Merge tag 'tegra-for-4.8-i2c' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra...
[cascardo/linux.git] / drivers / iommu / arm-smmu.c
index 2db74eb..c841eb7 100644 (file)
@@ -28,6 +28,7 @@
 
 #define pr_fmt(fmt) "arm-smmu: " fmt
 
+#include <linux/atomic.h>
 #include <linux/delay.h>
 #include <linux/dma-iommu.h>
 #include <linux/dma-mapping.h>
@@ -40,6 +41,8 @@
 #include <linux/module.h>
 #include <linux/of.h>
 #include <linux/of_address.h>
+#include <linux/of_device.h>
+#include <linux/of_iommu.h>
 #include <linux/pci.h>
 #include <linux/platform_device.h>
 #include <linux/slab.h>
 
 #include "io-pgtable.h"
 
-/* Maximum number of stream IDs assigned to a single device */
-#define MAX_MASTER_STREAMIDS           128
-
 /* Maximum number of context banks per SMMU */
 #define ARM_SMMU_MAX_CBS               128
 
-/* Maximum number of mapping groups per SMMU */
-#define ARM_SMMU_MAX_SMRS              128
-
 /* SMMU global address space */
 #define ARM_SMMU_GR0(smmu)             ((smmu)->base)
 #define ARM_SMMU_GR1(smmu)             ((smmu)->base + (1 << (smmu)->pgshift))
 #define ARM_SMMU_GR0_SMR(n)            (0x800 + ((n) << 2))
 #define SMR_VALID                      (1 << 31)
 #define SMR_MASK_SHIFT                 16
-#define SMR_MASK_MASK                  0x7fff
 #define SMR_ID_SHIFT                   0
-#define SMR_ID_MASK                    0x7fff
 
 #define ARM_SMMU_GR0_S2CR(n)           (0xc00 + ((n) << 2))
 #define S2CR_CBNDX_SHIFT               0
 #define S2CR_CBNDX_MASK                        0xff
 #define S2CR_TYPE_SHIFT                        16
 #define S2CR_TYPE_MASK                 0x3
-#define S2CR_TYPE_TRANS                        (0 << S2CR_TYPE_SHIFT)
-#define S2CR_TYPE_BYPASS               (1 << S2CR_TYPE_SHIFT)
-#define S2CR_TYPE_FAULT                        (2 << S2CR_TYPE_SHIFT)
+enum arm_smmu_s2cr_type {
+       S2CR_TYPE_TRANS,
+       S2CR_TYPE_BYPASS,
+       S2CR_TYPE_FAULT,
+};
 
 #define S2CR_PRIVCFG_SHIFT             24
-#define S2CR_PRIVCFG_UNPRIV            (2 << S2CR_PRIVCFG_SHIFT)
+#define S2CR_PRIVCFG_MASK              0x3
+enum arm_smmu_s2cr_privcfg {
+       S2CR_PRIVCFG_DEFAULT,
+       S2CR_PRIVCFG_DIPAN,
+       S2CR_PRIVCFG_UNPRIV,
+       S2CR_PRIVCFG_PRIV,
+};
 
 /* Context bank attribute registers */
 #define ARM_SMMU_GR1_CBAR(n)           (0x0 + ((n) << 2))
 #define ARM_SMMU_CB_TTBR0              0x20
 #define ARM_SMMU_CB_TTBR1              0x28
 #define ARM_SMMU_CB_TTBCR              0x30
+#define ARM_SMMU_CB_CONTEXTIDR         0x34
 #define ARM_SMMU_CB_S1_MAIR0           0x38
 #define ARM_SMMU_CB_S1_MAIR1           0x3c
 #define ARM_SMMU_CB_PAR                        0x50
 #define SCTLR_AFE                      (1 << 2)
 #define SCTLR_TRE                      (1 << 1)
 #define SCTLR_M                                (1 << 0)
-#define SCTLR_EAE_SBOP                 (SCTLR_AFE | SCTLR_TRE)
 
 #define ARM_MMU500_ACTLR_CPRE          (1 << 1)
 
@@ -296,23 +299,33 @@ enum arm_smmu_implementation {
        CAVIUM_SMMUV2,
 };
 
+struct arm_smmu_s2cr {
+       struct iommu_group              *group;
+       int                             count;
+       enum arm_smmu_s2cr_type         type;
+       enum arm_smmu_s2cr_privcfg      privcfg;
+       u8                              cbndx;
+};
+
+#define s2cr_init_val (struct arm_smmu_s2cr){                          \
+       .type = disable_bypass ? S2CR_TYPE_FAULT : S2CR_TYPE_BYPASS,    \
+}
+
 struct arm_smmu_smr {
-       u8                              idx;
        u16                             mask;
        u16                             id;
+       bool                            valid;
 };
 
 struct arm_smmu_master_cfg {
-       int                             num_streamids;
-       u16                             streamids[MAX_MASTER_STREAMIDS];
-       struct arm_smmu_smr             *smrs;
-};
-
-struct arm_smmu_master {
-       struct device_node              *of_node;
-       struct rb_node                  node;
-       struct arm_smmu_master_cfg      cfg;
+       struct arm_smmu_device          *smmu;
+       s16                             smendx[];
 };
+#define INVALID_SMENDX                 -1
+#define __fwspec_cfg(fw) ((struct arm_smmu_master_cfg *)fw->iommu_priv)
+#define fwspec_smmu(fw)  (__fwspec_cfg(fw)->smmu)
+#define for_each_cfg_sme(fw, i, idx) \
+       for (i = 0; idx = __fwspec_cfg(fw)->smendx[i], i < fw->num_ids; ++i)
 
