From patchwork Tue Apr 19 15:08:34 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dave Gordon X-Patchwork-Id: 8881141 Return-Path: X-Original-To: patchwork-intel-gfx@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork2.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.136]) by patchwork2.web.kernel.org (Postfix) with ESMTP id D5DA6BF29F for ; Tue, 19 Apr 2016 15:09:02 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 98377200C6 for ; Tue, 19 Apr 2016 15:09:01 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by mail.kernel.org (Postfix) with ESMTP id 0B9E62026F for ; Tue, 19 Apr 2016 15:08:56 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id B301A6E7CE; Tue, 19 Apr 2016 15:08:53 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by gabe.freedesktop.org (Postfix) with ESMTP id 74BF36E1EA for ; Tue, 19 Apr 2016 15:08:51 +0000 (UTC) Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by fmsmga104.fm.intel.com with ESMTP; 19 Apr 2016 08:08:49 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.24,506,1455004800"; d="scan'208";a="688791039" Received: from dsgordon-linux2.isw.intel.com ([10.102.226.88]) by FMSMGA003.fm.intel.com with ESMTP; 19 Apr 2016 08:08:49 -0700 From: Dave Gordon To: intel-gfx@lists.freedesktop.org Date: Tue, 19 Apr 2016 16:08:34 +0100 Message-Id: <1461078516-28678-1-git-send-email-david.s.gordon@intel.com> X-Mailer: git-send-email 1.9.1 Organization: Intel Corporation (UK) Ltd. - Co. Reg. #1134945 - Pipers Way, Swindon SN3 1RJ Subject: [Intel-gfx] [PATCH 1/3] drm/i915/guc: keep GuC doorbell & process descriptor mapped in kernel X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.18 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-Spam-Status: No, score=-5.2 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_MED, RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=unavailable version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Don't use kmap_atomic() for doorbell & process descriptor access. This patch fixes the BUG shown below, where the thread could sleep while holding a kmap_atomic mapping. In order not to need to call kmap_atomic() in this code path, we now set up a permanent kernel mapping of the shared doorbell and process-descriptor page, and use that in all doorbell and process-descriptor related code. BUG: scheduling while atomic: gem_close_race/1941/0x00000002 Modules linked in: hid_generic usbhid i915 asix usbnet libphy mii i2c_algo_bit drm_kms_helper cfbfillrect syscopyarea cfbimgblt sysfillrect sysimgblt fb_sys_fops cfbcopyarea drm coretemp i2c_hid hid video pinctrl_sunrisepoint pinctrl_intel acpi_pad nls_iso8859_1 e1000e ptp psmouse pps_core ahci libahci CPU: 0 PID: 1941 Comm: gem_close_race Tainted: G U 4.4.0-160121+ #123 Hardware name: Intel Corporation Skylake Client platform/Skylake AIO DDR3L RVP10, BIOS SKLSE2R1.R00.X100.B01.1509220551 09/22/2015 0000000000013e40 ffff880166c27a78 ffffffff81280d02 ffff880172c13e40 ffff880166c27a88 ffffffff810c203a ffff880166c27ac8 ffffffff814ec808 ffff88016b7c6000 ffff880166c28000 00000000000f4240 0000000000000001 Call Trace: [] dump_stack+0x4b/0x79 [] __schedule_bug+0x41/0x4f [] __schedule+0x5a8/0x690 [] schedule+0x37/0x80 [] schedule_hrtimeout_range_clock+0xad/0x130 [] ? hrtimer_init+0x10/0x10 [] ? schedule_hrtimeout_range_clock+0xa1/0x130 [] schedule_hrtimeout_range+0xe/0x10 [] usleep_range+0x3b/0x40 [] i915_guc_wq_check_space+0x119/0x210 [i915] [] intel_logical_ring_alloc_request_extras+0x5c/0x70 [i915] [] i915_gem_request_alloc+0x91/0x170 [i915] [] i915_gem_do_execbuffer.isra.25+0xbc7/0x12a0 [i915] [] ? i915_gem_object_get_pages_gtt+0x225/0x3c0 [i915] [] ? i915_gem_pwrite_ioctl+0xd6/0x9f0 [i915] [] i915_gem_execbuffer2+0xa8/0x250 [i915] [] drm_ioctl+0x258/0x4f0 [drm] [] ? i915_gem_execbuffer+0x340/0x340 [i915] [] do_vfs_ioctl+0x2cd/0x4a0 [] ? __fget+0x72/0xb0 [] SyS_ioctl+0x3c/0x70 [] entry_SYSCALL_64_fastpath+0x12/0x6a ------------[ cut here ]------------ v4: Only tear down doorbell & kunmap() client object if we actually succeeded in allocating a client object (Tvrtko Ursulin) Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93847 Original-version-by: Alex Dai Signed-off-by: Dave Gordon Cc: Tvtrko Ursulin Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_guc_submission.c | 66 +++++++++++++----------------- drivers/gpu/drm/i915/intel_guc.h | 2 + 2 files changed, 30 insertions(+), 38 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index da86bdb..d251699 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -179,15 +179,11 @@ static void guc_init_doorbell(struct intel_guc *guc, struct i915_guc_client *client) { struct guc_doorbell_info *doorbell; - void *base; - base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0)); - doorbell = base + client->doorbell_offset; + doorbell = client->client_base + client->doorbell_offset; - doorbell->db_status = 1; + doorbell->db_status = GUC_DOORBELL_ENABLED; doorbell->cookie = 0; - - kunmap_atomic(base); } static int guc_ring_doorbell(struct i915_guc_client *gc) @@ -195,11 +191,9 @@ static int guc_ring_doorbell(struct i915_guc_client *gc) struct guc_process_desc *desc; union guc_doorbell_qw db_cmp, db_exc, db_ret; union guc_doorbell_qw *db; - void *base; int attempt = 2, ret = -EAGAIN; - base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0)); - desc = base + gc->proc_desc_offset; + desc = gc->client_base + gc->proc_desc_offset; /* Update the tail so it is visible to GuC */ desc->tail = gc->wq_tail; @@ -215,7 +209,7 @@ static int guc_ring_doorbell(struct i915_guc_client *gc) db_exc.cookie = 1; /* pointer of current doorbell cacheline */ - db = base + gc->doorbell_offset; + db = gc->client_base + gc->doorbell_offset; while (attempt--) { /* lets ring the doorbell */ @@ -247,7 +241,6 @@ static int guc_ring_doorbell(struct i915_guc_client *gc) /* Finally, update the cached copy of the GuC's WQ head */ gc->wq_head = desc->head; - kunmap_atomic(base); return ret; } @@ -256,16 +249,12 @@ static void guc_disable_doorbell(struct intel_guc *guc, { struct drm_i915_private *dev_priv = guc_to_i915(guc); struct guc_doorbell_info *doorbell; - void *base; i915_reg_t drbreg = GEN8_DRBREGL(client->doorbell_id); int value; - base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0)); - doorbell = base + client->doorbell_offset; + doorbell = client->client_base + client->doorbell_offset; - doorbell->db_status = 0; - - kunmap_atomic(base); + doorbell->db_status = GUC_DOORBELL_DISABLED; I915_WRITE(drbreg, I915_READ(drbreg) & ~GEN8_DRB_VALID); @@ -341,10 +330,8 @@ static void guc_init_proc_desc(struct intel_guc *guc, struct i915_guc_client *client) { struct guc_process_desc *desc; - void *base; - base = kmap_atomic(i915_gem_object_get_page(client->client_obj, 0)); - desc = base + client->proc_desc_offset; + desc = client->client_base + client->proc_desc_offset; memset(desc, 0, sizeof(*desc)); @@ -361,8 +348,6 @@ static void guc_init_proc_desc(struct intel_guc *guc, desc->wq_size_bytes = client->wq_size; desc->wq_status = WQ_STATUS_ACTIVE; desc->priority = client->priority; - - kunmap_atomic(base); } /* @@ -474,7 +459,6 @@ static void guc_fini_ctx_desc(struct intel_guc *guc, int i915_guc_wq_check_space(struct i915_guc_client *gc) { struct guc_process_desc *desc; - void *base; u32 size = sizeof(struct guc_wq_item); int ret = -ETIMEDOUT, timeout_counter = 200; @@ -486,8 +470,7 @@ int i915_guc_wq_check_space(struct i915_guc_client *gc) if (CIRC_SPACE(gc->wq_tail, gc->wq_head, gc->wq_size) >= size) return 0; - base = kmap_atomic(i915_gem_object_get_page(gc->client_obj, 0)); - desc = base + gc->proc_desc_offset; + desc = gc->client_base + gc->proc_desc_offset; while (timeout_counter-- > 0) { gc->wq_head = desc->head; @@ -501,8 +484,6 @@ int i915_guc_wq_check_space(struct i915_guc_client *gc) usleep_range(1000, 2000); }; - kunmap_atomic(base); - return ret; } @@ -661,21 +642,28 @@ static void guc_client_free(struct drm_device *dev, if (!client) return; - if (client->doorbell_id != GUC_INVALID_DOORBELL_ID) { - /* - * First disable the doorbell, then tell the GuC we've - * finished with it, finally deallocate it in our bitmap - */ - guc_disable_doorbell(guc, client); - host2guc_release_doorbell(guc, client); - release_doorbell(guc, client->doorbell_id); - } - /* * XXX: wait for any outstanding submissions before freeing memory. * Be sure to drop any locks */ + if (client->client_base) { + /* + * If we got as far as setting up a doorbell, make sure + * we shut it down before unmapping & deallocating the + * memory. So first disable the doorbell, then tell the + * GuC that we've finished with it, finally deallocate + * it in our bitmap + */ + if (client->doorbell_id != GUC_INVALID_DOORBELL_ID) { + guc_disable_doorbell(guc, client); + host2guc_release_doorbell(guc, client); + release_doorbell(guc, client->doorbell_id); + } + + kunmap(kmap_to_page(client->client_base)); + } + gem_release_guc_obj(client->client_obj); if (client->ctx_index != GUC_INVALID_CTX_ID) { @@ -696,7 +684,7 @@ static void guc_client_free(struct drm_device *dev, * @ctx: the context that owns the client (we use the default render * context) * - * Return: An i915_guc_client object if success. + * Return: An i915_guc_client object if success, else NULL. */ static struct i915_guc_client *guc_client_alloc(struct drm_device *dev, uint32_t priority, @@ -728,7 +716,9 @@ static struct i915_guc_client *guc_client_alloc(struct drm_device *dev, if (!obj) goto err; + /* We'll keep just the first (doorbell/proc) page permanently kmap'd. */ client->client_obj = obj; + client->client_base = kmap(i915_gem_object_get_page(obj, 0)); client->wq_offset = GUC_DB_SIZE; client->wq_size = GUC_WQ_SIZE; diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index 3bb85b1..06050c241a 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -31,6 +31,7 @@ struct drm_i915_gem_request; struct i915_guc_client { struct drm_i915_gem_object *client_obj; + void *client_base; /* first page (only) of above */ struct intel_context *owner; struct intel_guc *guc; uint32_t priority; @@ -52,6 +53,7 @@ struct i915_guc_client { uint32_t q_fail; uint32_t b_fail; int retcode; + int spare; /* pad to 32 DWords */ }; enum intel_guc_fw_status {