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 4/5] iommu/s390: Optimize IOMMU table walking
Date:   Tue, 18 Oct 2022 16:51:31 +0200
Message-Id: <20221018145132.998866-5-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

When invalidating existing table entries for unmap there is no need to
know the physical address beforehand so don't do an extra walk of the
IOMMU table to get it. Also when invalidating entries not finding an
entry indicates an invalid unmap and not a lack of memory we also don't
need to undo updates in this case. Implement this by splitting
s390_iommu_update_trans() in a variant for validating and one for
invalidating translations.

Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
 drivers/iommu/s390-iommu.c | 69 ++++++++++++++++++++++++--------------
 1 file changed, 43 insertions(+), 26 deletions(-)

diff --git a/drivers/iommu/s390-iommu.c b/drivers/iommu/s390-iommu.c
index 4e90987be387..efd258e3f74b 100644
--- a/drivers/iommu/s390-iommu.c
+++ b/drivers/iommu/s390-iommu.c
@@ -260,14 +260,14 @@ static void s390_iommu_iotlb_sync_map(struct iommu_domain *domain,
 	rcu_read_unlock();
 }
 
-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)
+static int s390_iommu_validate_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;
 	unsigned long irq_flags, i;
 	unsigned long *entry;
-	int rc = 0;
+	int rc;
 
 	if (!nr_pages)
 		return 0;
@@ -275,7 +275,7 @@ static int s390_iommu_update_trans(struct s390_domain *s390_domain,
 	spin_lock_irqsave(&s390_domain->dma_table_lock, irq_flags);
 	for (i = 0; i < nr_pages; i++) {
 		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr);
-		if (!entry) {
+		if (unlikely(!entry)) {
 			rc = -ENOMEM;
 			goto undo_cpu_trans;
 		}
@@ -283,19 +283,43 @@ static int s390_iommu_update_trans(struct s390_domain *s390_domain,
 		page_addr += PAGE_SIZE;
 		dma_addr += PAGE_SIZE;
 	}
+	spin_unlock_irqrestore(&s390_domain->dma_table_lock, irq_flags);
+
+	return 0;
 
 undo_cpu_trans:
-	if (rc && ((flags & ZPCI_PTE_VALID_MASK) == ZPCI_PTE_VALID)) {
-		flags = ZPCI_PTE_INVALID;
-		while (i-- > 0) {
-			page_addr -= PAGE_SIZE;
-			dma_addr -= PAGE_SIZE;
-			entry = dma_walk_cpu_trans(s390_domain->dma_table,
-						   dma_addr);
-			if (!entry)
-				break;
-			dma_update_cpu_trans(entry, page_addr, flags);
+	while (i-- > 0) {
+		dma_addr -= PAGE_SIZE;
+		entry = dma_walk_cpu_trans(s390_domain->dma_table,
+					   dma_addr);
+		if (!entry)
+			break;
+		dma_update_cpu_trans(entry, 0, ZPCI_PTE_INVALID);
+	}
+	spin_unlock_irqrestore(&s390_domain->dma_table_lock, irq_flags);
+
+	return rc;
+}
+
+static int s390_iommu_invalidate_trans(struct s390_domain *s390_domain,
+				       dma_addr_t dma_addr, unsigned long nr_pages)
+{
+	unsigned long irq_flags, i;
+	unsigned long *entry;
+	int rc = 0;
+
+	if (!nr_pages)
+		return 0;
+
+	spin_lock_irqsave(&s390_domain->dma_table_lock, irq_flags);
+	for (i = 0; i < nr_pages; i++) {
+		entry = dma_walk_cpu_trans(s390_domain->dma_table, dma_addr);
+		if (unlikely(!entry)) {
+			rc = -EINVAL;
+			break;
 		}
+		dma_update_cpu_trans(entry, 0, ZPCI_PTE_INVALID);
+		dma_addr += PAGE_SIZE;
 	}
 	spin_unlock_irqrestore(&s390_domain->dma_table_lock, irq_flags);
 
@@ -308,8 +332,8 @@ static int s390_iommu_map_pages(struct iommu_domain *domain,
 				int prot, gfp_t gfp, size_t *mapped)
 {
 	struct s390_domain *s390_domain = to_s390_domain(domain);
-	int flags = ZPCI_PTE_VALID, rc = 0;
 	size_t size = pgcount << __ffs(pgsize);
+	int flags = ZPCI_PTE_VALID, rc = 0;
 
 	if (pgsize != SZ_4K)
 		return -EINVAL;
@@ -327,8 +351,8 @@ static int s390_iommu_map_pages(struct iommu_domain *domain,
 	if (!(prot & IOMMU_WRITE))
 		flags |= ZPCI_TABLE_PROTECTED;
 
-	rc = s390_iommu_update_trans(s390_domain, paddr, iova,
-				     pgcount, flags);
+	rc = s390_iommu_validate_trans(s390_domain, paddr, iova,
+				       pgcount, flags);
 	if (!rc)
 		*mapped = size;
 
@@ -373,20 +397,13 @@ static size_t s390_iommu_unmap_pages(struct iommu_domain *domain,
 {
 	struct s390_domain *s390_domain = to_s390_domain(domain);
 	size_t size = pgcount << __ffs(pgsize);
-	int flags = ZPCI_PTE_INVALID;
-	phys_addr_t paddr;
 	int rc;
 
 	if (WARN_ON(iova < s390_domain->domain.geometry.aperture_start ||
 	    (iova + size - 1) > s390_domain->domain.geometry.aperture_end))
 		return 0;
 
-	paddr = s390_iommu_iova_to_phys(domain, iova);
-	if (!paddr)
-		return 0;
-
-	rc = s390_iommu_update_trans(s390_domain, paddr, iova,
-				     pgcount, flags);
+	rc = s390_iommu_invalidate_trans(s390_domain, iova, pgcount);
 	if (rc)
 		return 0;
 
-- 
2.34.1