 struct arm_smmu_device {
        struct device                   *dev;
@@ -346,7 +359,11 @@ struct arm_smmu_device {
        atomic_t                        irptndx;
 
        u32                             num_mapping_groups;
-       DECLARE_BITMAP(smr_map, ARM_SMMU_MAX_SMRS);
+       u16                             streamid_mask;
+       u16                             smr_mask_mask;
+       struct arm_smmu_smr             *smrs;
+       struct arm_smmu_s2cr            *s2crs;
+       struct mutex                    stream_map_mutex;
 
        unsigned long                   va_size;
        unsigned long                   ipa_size;
@@ -357,9 +374,6 @@ struct arm_smmu_device {
        u32                             num_context_irqs;
        unsigned int                    *irqs;
 
-       struct list_head                list;
-       struct rb_root                  masters;
-
        u32                             cavium_id_base; /* Specific to Cavium */
 };
 
@@ -397,15 +411,6 @@ struct arm_smmu_domain {
        struct iommu_domain             domain;
 };
 
-struct arm_smmu_phandle_args {
-       struct device_node *np;
-       int args_count;
-       uint32_t args[MAX_MASTER_STREAMIDS];
-};
-
-static DEFINE_SPINLOCK(arm_smmu_devices_lock);
-static LIST_HEAD(arm_smmu_devices);
-
 struct arm_smmu_option_prop {
        u32 opt;
        const char *prop;
@@ -413,6 +418,8 @@ struct arm_smmu_option_prop {
 
 static atomic_t cavium_smmu_context_count = ATOMIC_INIT(0);
 
+static bool using_legacy_binding, using_generic_binding;
+
 static struct arm_smmu_option_prop arm_smmu_options[] = {
        { ARM_SMMU_OPT_SECURE_CFG_ACCESS, "calxeda,smmu-secure-config-access" },
        { 0, NULL},
@@ -444,131 +451,86 @@ static struct device_node *dev_get_dev_node(struct device *dev)
 
                while (!pci_is_root_bus(bus))
                        bus = bus->parent;
-               return bus->bridge->parent->of_node;
+               return of_node_get(bus->bridge->parent->of_node);
        }
 
-       return dev->of_node;
+       return of_node_get(dev->of_node);
 }
 
-static struct arm_smmu_master *find_smmu_master(struct arm_smmu_device *smmu,
-                                               struct device_node *dev_node)
+static int __arm_smmu_get_pci_sid(struct pci_dev *pdev, u16 alias, void *data)
 {
-       struct rb_node *node = smmu->masters.rb_node;
-
-       while (node) {
-               struct arm_smmu_master *master;
-
-               master = container_of(node, struct arm_smmu_master, node);
-
-               if (dev_node < master->of_node)
-                       node = node->rb_left;
-               else if (dev_node > master->of_node)
-                       node = node->rb_right;
-               else
-                       return master;
-       }
-
-       return NULL;
+       *((__be32 *)data) = cpu_to_be32(alias);
+       return 0; /* Continue walking */
 }
 
-static struct arm_smmu_master_cfg *
-find_smmu_master_cfg(struct device *dev)
+static int __find_legacy_master_phandle(struct device *dev, void *data)
 {
-       struct arm_smmu_master_cfg *cfg = NULL;
-       struct iommu_group *group = iommu_group_get(dev);
-
-       if (group) {
-               cfg = iommu_group_get_iommudata(group);
-               iommu_group_put(group);
-       }
-
-       return cfg;
+       struct of_phandle_iterator *it = *(void **)data;
+       struct device_node *np = it->node;
+       int err;
+
+       of_for_each_phandle(it, err, dev->of_node, "mmu-masters",
+                           "#stream-id-cells", 0)
+               if (it->node == np) {
+                       *(void **)data = dev;
+                       return 1;
+               }
+       it->node = np;
+       return err == -ENOENT ? 0 : err;
 }
 
-static int insert_smmu_master(struct arm_smmu_device *smmu,
-                             struct arm_smmu_master *master)
+static struct platform_driver arm_smmu_driver;
+static struct iommu_ops arm_smmu_ops;
+
+static int arm_smmu_register_legacy_master(struct device *dev,
+                                          struct arm_smmu_device **smmu)
 {
-       struct rb_node **new, *parent;
-
-       new = &smmu->masters.rb_node;
-       parent = NULL;
-       while (*new) {
-               struct arm_smmu_master *this
-                       = container_of(*new, struct arm_smmu_master, node);
-
-               parent = *new;
-               if (master->of_node < this->of_node)
-                       new = &((*new)->rb_left);
-               else if (master->of_node > this->of_node)
-                       new = &((*new)->rb_right);
-               else
-                       return -EEXIST;
+       struct device *smmu_dev;
+       struct device_node *np;
+       struct of_phandle_iterator it;
+       void *data = &it;
+       u32 *sids;
+       __be32 pci_sid;
+       int err;
+
+       np = dev_get_dev_node(dev);
+       if (!np || !of_find_property(np, "#stream-id-cells", NULL)) {
+               of_node_put(np);
+               return -ENODEV;
        }
 
