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 E1121CA0EC4 for ; Tue, 12 Aug 2025 18:27:05 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 8D9A210E0B0; Tue, 12 Aug 2025 18:27:05 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="nfhUqOvM"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.10]) by gabe.freedesktop.org (Postfix) with ESMTPS id 350C010E0B0 for ; Tue, 12 Aug 2025 18:27:05 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1755023225; x=1786559225; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=fsiXbfayP8XVScLoynfut7jBDMojEs3ltFGE12UfHxE=; b=nfhUqOvMpFaSP+5++m9YlxE0/YaC2F6qTuZDvxszV5+GJKbIGLL8j60y MyLTftbRfLJ8q/XAR3OxI4JCLeRvdKfePDsimnjYeR6QhX4P2t4cxb4bW 82L2L3hck3xiczk/SW3XEk1vUccKsdHszGK2vAFtKQKxSp+ZEKXvSP1yb vHp5ONBKhDATAd3SGwZsshUX3bAG0pWL5gPKEKc1nitK/okaFWHq53NkA uRNg4tMnsCmlH2q7BSb6wZ2gwdcIrPr5aSRjC17EafWKhfLvtWdtgoxd2 Chcfyj9SCueqLt+jvlLSo+XVoM2iop+Ujcy0pOBjZdWwfVYsmx9WUO5Zk A==; X-CSE-ConnectionGUID: HicEYzeJR0yDqwCTv8zpag== X-CSE-MsgGUID: lYBbCsciStqiIqk6ny4tRw== X-IronPort-AV: E=McAfee;i="6800,10657,11520"; a="74757082" X-IronPort-AV: E=Sophos;i="6.17,284,1747724400"; d="scan'208";a="74757082" Received: from fmviesa001.fm.intel.com ([10.60.135.141]) by orvoesa102.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 12 Aug 2025 11:27:05 -0700 X-CSE-ConnectionGUID: TwmgDRm3QcaHcWxMj8gxnQ== X-CSE-MsgGUID: aeFEYx19T2aAEvctIxCQ8Q== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.17,284,1747724400"; d="scan'208";a="197124875" Received: from fyang16-desk.jf.intel.com ([10.165.21.214]) by smtpauth.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 12 Aug 2025 11:27:04 -0700 From: fei.yang@intel.com To: intel-xe@lists.freedesktop.org Cc: matthew.d.roper@intel.com, lucas.demarchi@intel.com, matthew.brost@intel.com, Fei Yang Subject: [PATCH 1/1] drm/xe: serialize store_data and user_interrupt for ufence wait Date: Tue, 12 Aug 2025 11:28:46 -0700 Message-ID: <20250812182846.1608209-2-fei.yang@intel.com> X-Mailer: git-send-email 2.43.0 In-Reply-To: <20250812182846.1608209-1-fei.yang@intel.com> References: <20250812182846.1608209-1-fei.yang@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit 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: , Errors-To: intel-xe-bounces@lists.freedesktop.org Sender: "Intel-xe" From: Fei Yang Quote BSpec, MI_STORE_DATA_IMM "simply initiates the write operation with command execution proceeding normally. Although the write operation is guaranteed to complete eventually, there is no mechanism to synchronize command execution with the completion (or even initiation) of these operations." The KMD currently emit MI_STORE_DATA_IMM and MI_USER_INTERRUPT consecutively to implement user fence. However, according to the BSpec, the data write operation is not guaranteed to be completed when triggering the interrupt, that would cause the xe_wait_user_fence_ioctl to wait until the full user specified timeout is reached before checking the fence value again. Great performance degradation has been observed in IGT xe_exec_fault_mode test cases due to this unnecessary wait. The worst case is that if user set the timeout to MAX_INT32, the wait could end up being a hang until some other random program triggers a user interrupt to wake it up. A semaphore wait is added right after the data write to avoid the unexpected wait. Signed-off-by: Fei Yang --- drivers/gpu/drm/xe/instructions/xe_mi_commands.h | 11 +++++++++++ drivers/gpu/drm/xe/xe_ring_ops.c | 14 ++++++++++++++ 2 files changed, 25 insertions(+) diff --git a/drivers/gpu/drm/xe/instructions/xe_mi_commands.h b/drivers/gpu/drm/xe/instructions/xe_mi_commands.h index c47b290e0e9f..1c9e7b35c665 100644 --- a/drivers/gpu/drm/xe/instructions/xe_mi_commands.h +++ b/drivers/gpu/drm/xe/instructions/xe_mi_commands.h @@ -34,6 +34,17 @@ #define MI_FORCE_WAKEUP __MI_INSTR(0x1D) #define MI_MATH(n) (__MI_INSTR(0x1A) | XE_INSTR_NUM_DW((n) + 1)) +#define MI_SEMAPHORE_WAIT (__MI_INSTR(0x1c) | XE_INSTR_NUM_DW(5)) +#define MI_SEMAPHORE_REGISTER_POLL REG_BIT(16) +#define MI_SEMAPHORE_POLL REG_BIT(15) +#define MI_SEMAPHORE_COMP_OP GENMASK(14, 12) +#define MI_SEMAPHORE_SAD_GT_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 0) +#define MI_SEMAPHORE_SAD_GTE_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 1) +#define MI_SEMAPHORE_SAD_LT_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 2) +#define MI_SEMAPHORE_SAD_LTE_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 3) +#define MI_SEMAPHORE_SAD_EQ_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 4) +#define MI_SEMAPHORE_SAD_NEQ_SDD REG_FIELD_PREP(MI_SEMAPHORE_COMP_OP, 5) + #define MI_STORE_DATA_IMM __MI_INSTR(0x20) #define MI_SDI_GGTT REG_BIT(22) #define MI_SDI_LEN_DW GENMASK(9, 0) diff --git a/drivers/gpu/drm/xe/xe_ring_ops.c b/drivers/gpu/drm/xe/xe_ring_ops.c index 5f15360d14bf..189e764e3914 100644 --- a/drivers/gpu/drm/xe/xe_ring_ops.c +++ b/drivers/gpu/drm/xe/xe_ring_ops.c @@ -169,6 +169,20 @@ static int emit_store_imm_ppgtt_posted(u64 addr, u64 value, dw[i++] = upper_32_bits(addr); dw[i++] = lower_32_bits(value); dw[i++] = upper_32_bits(value); + dw[i++] = MI_SEMAPHORE_WAIT | + MI_SEMAPHORE_POLL | + MI_SEMAPHORE_SAD_EQ_SDD; + dw[i++] = lower_32_bits(value); + dw[i++] = lower_32_bits(addr); + dw[i++] = upper_32_bits(addr); + dw[i++] = 0; + dw[i++] = MI_SEMAPHORE_WAIT | + MI_SEMAPHORE_POLL | + MI_SEMAPHORE_SAD_EQ_SDD; + dw[i++] = upper_32_bits(value); + dw[i++] = lower_32_bits(addr + 4); + dw[i++] = upper_32_bits(addr); + dw[i++] = 0; return i; } -- 2.43.0