STATE_SUSPENDED (1 << 5) > > #define ENGINE_STATE_RESET (1 << 6) > > #define ENGINE_STATE_KILLED (1 << 7) > > +#define ENGINE_STATE_DB_REGISTERED (1 << 8) > > > > static bool engine_registered(struct xe_engine *e) > > { > > @@ -179,6 +193,16 @@ static void set_engine_killed(struct xe_engine *e) > > atomic_or(ENGINE_STATE_KILLED, &e->guc->state); > > } > > > > +static bool engine_doorbell_registered(struct xe_engine *e) > > +{ > > + return atomic_read(&e->guc->state) & ENGINE_STATE_DB_REGISTERED; > > +} > > + > > +static void set_engine_doorbell_registered(struct xe_engine *e) > > +{ > > + atomic_or(ENGINE_STATE_DB_REGISTERED, &e->guc->state); > > +} > > + > > static bool engine_killed_or_banned(struct xe_engine *e) > > { > > return engine_killed(e) || engine_banned(e); > > @@ -190,6 +214,7 @@ static void guc_submit_fini(struct drm_device *drm, void *arg) > > > > xa_destroy(&guc->submission_state.engine_lookup); > > ida_destroy(&guc->submission_state.guc_ids); > > + ida_destroy(&guc->submission_state.doorbell_ids); > > bitmap_free(guc->submission_state.guc_ids_bitmap); > > } > > > > @@ -230,6 +255,7 @@ int xe_guc_submit_init(struct xe_guc *guc) > > mutex_init(&guc->submission_state.lock); > > xa_init(&guc->submission_state.engine_lookup); > > ida_init(&guc->submission_state.guc_ids); > > + ida_init(&guc->submission_state.doorbell_ids); > > > > spin_lock_init(&guc->submission_state.suspend.lock); > > guc->submission_state.suspend.context = dma_fence_context_alloc(1); > > @@ -243,6 +269,237 @@ int xe_guc_submit_init(struct xe_guc *guc) > > return 0; > > } > > > > +int xe_guc_submit_init_post_hwconfig(struct xe_guc *guc) > > +{ > > + if (HAS_GUC_DIST_DB(guc_to_xe(guc))) { > > + u32 distdbreg = xe_mmio_read32(guc_to_gt(guc), > > + DIST_DBS_POPULATED.reg); > > + u32 num_sqidi = > > + hweight32(distdbreg & SQIDIS_DOORBELL_EXIST_MASK); > > + u32 doorbells_per_sqidi = > > + ((distdbreg >> DOORBELLS_PER_SQIDI_SHIFT) & > > + DOORBELLS_PER_SQIDI_MASK) + 1; > > + > > + guc->submission_state.num_doorbells = > > + num_sqidi * doorbells_per_sqidi; > > + } else { > > + guc->submission_state.num_doorbells = GUC_NUM_HW_DOORBELLS; > > + } > > + > > + return 0; > > +} > > + > > +static bool alloc_doorbell_id(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + int ret; > > + > > + lockdep_assert_held(&guc->submission_state.lock); > > + > > + e->guc->doorbell_id = GUC_NUM_HW_DOORBELLS; > > + ret = ida_simple_get(&guc->submission_state.doorbell_ids, 0, > > + guc->submission_state.num_doorbells, GFP_NOWAIT); > > + if (ret < 0) > > + return false; > > + > > + e->guc->doorbell_id = ret; > > + > > + return true; > > +} > > + > > +static void release_doorbell_id(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + mutex_lock(&guc->submission_state.lock); > > + ida_simple_remove(&guc->submission_state.doorbell_ids, > > + e->guc->doorbell_id); > > + mutex_unlock(&guc->submission_state.lock); > > + > > + e->guc->doorbell_id = GUC_NUM_HW_DOORBELLS; > > +} > > + > > +static int allocate_doorbell(struct xe_guc *guc, u16 guc_id, u16 doorbell_id, > > + u64 gpa, u32 gtt_addr) > > +{ > > + u32 action[] = { > > + XE_GUC_ACTION_ALLOCATE_DOORBELL, > > + guc_id, > > + doorbell_id, > > + lower_32_bits(gpa), > > + upper_32_bits(gpa), > > + gtt_addr > > + }; > > + > > + return xe_guc_ct_send_block(&guc->ct, action, ARRAY_SIZE(action)); > > +} > > + > > +static void deallocate_doorbell(struct xe_guc *guc, u16 guc_id) > > +{ > > + u32 action[] = { > > + XE_GUC_ACTION_DEALLOCATE_DOORBELL, > > + guc_id > > + }; > > + > > + xe_guc_ct_send(&guc->ct, action, ARRAY_SIZE(action), 