-       rb_link_node(&master->node, parent, new);
-       rb_insert_color(&master->node, &smmu->masters);
-       return 0;
-}
-
-static int register_smmu_master(struct arm_smmu_device *smmu,
-                               struct device *dev,
-                               struct arm_smmu_phandle_args *masterspec)
-{
-       int i;
-       struct arm_smmu_master *master;
+       it.node = np;
+       err = driver_for_each_device(&arm_smmu_driver.driver, NULL, &data,
+                                    __find_legacy_master_phandle);
+       smmu_dev = data;
+       of_node_put(np);
+       if (err == 0)
+               return -ENODEV;
+       if (err < 0)
+               return err;
 
-       master = find_smmu_master(smmu, masterspec->np);
-       if (master) {
-               dev_err(dev,
-                       "rejecting multiple registrations for master device %s\n",
-                       masterspec->np->name);
-               return -EBUSY;
+       if (dev_is_pci(dev)) {
+               /* "mmu-masters" assumes Stream ID == Requester ID */
+               pci_for_each_dma_alias(to_pci_dev(dev), __arm_smmu_get_pci_sid,
+                                      &pci_sid);
+               it.cur = &pci_sid;
+               it.cur_count = 1;
        }
 
-       if (masterspec->args_count > MAX_MASTER_STREAMIDS) {
-               dev_err(dev,
-                       "reached maximum number (%d) of stream IDs for master device %s\n",
-                       MAX_MASTER_STREAMIDS, masterspec->np->name);
-               return -ENOSPC;
-       }
+       err = iommu_fwspec_init(dev, &smmu_dev->of_node->fwnode,
+                               &arm_smmu_ops);
+       if (err)
+               return err;
 
-       master = devm_kzalloc(dev, sizeof(*master), GFP_KERNEL);
-       if (!master)
+       sids = kcalloc(it.cur_count, sizeof(*sids), GFP_KERNEL);
+       if (!sids)
                return -ENOMEM;
 
-       master->of_node                 = masterspec->np;
-       master->cfg.num_streamids       = masterspec->args_count;
-
-       for (i = 0; i < master->cfg.num_streamids; ++i) {
-               u16 streamid = masterspec->args[i];
-
-               if (!(smmu->features & ARM_SMMU_FEAT_STREAM_MATCH) &&
-                    (streamid >= smmu->num_mapping_groups)) {
-                       dev_err(dev,
-                               "stream ID for master device %s greater than maximum allowed (%d)\n",
-                               masterspec->np->name, smmu->num_mapping_groups);
-                       return -ERANGE;
-               }
-               master->cfg.streamids[i] = streamid;
-       }
-       return insert_smmu_master(smmu, master);
-}
-
-static struct arm_smmu_device *find_smmu_for_device(struct device *dev)
-{
-       struct arm_smmu_device *smmu;
-       struct arm_smmu_master *master = NULL;
-       struct device_node *dev_node = dev_get_dev_node(dev);
-
-       spin_lock(&arm_smmu_devices_lock);
-       list_for_each_entry(smmu, &arm_smmu_devices, list) {
-               master = find_smmu_master(smmu, dev_node);
-               if (master)
-                       break;
-       }
-       spin_unlock(&arm_smmu_devices_lock);
-
-       return master ? smmu : NULL;
+       *smmu = dev_get_drvdata(smmu_dev);
+       of_phandle_iterator_args(&it, sids, it.cur_count);
+       err = iommu_fwspec_add_ids(dev, sids, it.cur_count);
+       kfree(sids);
+       return err;
 }
 
 static int __arm_smmu_alloc_bitmap(unsigned long *map, int start, int end)
@@ -738,7 +700,7 @@ static irqreturn_t arm_smmu_global_fault(int irq, void *dev)
 static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
                                       struct io_pgtable_cfg *pgtbl_cfg)
 {
-       u32 reg;
+       u32 reg, reg2;
        u64 reg64;
        bool stage1;
        struct arm_smmu_cfg *cfg = &smmu_domain->cfg;
@@ -781,14 +743,22 @@ static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
 
