From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 9A1F7BA49 for ; Tue, 7 Mar 2023 18:43:12 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id F224FC433EF; Tue, 7 Mar 2023 18:43:11 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1678214592; bh=v6aqi8yIjHOzZ1G5m6H5in+wgo6c9Wzy4oml/pd+O2Q=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=BUq8UFzX3GSVrerPtRsIld3kIKLBSucO5OtPYZIJ+GOHCa31K+K4z1ToOAcM6em9n oDbs3Km9IdKvyruNNd5BSiU1li3GsnZIzx5J0zBx9LQIekWqaA3EMH5zcuUqlUQhVa smhWan/OK7o20Fz6dquBIEBdvenHrfuXOzF8cbcw= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Robin Murphy , Kevin Tian , Sanjay Kumar , Jacob Pan , Lu Baolu , Joerg Roedel Subject: [PATCH 6.1 872/885] iommu/vt-d: Avoid superfluous IOTLB tracking in lazy mode Date: Tue, 7 Mar 2023 18:03:26 +0100 Message-Id: <20230307170039.671009849@linuxfoundation.org> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230307170001.594919529@linuxfoundation.org> References: <20230307170001.594919529@linuxfoundation.org> User-Agent: quilt/0.67 Precedence: bulk X-Mailing-List: patches@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Jacob Pan commit 16a75bbe480c3598b3af57a2504ea89b1e32c3ac upstream. Intel IOMMU driver implements IOTLB flush queue with domain selective or PASID selective invalidations. In this case there's no need to track IOVA page range and sync IOTLBs, which may cause significant performance hit. This patch adds a check to avoid IOVA gather page and IOTLB sync for the lazy path. The performance difference on Sapphire Rapids 100Gb NIC is improved by the following (as measured by iperf send): w/o this fix~48 Gbits/s. with this fix ~54 Gbits/s Cc: Fixes: 2a2b8eaa5b25 ("iommu: Handle freelists when using deferred flushing in iommu drivers") Reviewed-by: Robin Murphy Reviewed-by: Kevin Tian Tested-by: Sanjay Kumar Signed-off-by: Sanjay Kumar Signed-off-by: Jacob Pan Link: https://lore.kernel.org/r/20230209175330.1783556-1-jacob.jun.pan@linux.intel.com Signed-off-by: Lu Baolu Signed-off-by: Joerg Roedel Signed-off-by: Greg Kroah-Hartman --- drivers/iommu/intel/iommu.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) --- a/drivers/iommu/intel/iommu.c +++ b/drivers/iommu/intel/iommu.c @@ -4359,7 +4359,12 @@ static size_t intel_iommu_unmap(struct i if (dmar_domain->max_addr == iova + size) dmar_domain->max_addr = iova; - iommu_iotlb_gather_add_page(domain, gather, iova, size); + /* + * We do not use page-selective IOTLB invalidation in flush queue, + * so there is no need to track page and sync iotlb. + */ + if (!iommu_iotlb_gather_queued(gather)) + iommu_iotlb_gather_add_page(domain, gather, iova, size); return size; }