0, 0); > > +} > > + > > +static bool has_doorbell(struct xe_engine *e) > > +{ > > + return e->guc->doorbell_id != GUC_NUM_HW_DOORBELLS; > > +} > > + > > +#define doorbell_read(guc_, e_, field_) ({ \ > > + struct iosys_map _vmap = (e_)->guc->doorbell_bo->vmap; \ > > + iosys_map_incr(&_vmap, (e_)->guc->doorbell_offset); \ > > + xe_map_rd_field(guc_to_xe((guc_)), &_vmap, 0, \ > > + struct guc_doorbell_info, field_); \ > > + }) > > +#define doorbell_write(guc_, e_, field_, val_) ({ \ > > + struct iosys_map _vmap = (e_)->guc->doorbell_bo->vmap; \ > > + iosys_map_incr(&_vmap, (e_)->guc->doorbell_offset); \ > > + xe_map_wr_field(guc_to_xe((guc_)), &_vmap, 0, \ > > + struct guc_doorbell_info, field_, val_); \ > > + }) > > + > > +static void init_doorbell(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + struct xe_device *xe = guc_to_xe(guc); > > + > > + /* GuC does the initialization with distributed and MMIO doorbells */ > > + if (!HAS_GUC_DIST_DB(xe) && !HAS_GUC_MMIO_DB(xe)) { > > + doorbell_write(guc, e, db_status, GUC_DOORBELL_ENABLED); > > + doorbell_write(guc, e, cookie, 0); > > + } > > +} > > + > > +static void fini_doorbell(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + if (!HAS_GUC_MMIO_DB(guc_to_xe(guc)) && > > + xe_device_mem_access_ongoing(guc_to_xe(guc))) > > + doorbell_write(guc, e, db_status, GUC_DOORBELL_DISABLED); > > +} > > + > > +static void destroy_doorbell(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + if (has_doorbell(e)) { > > + release_doorbell_id(guc, e); > > + xe_bo_unpin_map_no_vm(e->guc->doorbell_bo); > > + } > > +} > > + > > +static void ring_memory_doorbell(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + u32 cookie; > > + > > + cookie = doorbell_read(guc, e, cookie); > > + doorbell_write(guc, e, cookie, cookie + 1 ?: cookie + 2); > > + > > + XE_WARN_ON(doorbell_read(guc, e, db_status) != GUC_DOORBELL_ENABLED); > > +} > > + > > +#define GUC_MMIO_DOORBELL_RING_ACK 0xACEDBEEF > > +#define GUC_MMIO_DOORBELL_RING_NACK 0xDEADBEEF > > Is this a guc abi? should it be in the guc abi files? > Probably? I can move this. > I feel that we need someone with deeper guc knowledge on this review > although based on what I followed on the discussion with Faith and others > it looks like a good move in general. > We probably need Daniele or JOhn H. to review this one. Matt > > +static void ring_mmio_doorbell(struct xe_guc *guc, u32 doorbell_offset) > > +{ > > + u32 db_value; > > + > > + db_value = xe_mmio_read32(guc_to_gt(guc), GUC_MMIO_DB_BAR_OFFSET + > > + doorbell_offset); > > + > > + /* > > + * The read from the doorbell page will return ack/nack. We don't remove > > + * doorbells from active clients so we don't expect to ever get a nack. > > + * XXX: if doorbell is lost, re-acquire it? > > + */ > > + XE_WARN_ON(db_value == GUC_MMIO_DOORBELL_RING_NACK); > > + XE_WARN_ON(db_value != GUC_MMIO_DOORBELL_RING_ACK); > > +} > > + > > +static void ring_doorbell(struct xe_guc *guc, struct xe_engine *e) > > +{ > > + XE_BUG_ON(!has_doorbell(e)); > > + > > + if (HAS_GUC_MMIO_DB(guc_to_xe(guc))) > > + ring_mmio_doorbell(guc, e->guc->doorbell_offset); > > + else > > + ring_memory_doorbell(guc, e); > > + > > + trace_xe_engine_ring_db(e); > > +} > > + > > +static void register_engine(struct xe_engine *e); > > + > > +static int create_doorbell(struct xe_guc *guc, struct xe_engine *e, bool init) > > +{ > > + struct xe_gt *gt = guc_to_gt(guc); > > + struct xe_device *xe = gt_to_xe(gt); > > + u64 gpa; > > + u32 gtt_addr; > > + int