        /* TTBRs */
        if (stage1) {
-               reg64 = pgtbl_cfg->arm_lpae_s1_cfg.ttbr[0];
-
-               reg64 |= ((u64)ARM_SMMU_CB_ASID(smmu, cfg)) << TTBRn_ASID_SHIFT;
-               writeq_relaxed(reg64, cb_base + ARM_SMMU_CB_TTBR0);
-
-               reg64 = pgtbl_cfg->arm_lpae_s1_cfg.ttbr[1];
-               reg64 |= ((u64)ARM_SMMU_CB_ASID(smmu, cfg)) << TTBRn_ASID_SHIFT;
-               writeq_relaxed(reg64, cb_base + ARM_SMMU_CB_TTBR1);
+               u16 asid = ARM_SMMU_CB_ASID(smmu, cfg);
+
+               if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH32_S) {
+                       reg = pgtbl_cfg->arm_v7s_cfg.ttbr[0];
+                       writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBR0);
+                       reg = pgtbl_cfg->arm_v7s_cfg.ttbr[1];
+                       writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBR1);
+                       writel_relaxed(asid, cb_base + ARM_SMMU_CB_CONTEXTIDR);
+               } else {
+                       reg64 = pgtbl_cfg->arm_lpae_s1_cfg.ttbr[0];
+                       reg64 |= (u64)asid << TTBRn_ASID_SHIFT;
+                       writeq_relaxed(reg64, cb_base + ARM_SMMU_CB_TTBR0);
+                       reg64 = pgtbl_cfg->arm_lpae_s1_cfg.ttbr[1];
+                       reg64 |= (u64)asid << TTBRn_ASID_SHIFT;
+                       writeq_relaxed(reg64, cb_base + ARM_SMMU_CB_TTBR1);
+               }
        } else {
                reg64 = pgtbl_cfg->arm_lpae_s2_cfg.vttbr;
                writeq_relaxed(reg64, cb_base + ARM_SMMU_CB_TTBR0);
@@ -796,28 +766,36 @@ static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
 
        /* TTBCR */
        if (stage1) {
-               reg = pgtbl_cfg->arm_lpae_s1_cfg.tcr;
-               writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBCR);
-               if (smmu->version > ARM_SMMU_V1) {
-                       reg = pgtbl_cfg->arm_lpae_s1_cfg.tcr >> 32;
-                       reg |= TTBCR2_SEP_UPSTREAM;
-                       writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBCR2);
+               if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH32_S) {
+                       reg = pgtbl_cfg->arm_v7s_cfg.tcr;
+                       reg2 = 0;
+               } else {
+                       reg = pgtbl_cfg->arm_lpae_s1_cfg.tcr;
+                       reg2 = pgtbl_cfg->arm_lpae_s1_cfg.tcr >> 32;
+                       reg2 |= TTBCR2_SEP_UPSTREAM;
                }
+               if (smmu->version > ARM_SMMU_V1)
+                       writel_relaxed(reg2, cb_base + ARM_SMMU_CB_TTBCR2);
        } else {
                reg = pgtbl_cfg->arm_lpae_s2_cfg.vtcr;
-               writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBCR);
        }
+       writel_relaxed(reg, cb_base + ARM_SMMU_CB_TTBCR);
 
        /* MAIRs (stage-1 only) */
        if (stage1) {
-               reg = pgtbl_cfg->arm_lpae_s1_cfg.mair[0];
+               if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH32_S) {
+                       reg = pgtbl_cfg->arm_v7s_cfg.prrr;
+                       reg2 = pgtbl_cfg->arm_v7s_cfg.nmrr;
+               } else {
+                       reg = pgtbl_cfg->arm_lpae_s1_cfg.mair[0];
+                       reg2 = pgtbl_cfg->arm_lpae_s1_cfg.mair[1];
+               }
                writel_relaxed(reg, cb_base + ARM_SMMU_CB_S1_MAIR0);
-               reg = pgtbl_cfg->arm_lpae_s1_cfg.mair[1];
-               writel_relaxed(reg, cb_base + ARM_SMMU_CB_S1_MAIR1);
+               writel_relaxed(reg2, cb_base + ARM_SMMU_CB_S1_MAIR1);
        }
 
        /* SCTLR */
-       reg = SCTLR_CFIE | SCTLR_CFRE | SCTLR_M | SCTLR_EAE_SBOP;
+       reg = SCTLR_CFIE | SCTLR_CFRE | SCTLR_AFE | SCTLR_TRE | SCTLR_M;
        if (stage1)
                reg |= SCTLR_S1_ASIDPNE;
 #ifdef __BIG_ENDIAN
@@ -841,12 +819,6 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
        if (smmu_domain->smmu)
                goto out_unlock;
 
-       /* We're bypassing these SIDs, so don't allocate an actual context */
-       if (domain->type == IOMMU_DOMAIN_DMA) {
-               smmu_domain->smmu = smmu;
-               goto out_unlock;
-       }
-
        /*
         * Mapping the requested stage onto what we support is surprisingly
         * complicated, mainly because the spec allows S1+S2 SMMUs without
@@ -880,6 +852,11 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
         */
        if (smmu->features & ARM_SMMU_FEAT_FMT_AARCH32_L)
                cfg->fmt = ARM_SMMU_CTX_FMT_AARCH32_L;
+       if (IS_ENABLED(CONFIG_IOMMU_IO_PGTABLE_ARMV7S) &&
+           !IS_ENABLED(CONFIG_64BIT) && !IS_ENABLED(CONFIG_ARM_LPAE) &&
+           (smmu->features & ARM_SMMU_FEAT_FMT_AARCH32_S) &&
+           (smmu_domain->stage == ARM_SMMU_DOMAIN_S1))
+               cfg->fmt = ARM_SMMU_CTX_FMT_AARCH32_S;
        if ((IS_ENABLED(CONFIG_64BIT) || cfg->fmt == ARM_SMMU_CTX_FMT_NONE) &&
            (smmu->features & (ARM_SMMU_FEAT_FMT_AARCH64_64K |
                               ARM_SMMU_FEAT_FMT_AARCH64_16K |
@@ -899,10 +876,14 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
                oas = smmu->ipa_size;
                if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH64) {
                        fmt = ARM_64_LPAE_S1;
-               } else {
+               } else if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH32_L) {
                        fmt = ARM_32_LPAE_S1;
                        ias = min(ias, 32UL);
                        oas = min(oas, 40UL);
+               } else {
+                       fmt = ARM_V7S;
+                       ias = min(ias, 32UL);
+                       oas = min(oas, 32UL);
                }
                break;
        case ARM_SMMU_DOMAIN_NESTED:
@@ -958,6 +939,8 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
 
