iommu/io-pgtable-arm: Add support for non-strict mode

commit b6b65ca20bc93d14319f9b5cf98fd3c19a4244e3 upstream

Non-strict mode is simply a case of skipping 'regular' leaf TLBIs, since
the sync is already factored out into ops->iotlb_sync at the core API
level. Non-leaf invalidations where we change the page table structure
itself still have to be issued synchronously in order to maintain walk
caches correctly.

To save having to reason about it too much, make sure the invalidation
in arm_lpae_split_blk_unmap() just performs its own unconditional sync
to minimise the window in which we're technically violating the break-
before-make requirement on a live mapping. This might work out redundant
with an outer-level sync for strict unmaps, but we'll never be splitting
blocks on a DMA fastpath anyway.
Signed-off-by: NZhen Lei <thunder.leizhen@huawei.com>
[rm: tweak comment, commit message, split_blk_unmap logic and barriers]
Signed-off-by: NRobin Murphy <robin.murphy@arm.com>
Signed-off-by: NWill Deacon <will.deacon@arm.com>
Signed-off-by: NZou Cao <zoucao@linux.alibaba.com>
Reviewed-by: NBaoyou Xie <xie.baoyou@linux.alibaba.com>

drivers/iommu/io-pgtable-arm.c

@@ -576,6 +576,7 @@ static size_t arm_lpae_split_blk_unmap(struct arm_lpae_io_pgtable *data,
 		tablep = iopte_deref(pte, data);
 	} else if (unmap_idx >= 0) {
 		io_pgtable_tlb_add_flush(&data->iop, iova, size, size, true);
+		io_pgtable_tlb_sync(&data->iop);
 		return size;
 	}
 
@@ -609,6 +610,13 @@ static size_t __arm_lpae_unmap(struct arm_lpae_io_pgtable *data,
 			io_pgtable_tlb_sync(iop);
 			ptep = iopte_deref(pte, data);
 			__arm_lpae_free_pgtable(data, lvl + 1, ptep);
+		} else if (iop->cfg.quirks & IO_PGTABLE_QUIRK_NON_STRICT) {
+			/*
+			 * Order the PTE update against queueing the IOVA, to
+			 * guarantee that a flush callback from a different CPU
+			 * has observed it before the TLBIALL can be issued.
+			 */
+			smp_wmb();
 		} else {
 			io_pgtable_tlb_add_flush(iop, iova, size, size, true);
 		}
@@ -771,7 +779,8 @@ arm_64_lpae_alloc_pgtable_s1(struct io_pgtable_cfg *cfg, void *cookie)
 	u64 reg;
 	struct arm_lpae_io_pgtable *data;
 
-	if (cfg->quirks & ~(IO_PGTABLE_QUIRK_ARM_NS | IO_PGTABLE_QUIRK_NO_DMA))
+	if (cfg->quirks & ~(IO_PGTABLE_QUIRK_ARM_NS | IO_PGTABLE_QUIRK_NO_DMA |
+			    IO_PGTABLE_QUIRK_NON_STRICT))
 		return NULL;
 
 	data = arm_lpae_alloc_pgtable(cfg);
@@ -863,7 +872,8 @@ arm_64_lpae_alloc_pgtable_s2(struct io_pgtable_cfg *cfg, void *cookie)
 	struct arm_lpae_io_pgtable *data;
 
 	/* The NS quirk doesn't apply at stage 2 */
-	if (cfg->quirks & ~IO_PGTABLE_QUIRK_NO_DMA)
+	if (cfg->quirks & ~(IO_PGTABLE_QUIRK_NO_DMA |
+			    IO_PGTABLE_QUIRK_NON_STRICT))
 		return NULL;
 
 	data = arm_lpae_alloc_pgtable(cfg);

drivers/iommu/io-pgtable.h

@@ -71,12 +71,17 @@ struct io_pgtable_cfg {
 	 *	be accessed by a fully cache-coherent IOMMU or CPU (e.g. for a
 	 *	software-emulated IOMMU), such that pagetable updates need not
 	 *	be treated as explicit DMA data.
+	 *
+	 * IO_PGTABLE_QUIRK_NON_STRICT: Skip issuing synchronous leaf TLBIs
+	 *	on unmap, for DMA domains using the flush queue mechanism for
+	 *	delayed invalidation.
 	 */
 	#define IO_PGTABLE_QUIRK_ARM_NS		BIT(0)
 	#define IO_PGTABLE_QUIRK_NO_PERMS	BIT(1)
 	#define IO_PGTABLE_QUIRK_TLBI_ON_MAP	BIT(2)
 	#define IO_PGTABLE_QUIRK_ARM_MTK_4GB	BIT(3)
 	#define IO_PGTABLE_QUIRK_NO_DMA		BIT(4)
+	#define IO_PGTABLE_QUIRK_NON_STRICT	BIT(5)
 	unsigned long			quirks;
 	unsigned long			pgsize_bitmap;
 	unsigned int			ias;