From patchwork Wed Mar 28 22:30:56 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 10314221 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 A5D7A60325 for ; Wed, 28 Mar 2018 22:31:04 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 8B2F429A3B for ; Wed, 28 Mar 2018 22:31:04 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 7E55F29A69; Wed, 28 Mar 2018 22:31:04 +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=-4.2 required=2.0 tests=BAYES_00, 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 04E5729A3B for ; Wed, 28 Mar 2018 22:31:03 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 4405789104; Wed, 28 Mar 2018 22:31:02 +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 D099A89104 for ; Wed, 28 Mar 2018 22:31:00 +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 11194162-1500050 for ; Wed, 28 Mar 2018 23:30:56 +0100 Received: by haswell.alporthouse.com (sSMTP sendmail emulation); Wed, 28 Mar 2018 23:30:56 +0100 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Wed, 28 Mar 2018 23:30:56 +0100 Message-Id: <20180328223056.31439-1-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.16.3 MIME-Version: 1.0 X-Originating-IP: 78.156.65.138 X-Country: code=GB country="United Kingdom" ip=78.156.65.138 Subject: [Intel-gfx] [CI] drm/i915: Avoid sleeping inside per-engine reset 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: , Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" X-Virus-Scanned: ClamAV using ClamSMTP Only sleep and repeat when asked for a full device reset (ALL_ENGINES) and avoid using sleeping waits when asked for a per-engine reset. The goal is to be able to use a per-engine reset from hardirq/softirq/timer context. A consequence is that our individual wait timeouts are a thousand times shorter, on the order of a hundred microseconds rather than hundreds of millisecond. This may make hitting the timeouts more common, but hopefully the fallover to the full-device reset will be sufficient to pick up the pieces. Note, that the sleeps inside older gen (pre-gen8) have been left as they are only used in full device reset mode. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Michał Winiarski CC: Michel Thierry Cc: Jeff McGee Cc: Tvrtko Ursulin Reviewed-by: Michel Thierry --- drivers/gpu/drm/i915/intel_uncore.c | 51 ++++++++++++++++++++++++------------- 1 file changed, 34 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index f37ecfc69e49..a0d7e0cfbd32 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1702,11 +1702,10 @@ static void gen3_stop_engine(struct intel_engine_cs *engine) const i915_reg_t mode = RING_MI_MODE(base); I915_WRITE_FW(mode, _MASKED_BIT_ENABLE(STOP_RING)); - if (intel_wait_for_register_fw(dev_priv, - mode, - MODE_IDLE, - MODE_IDLE, - 500)) + if (__intel_wait_for_register_fw(dev_priv, + mode, MODE_IDLE, MODE_IDLE, + 500, 0, + NULL)) DRM_DEBUG_DRIVER("%s: timed out on STOP_RING\n", engine->name); @@ -1860,9 +1859,10 @@ static int gen6_hw_domain_reset(struct drm_i915_private *dev_priv, __raw_i915_write32(dev_priv, GEN6_GDRST, hw_domain_mask); /* Wait for the device to ack the reset requests */ - err = intel_wait_for_register_fw(dev_priv, - GEN6_GDRST, hw_domain_mask, 0, - 500); + err = __intel_wait_for_register_fw(dev_priv, + GEN6_GDRST, hw_domain_mask, 0, + 500, 0, + NULL); if (err) DRM_DEBUG_DRIVER("Wait for 0x%08x engines reset failed\n", hw_domain_mask); @@ -2027,11 +2027,12 @@ static int gen8_reset_engine_start(struct intel_engine_cs *engine) I915_WRITE_FW(RING_RESET_CTL(engine->mmio_base), _MASKED_BIT_ENABLE(RESET_CTL_REQUEST_RESET)); - ret = intel_wait_for_register_fw(dev_priv, - RING_RESET_CTL(engine->mmio_base), - RESET_CTL_READY_TO_RESET, - RESET_CTL_READY_TO_RESET, - 700); + ret = __intel_wait_for_register_fw(dev_priv, + RING_RESET_CTL(engine->mmio_base), + RESET_CTL_READY_TO_RESET, + RESET_CTL_READY_TO_RESET, + 700, 0, + NULL); if (ret) DRM_ERROR("%s: reset request timeout\n", engine->name); @@ -2094,15 +2095,31 @@ int intel_gpu_reset(struct drm_i915_private *dev_priv, unsigned engine_mask) int retry; int ret; - might_sleep(); + /* + * We want to perform per-engine reset from atomic context (e.g. + * softirq), which imposes the constraint that we cannot sleep. + * However, experience suggests that spending a bit of time waiting + * for a reset helps in various cases, so for a full-device reset + * we apply the opposite rule and wait if we want to. As we should + * always follow up a failed per-engine reset with a full device reset, + * being a little faster, stricter and more error prone for the + * atomic case seems an acceptable compromise. + * + * Unfortunately this leads to a bimodal routine, when the goal was + * to have a single reset function that worked for resetting any + * number of engines simultaneously. + */ + might_sleep_if(engine_mask == ALL_ENGINES); - /* If the power well sleeps during the reset, the reset + /* + * If the power well sleeps during the reset, the reset * request may be dropped and never completes (causing -EIO). */ intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL); for (retry = 0; retry < 3; retry++) { - /* We stop engines, otherwise we might get failed reset and a + /* + * We stop engines, otherwise we might get failed reset and a * dead gpu (on elk). Also as modern gpu as kbl can suffer * from system hang if batchbuffer is progressing when * the reset is issued, regardless of READY_TO_RESET ack. @@ -2120,7 +2137,7 @@ int intel_gpu_reset(struct drm_i915_private *dev_priv, unsigned engine_mask) GEM_TRACE("engine_mask=%x\n", engine_mask); ret = reset(dev_priv, engine_mask); } - if (ret != -ETIMEDOUT) + if (ret != -ETIMEDOUT || engine_mask != ALL_ENGINES) break; cond_resched();