        /* Update the domain's page sizes to reflect the page table format */
        domain->pgsize_bitmap = pgtbl_cfg.pgsize_bitmap;
+       domain->geometry.aperture_end = (1UL << ias) - 1;
+       domain->geometry.force_aperture = true;
 
        /* Initialise the context bank with our page table cfg */
        arm_smmu_init_context_bank(smmu_domain, &pgtbl_cfg);
@@ -996,7 +979,7 @@ static void arm_smmu_destroy_domain_context(struct iommu_domain *domain)
        void __iomem *cb_base;
        int irq;
 
-       if (!smmu || domain->type == IOMMU_DOMAIN_DMA)
+       if (!smmu)
                return;
 
        /*
@@ -1030,8 +1013,8 @@ static struct iommu_domain *arm_smmu_domain_alloc(unsigned type)
        if (!smmu_domain)
                return NULL;
 
-       if (type == IOMMU_DOMAIN_DMA &&
-           iommu_get_dma_cookie(&smmu_domain->domain)) {
+       if (type == IOMMU_DOMAIN_DMA && (using_legacy_binding ||
+           iommu_get_dma_cookie(&smmu_domain->domain))) {
                kfree(smmu_domain);
                return NULL;
        }
@@ -1055,162 +1038,197 @@ static void arm_smmu_domain_free(struct iommu_domain *domain)
        kfree(smmu_domain);
 }
 
-static int arm_smmu_master_configure_smrs(struct arm_smmu_device *smmu,
-                                         struct arm_smmu_master_cfg *cfg)
+static void arm_smmu_write_smr(struct arm_smmu_device *smmu, int idx)
 {
-       int i;
-       struct arm_smmu_smr *smrs;
-       void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
+       struct arm_smmu_smr *smr = smmu->smrs + idx;
+       u32 reg = smr->id << SMR_ID_SHIFT | smr->mask << SMR_MASK_SHIFT;
 
-       if (!(smmu->features & ARM_SMMU_FEAT_STREAM_MATCH))
-               return 0;
+       if (smr->valid)
+               reg |= SMR_VALID;
+       writel_relaxed(reg, ARM_SMMU_GR0(smmu) + ARM_SMMU_GR0_SMR(idx));
+}
 
-       if (cfg->smrs)
-               return -EEXIST;
+static void arm_smmu_write_s2cr(struct arm_smmu_device *smmu, int idx)
+{
+       struct arm_smmu_s2cr *s2cr = smmu->s2crs + idx;
+       u32 reg = (s2cr->type & S2CR_TYPE_MASK) << S2CR_TYPE_SHIFT |
+                 (s2cr->cbndx & S2CR_CBNDX_MASK) << S2CR_CBNDX_SHIFT |
+                 (s2cr->privcfg & S2CR_PRIVCFG_MASK) << S2CR_PRIVCFG_SHIFT;
 
-       smrs = kmalloc_array(cfg->num_streamids, sizeof(*smrs), GFP_KERNEL);
-       if (!smrs) {
-               dev_err(smmu->dev, "failed to allocate %d SMRs\n",
-                       cfg->num_streamids);
-               return -ENOMEM;
-       }
+       writel_relaxed(reg, ARM_SMMU_GR0(smmu) + ARM_SMMU_GR0_S2CR(idx));
+}
 
-       /* Allocate the SMRs on the SMMU */
-       for (i = 0; i < cfg->num_streamids; ++i) {
-               int idx = __arm_smmu_alloc_bitmap(smmu->smr_map, 0,
-                                                 smmu->num_mapping_groups);
-               if (idx < 0) {
-                       dev_err(smmu->dev, "failed to allocate free SMR\n");
-                       goto err_free_smrs;
-               }
+static void arm_smmu_write_sme(struct arm_smmu_device *smmu, int idx)
+{
+       arm_smmu_write_s2cr(smmu, idx);
+       if (smmu->smrs)
+               arm_smmu_write_smr(smmu, idx);
+}
 
-               smrs[i] = (struct arm_smmu_smr) {
-                       .idx    = idx,
-                       .mask   = 0, /* We don't currently share SMRs */
-                       .id     = cfg->streamids[i],
-               };
-       }
+static int arm_smmu_find_sme(struct arm_smmu_device *smmu, u16 id, u16 mask)
+{
+       struct arm_smmu_smr *smrs = smmu->smrs;
+       int i, free_idx = -ENOSPC;
 