ret; > > + > > + XE_BUG_ON(!has_doorbell(e)); > > + > > + if (HAS_GUC_MMIO_DB(xe)) { > > + e->guc->doorbell_offset = PAGE_SIZE * e->guc->doorbell_id; > > + gpa = GUC_MMIO_DB_BAR_OFFSET + e->guc->doorbell_offset; > > + gtt_addr = 0; > > + } else { > > + struct xe_bo *bo; > > + > > + if (!e->guc->doorbell_bo) { > > + bo = xe_bo_create_pin_map(xe, gt, NULL, PAGE_SIZE, > > + ttm_bo_type_kernel, > > + XE_BO_CREATE_VRAM_IF_DGFX(gt) | > > + XE_BO_CREATE_GGTT_BIT); > > + if (IS_ERR(bo)) > > + return PTR_ERR(bo); > > + > > + e->guc->doorbell_bo = bo; > > + } else { > > + bo = e->guc->doorbell_bo; > > + } > > + > > + init_doorbell(guc, e); > > + gpa = xe_bo_main_addr(bo, PAGE_SIZE); > > + gtt_addr = xe_bo_ggtt_addr(bo); > > + } > > + > > + if (init && e->flags & ENGINE_FLAG_KERNEL) > > + return 0; > > + > > + register_engine(e); > > + ret = allocate_doorbell(guc, e->guc->id, e->guc->doorbell_id, gpa, > > + gtt_addr); > > + if (ret < 0) { > > + fini_doorbell(guc, e); > > + return ret; > > + } > > + > > + /* > > + * In distributed doorbells, guc is returning the cacheline selected > > + * by HW as part of the 7bit data from the allocate doorbell command: > > + * bit [22] - Cacheline allocated > > + * bit [21:16] - Cacheline offset address > > + * (bit 21 must be zero, or our assumption of only using half a page is > > + * no longer correct). > > + */ > > + if (HAS_GUC_DIST_DB(xe)) { > > + u32 dd_cacheline_info; > > + > > + XE_WARN_ON(!(ret & BIT(22))); > > + XE_WARN_ON(ret & BIT(21)); > > + > > + dd_cacheline_info = FIELD_GET(GENMASK(21, 16), ret); > > + e->guc->doorbell_offset = dd_cacheline_info * cache_line_size(); > > + > > + /* and verify db status was updated correctly by the guc fw */ > > + XE_WARN_ON(doorbell_read(guc, e, db_status) != > > + GUC_DOORBELL_ENABLED); > > + } > > + > > + set_engine_doorbell_registered(e); > > + > > + return 0; > > +} > > + > > static int alloc_guc_id(struct xe_guc *guc, struct xe_engine *e) > > { > > int ret; > > @@ -623,6 +880,7 @@ static void submit_engine(struct xe_engine *e) > > u32 num_g2h = 0; > > int len = 0; > > bool extra_submit = false; > > + bool enable = false; > > > > XE_BUG_ON(!engine_registered(e)); > > > > @@ -642,6 +900,7 @@ static void submit_engine(struct xe_engine *e) > > num_g2h = 1; > > if (xe_engine_is_parallel(e)) > > extra_submit = true; > > + enable = true; > > > > e->guc->resume_time = RESUME_PENDING; > > set_engine_pending_enable(e); > > @@ -653,7 +912,10 @@ static void submit_engine(struct xe_engine *e) > > trace_xe_engine_submit(e); > > } > > > > - xe_guc_ct_send(&guc->ct, action, len, g2h_len, num_g2h); > > + if (enable || !engine_doorbell_registered(e)) > > + xe_guc_ct_send(&guc->ct, action, len, g2h_len, num_g2h); > > + else > > + ring_doorbell(guc, e); > > > > if (extra_submit) { > > len = 0; > > @@ -678,8 +940,17 @@ guc_engine_run_job(struct drm_sched_job *drm_job) > > trace_xe_sched_job_run(job); > > > > if (!engine_killed_or_banned(e) && !xe_sched_job_is_error(job)) { > > - if (!engine_registered(e)) > > - register_engine(e); > > + if (!engine_registered(e)) { > > + if (has_doorbell(e)) { > > + int err = create_doorbell(engine_to_guc(e), e, > > + false); > > + > > + /* Not fatal, but let's warn */ > > + XE_WARN_ON(err); > > + } else { > > + register_engine(e); > > + } > > + } > > if (!lr) /* Written in IOCTL */ > > e->ring_ops->emit_job(job); > > submit_engine(e); > > @@ -722,6 +993,11 @@ static void disable_scheduling_deregister(struct xe_guc *guc, > > MAKE_SCHED_CONTEXT_ACTION(e, DISABLE); > > int ret; > > > > + if (has_doorbell(e)) { > > + fini_doorbell(guc, e); > > + deallocate_doorbell(guc, e->guc->id); > > + } > > + > > set_min_preemption_timeout(guc, e); > > smp_rmb(); > > ret = wait_event_timeout(guc->ct.wq, !engine_pending_enable(e) || > > @@ -958,6 +1234,7 @@ static void __guc_engine_fini_async(struct work_struct *w) > > cancel_work_sync(&ge->lr_tdr); > > if (e->flags & ENGINE_FLAG_PERSISTENT) > > xe_device_remove_persistent_engines(gt_to_xe(e->gt), e); > > + destroy_doorbell(guc, e); > > release_guc_id(guc, e); > > drm_sched_entity_fini(&ge->entity); > > drm_sched_fini(&ge->sched); > > @@ -1136,6 +1413,7 @@ static int guc_engine_init(struct xe_engine *e) > > struct xe_guc_engine *ge; > > long timeout; > > int err; > > + bool create_db = false; > > > > XE_BUG_ON(!xe_device_guc_submission_enabled(guc_to_xe(guc))); > > > > @@ -1177,8 +1455,17 @@ static int guc_engine_init(struct xe_engine *e) > > if (guc_read_stopped(guc)) > > drm_sched_stop(sched, NULL); > > > > + create_db = alloc_doorbell_id(guc, e); > > + > > mutex_unlock(&guc->submission_state.lock); > > > > + if (create_db) { > > + /* Error isn't fatal as we don't need a doorbell */ > > + err = create_doorbell(guc, e, true); > > + if (err) > > + release_doorbell_id(guc, e); > > + } > > + > > switch (e->class) { > > case XE_ENGINE_CLASS_RENDER: > > sprintf(e->name, "rcs%d", e->guc->id); > > @@ -1302,7 +1589,7 @@ static int guc_engine_set_job_timeout(struct xe_engine *e, u32 job_timeout_ms) > > { > > struct drm_gpu_scheduler *sched = &e->guc->sched; > > > > - XE_BUG_ON(engine_registered(e)); > > + XE_BUG_ON(engine_registered(e) && !has_doorbell(e)); > > XE_BUG_ON(engine_banned(e)); > > XE_BUG_ON(engine_killed(e)); > > > > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.h b/drivers/gpu/drm/xe/xe_guc_submit.h > > index 8002734d6f24..bada6c02d6aa 100644 > > --- a/drivers/gpu/drm/xe/xe_guc_submit.h > > +++ b/drivers/gpu/drm/xe/xe_guc_submit.h > > @@ -13,6 +13,7 @@ struct xe_engine; > > struct xe_guc; > > > > int xe_guc_submit_init(struct xe_guc *guc); > > +int xe_guc_submit_init_post_hwconfig(struct xe_guc *guc); > > void xe_guc_submit_print(struct xe_guc *guc, struct drm_printer *p); > > > > int xe_guc_submit_reset_prepare(struct xe_guc *guc); > > diff --git a/drivers/gpu/drm/xe/xe_guc_types.h b/drivers/gpu/drm/xe/xe_guc_types.h > > index ac7eec28934d..9ee4d572f4e0 100644 > > --- a/drivers/gpu/drm/xe/xe_guc_types.h > > +++ b/drivers/gpu/drm/xe/xe_guc_types.h > > @@ -36,10 +36,14 @@ struct xe_guc { > > struct xarray engine_lookup; > > /** @guc_ids: used to allocate new guc_ids, single-lrc */ > > struct ida guc_ids; > > + /** @doorbell_ids: use to allocate new doorbells */ > > + struct ida doorbell_ids; > > /** @guc_ids_bitmap: used to allocate new guc_ids, multi-lrc */ > > unsigned long *guc_ids_bitmap; > > /** @stopped: submissions are stopped */ > > atomic_t stopped; > > + /** @num_doorbells: number of doorbels */ > > + int num_doorbells; > > /** @lock: protects submission state */ > > struct mutex lock; > > /** @suspend: suspend fence state */ > > diff --git a/drivers/gpu/drm/xe/xe_trace.h b/drivers/gpu/drm/xe/xe_trace.h > > index 02861c26e145..38e9d7c6197b 100644 > > --- a/drivers/gpu/drm/xe/xe_trace.h > > +++ b/drivers/gpu/drm/xe/xe_trace.h > > @@ -149,6 +149,11 @@ DEFINE_EVENT(xe_engine, xe_engine_submit, > > TP_ARGS(e) > > ); > > > > +DEFINE_EVENT(xe_engine, xe_engine_ring_db, > > + TP_PROTO(struct xe_engine *e), > > + TP_ARGS(e) > > +); > > + > > DEFINE_EVENT(xe_engine, xe_engine_scheduling_enable, > > TP_PROTO(struct xe_engine *e), > > TP_ARGS(e) > > -- > > 2.34.1 > >