Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Niklas Schnelle <schnelle@linux.ibm.com>
To:     Matthew Rosato <mjrosato@linux.ibm.com>, iommu@lists.linux.dev,
        Joerg Roedel <joro@8bytes.org>, Will Deacon <will@kernel.org>,
        Robin Murphy <robin.murphy@arm.com>,
        Jason Gunthorpe <jgg@nvidia.com>
Cc:     Gerd Bayer <gbayer@linux.ibm.com>,
        Pierre Morel <pmorel@linux.ibm.com>,
        linux-s390@vger.kernel.org, borntraeger@linux.ibm.com,
        hca@linux.ibm.com, gor@linux.ibm.com,
        gerald.schaefer@linux.ibm.com, agordeev@linux.ibm.com,
        svens@linux.ibm.com, linux-kernel@vger.kernel.org
Subject: [PATCH 2/5] iommu/s390: Add I/O TLB ops
Date:   Tue, 18 Oct 2022 16:51:29 +0200
Message-Id: <20221018145132.998866-3-schnelle@linux.ibm.com>
In-Reply-To: <20221018145132.998866-1-schnelle@linux.ibm.com>
References: <20221018145132.998866-1-schnelle@linux.ibm.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Currently s390-iommu does an I/O TLB flush (RPCIT) for every update of
the I/O translation table explicitly. For one this is wasteful since
RPCIT can be skipped after a mapping operation if zdev->tlb_refresh is
unset. Moreover we can do a single RPCIT for a range of pages including
whne doing lazy unmapping.

Thankfully both of these optimizations can be achieved by implementing
the IOMMU operations common code provides for the different types of I/O
tlb flushes:

 * flush_iotlb_all: Flushes the I/O TLB for the entire IOVA space
 * iotlb_sync:  Flushes the I/O TLB for a range of pages that can be
   gathered up, for example to implement lazy unmapping.
 * iotlb_sync_map: Flushes the I/O TLB after a mapping operation

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
 drivers/iommu/s390-iommu.c | 76 ++++++++++++++++++++++++++++++++------
 1 file changed, 65 insertions(+), 11 deletions(-)

diff --git a/drivers/iommu/s390-iommu.c b/drivers/iommu/s390-iommu.c
index ee88e717254b..a4c2e9bc6d83 100644
--- a/drivers/iommu/s390-iommu.c
+++ b/drivers/iommu/s390-iommu.c
@@ -199,14 +199,72 @@ static void s390_iommu_release_device(struct device *dev)
 		__s390_iommu_detach_device(zdev);
 }
 
+static void s390_iommu_flush_iotlb_all(struct iommu_domain *domain)
+{
+	struct s390_domain *s390_domain = to_s390_domain(domain);
+	struct zpci_dev *zdev;
+	unsigned long flags;
+	int rc;
+
+	spin_lock_irqsave(&s390_domain->list_lock, flags);
+	list_for_each_entry(zdev, &s390_domain->devices, iommu_list) {
+		rc = zpci_refresh_trans((u64)zdev->fh << 32, zdev->start_dma,
+					zdev->end_dma - zdev->start_dma + 1);
+		if (rc)
+			break;
+	}
+	spin_unlock_irqrestore(&s390_domain->list_lock, flags);
+}
+
+static void s390_iommu_iotlb_sync(struct iommu_domain *domain,
+				  struct iommu_iotlb_gather *gather)
+{
+	struct s390_domain *s390_domain = to_s390_domain(domain);
+	size_t size = gather->end - gather->start + 1;
+	struct zpci_dev *zdev;
+	unsigned long flags;
+	int rc;
+
+	/* If gather was never added to there is nothing to flush */
+	if (gather->start == ULONG_MAX)
+		return;
+
+	spin_lock_irqsave(&s390_domain->list_lock, flags);
+	list_for_each_entry(zdev, &s390_domain->devices, iommu_list) {
+		rc = zpci_refresh_trans((u64)zdev->fh << 32, gather->start,
+					size);
+		if (rc)
+			break;
+	}
+	spin_unlock_irqrestore(&s390_domain->list_lock, flags);
+}
+
+static void s390_iommu_iotlb_sync_map(struct iommu_domain *domain,
+				      unsigned long iova, size_t size)
+{
+	struct s390_domain *s390_domain = to_s390_domain(domain);
+	struct zpci_dev *zdev;
+	unsigned long flags;
+	int rc;
+
+	spin_lock_irqsave(&s390_domain->list_lock, flags);
+	list_for_each_entry(zdev, &s390_domain->devices, iommu_list) {
+		if (!zdev->tlb_refresh)
+			continue;
+		rc = zpci_refresh_trans((u64)zdev->fh << 32,
+					iova, size);
+		if (rc)
+			break;
+	}
+	spin_unlock_irqrestore(&s390_domain->list_lock, flags);
+}
+
 static int s390_iommu_update_trans(struct s390_domain *s390_domain,
 				   phys_addr_t pa, dma_addr_t dma_addr,
 				   unsigned long nr_pages, int flags)
 {
 	phys_addr_t page_addr = pa & PAGE_MASK;
-	dma_addr_t start_dma_addr = dma_addr;
 	unsigned long irq_flags, i;
-	struct zpci_dev *zdev;
 	unsigned long *entry;
 	int rc = 0;
 
@@ -225,15 +283,6 @@ static int s390_iommu_update_trans(struct s390_domain *s390_domain,
 		dma_addr += PAGE_SIZE;
 	}
 
-	spin_lock(&s390_domain->list_lock);
-	list_for_each_entry(zdev, &s390_domain->devices, iommu_list) {
-		rc = zpci_refresh_trans((u64)zdev->fh << 32,
-					start_dma_addr, nr_pages * PAGE_SIZE);
-		if (rc)
-			break;
-	}
-	spin_unlock(&s390_domain->list_lock);
-
 undo_cpu_trans:
 	if (rc && ((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)) {
 		flags = ZPCI_PTE_INVALID;
@@ -340,6 +389,8 @@ static size_t s390_iommu_unmap_pages(struct iommu_domain *domain,
 	if (rc)
 		return 0;
 
+	iommu_iotlb_gather_add_range(gather, iova, size);
+
 	return size;
 }
 
@@ -384,6 +435,9 @@ static const struct iommu_ops s390_iommu_ops = {
 		.detach_dev	= s390_iommu_detach_device,
 		.map_pages	= s390_iommu_map_pages,
 		.unmap_pages	= s390_iommu_unmap_pages,
+		.flush_iotlb_all = s390_iommu_flush_iotlb_all,
+		.iotlb_sync      = s390_iommu_iotlb_sync,
+		.iotlb_sync_map  = s390_iommu_iotlb_sync_map,
 		.iova_to_phys	= s390_iommu_iova_to_phys,
 		.free		= s390_domain_free,
 	}
-- 
2.34.1