-       /* It worked! Now, poke the actual hardware */
-       for (i = 0; i < cfg->num_streamids; ++i) {
-               u32 reg = SMR_VALID | smrs[i].id << SMR_ID_SHIFT |
-                         smrs[i].mask << SMR_MASK_SHIFT;
-               writel_relaxed(reg, gr0_base + ARM_SMMU_GR0_SMR(smrs[i].idx));
-       }
+       /* Stream indexing is blissfully easy */
+       if (!smrs)
+               return id;
 
-       cfg->smrs = smrs;
-       return 0;
+       /* Validating SMRs is... less so */
+       for (i = 0; i < smmu->num_mapping_groups; ++i) {
+               if (!smrs[i].valid) {
+                       /*
+                        * Note the first free entry we come across, which
+                        * we'll claim in the end if nothing else matches.
+                        */
+                       if (free_idx < 0)
+                               free_idx = i;
+                       continue;
+               }
+               /*
+                * If the new entry is _entirely_ matched by an existing entry,
+                * then reuse that, with the guarantee that there also cannot
+                * be any subsequent conflicting entries. In normal use we'd
+                * expect simply identical entries for this case, but there's
+                * no harm in accommodating the generalisation.
+                */
+               if ((mask & smrs[i].mask) == mask &&
+                   !((id ^ smrs[i].id) & ~smrs[i].mask))
+                       return i;
+               /*
+                * If the new entry has any other overlap with an existing one,
+                * though, then there always exists at least one stream ID
+                * which would cause a conflict, and we can't allow that risk.
+                */
+               if (!((id ^ smrs[i].id) & ~(smrs[i].mask | mask)))
+                       return -EINVAL;
+       }
 
-err_free_smrs:
-       while (--i >= 0)
-               __arm_smmu_free_bitmap(smmu->smr_map, smrs[i].idx);
-       kfree(smrs);
-       return -ENOSPC;
+       return free_idx;
 }
 
-static void arm_smmu_master_free_smrs(struct arm_smmu_device *smmu,
-                                     struct arm_smmu_master_cfg *cfg)
+static bool arm_smmu_free_sme(struct arm_smmu_device *smmu, int idx)
 {
-       int i;
-       void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
-       struct arm_smmu_smr *smrs = cfg->smrs;
-
-       if (!smrs)
-               return;
-
-       /* Invalidate the SMRs before freeing back to the allocator */
-       for (i = 0; i < cfg->num_streamids; ++i) {
-               u8 idx = smrs[i].idx;
+       if (--smmu->s2crs[idx].count)
+               return false;
 
-               writel_relaxed(~SMR_VALID, gr0_base + ARM_SMMU_GR0_SMR(idx));
-               __arm_smmu_free_bitmap(smmu->smr_map, idx);
-       }
+       smmu->s2crs[idx] = s2cr_init_val;
+       if (smmu->smrs)
+               smmu->smrs[idx].valid = false;
 
-       cfg->smrs = NULL;
-       kfree(smrs);
+       return true;
 }
 
-static int arm_smmu_domain_add_master(struct arm_smmu_domain *smmu_domain,
-                                     struct arm_smmu_master_cfg *cfg)
+static int arm_smmu_master_alloc_smes(struct device *dev)
 {
-       int i, ret;
-       struct arm_smmu_device *smmu = smmu_domain->smmu;
-       void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
+       struct iommu_fwspec *fwspec = dev->iommu_fwspec;
+       struct arm_smmu_master_cfg *cfg = fwspec->iommu_priv;
+       struct arm_smmu_device *smmu = cfg->smmu;
+       struct arm_smmu_smr *smrs = smmu->smrs;
+       struct iommu_group *group;
+       int i, idx, ret;
 
-       /*
-        * FIXME: This won't be needed once we have IOMMU-backed DMA ops
-        * for all devices behind the SMMU. Note that we need to take
-        * care configuring SMRs for devices both a platform_device and
-        * and a PCI device (i.e. a PCI host controller)
-        */
-       if (smmu_domain->domain.type == IOMMU_DOMAIN_DMA)
-               return 0;
+       mutex_lock(&smmu->stream_map_mutex);
+       /* Figure out a viable stream map entry allocation */
+       for_each_cfg_sme(fwspec, i, idx) {
+               u16 sid = fwspec->ids[i];
+               u16 mask = fwspec->ids[i] >> SMR_MASK_SHIFT;
 
-       /* Devices in an IOMMU group may already be configured */
-       ret = arm_smmu_master_configure_smrs(smmu, cfg);
-       if (ret)
-               return ret == -EEXIST ? 0 : ret;
+               if (idx != INVALID_SMENDX) {
+                       ret = -EEXIST;
+                       goto out_err;
+               }
 
-       for (i = 0; i < cfg->num_streamids; ++i) {
-               u32 idx, s2cr;
+               ret = arm_smmu_find_sme(smmu, sid, mask);
+               if (ret < 0)
+                       goto out_err;
+
+               idx = ret;
+               if (smrs && smmu->s2crs[idx].count == 0) {
+                       smrs[idx].id = sid;
+                       smrs[idx].mask = mask;
+                       smrs[idx].valid = true;
+               }
+               smmu->s2crs[idx].count++;
+               cfg->smendx[i] = (s16)idx;
+       }
 
-               idx = cfg->smrs ? cfg->smrs[i].idx : cfg->streamids[i];
-               s2cr = S2CR_TYPE_TRANS | S2CR_PRIVCFG_UNPRIV |
-                      (smmu_domain->cfg.cbndx << S2CR_CBNDX_SHIFT);
-               writel_relaxed(s2cr, gr0_base + ARM_SMMU_GR0_S2CR(idx));
+       group = iommu_group_get_for_dev(dev);
+       if (!group)
+               group = ERR_PTR(-ENOMEM);
+       if (IS_ERR(group)) {
+               ret = PTR_ERR(group);
+               goto out_err;
        }
+       iommu_group_put(group);
 
+       /* It worked! Now, poke the actual hardware */
+       for_each_cfg_sme(fwspec, i, idx) {
+               arm_smmu_write_sme(smmu, idx);
+               smmu->s2crs[idx].group = group;
+       }
+
+       mutex_unlock(&smmu->stream_map_mutex);
        return 0;
+
+out_err:
+       while (i--) {
+               arm_smmu_free_sme(smmu, cfg->smendx[i]);
+               cfg->smendx[i] = INVALID_SMENDX;
+       }
+       mutex_unlock(&smmu->stream_map_mutex);
+       return ret;
 }
 
-static void arm_smmu_domain_remove_master(struct arm_smmu_domain *smmu_domain,
-                                         struct arm_smmu_master_cfg *cfg)
+static void arm_smmu_master_free_smes(struct iommu_fwspec *fwspec)
 {
-       int i;
-       struct arm_smmu_device *smmu = smmu_domain->smmu;
-       void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
-
-       /* An IOMMU group is torn down by the first device to be removed */
-       if ((smmu->features & ARM_SMMU_FEAT_STREAM_MATCH) && !cfg->smrs)
-               return;
+       struct arm_smmu_device *smmu = fwspec_smmu(fwspec);
+       struct arm_smmu_master_cfg *cfg = fwspec->iommu_priv;
+       int i, idx;
 
-       /*
-        * We *must* clear the S2CR first, because freeing the SMR means
-        * that it can be re-allocated immediately.
-        */
-       for (i = 0; i < cfg->num_streamids; ++i) {
-               u32 idx = cfg->smrs ? cfg->smrs[i].idx : cfg->streamids[i];
-               u32 reg = disable_bypass ? S2CR_TYPE_FAULT : S2CR_TYPE_BYPASS;
-
-               writel_relaxed(reg, gr0_base + ARM_SMMU_GR0_S2CR(idx));
+       mutex_lock(&smmu->stream_map_mutex);
+       for_each_cfg_sme(fwspec, i, idx) {
+               if (arm_smmu_free_sme(smmu, idx))
+                       arm_smmu_write_sme(smmu, idx);
+               cfg->smendx[i] = INVALID_SMENDX;
        }
-
-       arm_smmu_master_free_smrs(smmu, cfg);
+       mutex_unlock(&smmu->stream_map_mutex);
 }
 
-static void arm_smmu_detach_dev(struct device *dev,
-                               struct arm_smmu_master_cfg *cfg)
+static int arm_smmu_domain_add_master(struct arm_smmu_domain *smmu_domain,
+                                     struct iommu_fwspec *fwspec)
 {
-       struct iommu_domain *domain = dev->archdata.iommu;
-       struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
+       struct arm_smmu_device *smmu = smmu_domain->smmu;
+       struct arm_smmu_s2cr *s2cr = smmu->s2crs;
+       enum arm_smmu_s2cr_type type = S2CR_TYPE_TRANS;
+       u8 cbndx = smmu_domain->cfg.cbndx;
+       int i, idx;
+
+       for_each_cfg_sme(fwspec, i, idx) {
+               if (type == s2cr[idx].type && cbndx == s2cr[idx].cbndx)
+                       continue;
 
-       dev->archdata.iommu = NULL;
-       arm_smmu_domain_remove_master(smmu_domain, cfg);
+               s2cr[idx].type = type;
+               s2cr[idx].privcfg = S2CR_PRIVCFG_UNPRIV;
+               s2cr[idx].cbndx = cbndx;
+               arm_smmu_write_s2cr(smmu, idx);
+       }
+       return 0;
 }
 
 static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
 {
        int ret;
-       struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
+       struct iommu_fwspec *fwspec = dev->iommu_fwspec;
        struct arm_smmu_device *smmu;
-       struct arm_smmu_master_cfg *cfg;
+       struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
 
-       smmu = find_smmu_for_device(dev);
-       if (!smmu) {
+       if (!fwspec || fwspec->ops != &arm_smmu_ops) {
                dev_err(dev, "cannot attach to SMMU, is it on the same bus?\n");
                return -ENXIO;
        }
 
+       smmu = fwspec_smmu(fwspec);
        /* Ensure that the domain is finalised */
        ret = arm_smmu_init_domain_context(domain, smmu);
        if (ret < 0)
@@ -1228,18 +1246,7 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
        }
 
        /* Looks ok, so add the device to the domain */
-       cfg = find_smmu_master_cfg(dev);
-       if (!cfg)
-               return -ENODEV;
-
-       /* Detach the dev from its current domain */
-       if (dev->archdata.iommu)
-               arm_smmu_detach_dev(dev, cfg);
-
-       ret = arm_smmu_domain_add_master(smmu_domain, cfg);
-       if (!ret)
-               dev->archdata.iommu = domain;
-       return ret;
+       return arm_smmu_domain_add_master(smmu_domain, fwspec);
 }
 
 static int arm_smmu_map(struct iommu_domain *domain, unsigned long iova,
@@ -1358,110 +1365,113 @@ static bool arm_smmu_capable(enum iommu_cap cap)
        }
 }
 
-static int __arm_smmu_get_pci_sid(struct pci_dev *pdev, u16 alias, void *data)
+static int arm_smmu_match_node(struct device *dev, void *data)
 {
-       *((u16 *)data) = alias;
-       return 0; /* Continue walking */
+       return dev->of_node == data;
 }
 
-static void __arm_smmu_release_pci_iommudata(void *data)
+static struct arm_smmu_device *arm_smmu_get_by_node(struct device_node *np)
 {
-       kfree(data);
+       struct device *dev = driver_find_device(&arm_smmu_driver.driver, NULL,
+                                               np, arm_smmu_match_node);
+       put_device(dev);
+       return dev ? dev_get_drvdata(dev) : NULL;
 }
 
-static int arm_smmu_init_pci_device(struct pci_dev *pdev,
-                                   struct iommu_group *group)
+static int arm_smmu_add_device(struct device *dev)
 {
+       struct arm_smmu_device *smmu;
        struct arm_smmu_master_cfg *cfg;
-       u16 sid;
-       int i;
-
-       cfg = iommu_group_get_iommudata(group);
-       if (!cfg) {
-               cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
-               if (!cfg)
-                       return -ENOMEM;
+       struct iommu_fwspec *fwspec = dev->iommu_fwspec;
+       int i, ret;
 
-               iommu_group_set_iommudata(group, cfg,
-                                         __arm_smmu_release_pci_iommudata);
+       if (using_legacy_binding) {
+               ret = arm_smmu_register_legacy_master(dev, &smmu);
+               fwspec = dev->iommu_fwspec;
+               if (ret)
+                       goto out_free;
+       } else if (fwspec) {
+               smmu = arm_smmu_get_by_node(to_of_node(fwspec->iommu_fwnode));
+       } else {
+               return -ENODEV;
        }
 
-       if (cfg->num_streamids >= MAX_MASTER_STREAMIDS)
-               return -ENOSPC;
+       ret = -EINVAL;
+       for (i = 0; i < fwspec->num_ids; i++) {
+               u16 sid = fwspec->ids[i];
+               u16 mask = fwspec->ids[i] >> SMR_MASK_SHIFT;
 
-       /*
-        * Assume Stream ID == Requester ID for now.
-        * We need a way to describe the ID mappings in FDT.
-        */
-       pci_for_each_dma_alias(pdev, __arm_smmu_get_pci_sid, &sid);
-       for (i = 0; i < cfg->num_streamids; ++i)
-               if (cfg->streamids[i] == sid)
-                       break;
-
-       /* Avoid duplicate SIDs, as this can lead to SMR conflicts */
-       if (i == cfg->num_streamids)
-               cfg->streamids[cfg->num_streamids++] = sid;
-
-       return 0;
-}
-
-static int arm_smmu_init_platform_device(struct device *dev,
-                                        struct iommu_group *group)
-{
-       struct arm_smmu_device *smmu = find_smmu_for_device(dev);
-       struct arm_smmu_master *master;
+               if (sid & ~smmu->streamid_mask) {
+                       dev_err(dev, "stream ID 0x%x out of range for SMMU (0x%x)\n",
+                               sid, smmu->streamid_mask);
+                       goto out_free;
+               }
+               if (mask & ~smmu->smr_mask_mask) {
+                       dev_err(dev, "SMR mask 0x%x out of range for SMMU (0x%x)\n",
+                               sid, smmu->smr_mask_mask);
+                       goto out_free;
+               }
+       }
 
-       if (!smmu)
-               return -ENODEV;
+       ret = -ENOMEM;
+       cfg = kzalloc(offsetof(struct arm_smmu_master_cfg, smendx[i]),
+                     GFP_KERNEL);
+       if (!cfg)
+               goto out_free;
 
-       master = find_smmu_master(smmu, dev->of_node);
-       if (!master)
-               return -ENODEV;
+       cfg->smmu = smmu;
+       fwspec->iommu_priv = cfg;
+       while (i--)
+               cfg->smendx[i] = INVALID_SMENDX;
 
-       iommu_group_set_iommudata(group, &master->cfg, NULL);
+       ret = arm_smmu_master_alloc_smes(dev);
+       if (ret)
+               goto out_free;
 
        return 0;
-}
 
-static int arm_smmu_add_device(struct device *dev)
-{
-       struct iommu_group *group;
-
-       group = iommu_group_get_for_dev(dev);
-       if (IS_ERR(group))
-               return PTR_ERR(group);
-
-       iommu_group_put(group);
-       return 0;
+out_free:
+       if (fwspec)
+               kfree(fwspec->iommu_priv);
+       iommu_fwspec_free(dev);
+       return ret;
 }
 
 static void arm_smmu_remove_device(struct device *dev)
 {
+       struct iommu_fwspec *fwspec = dev->iommu_fwspec;
+
+       if (!fwspec || fwspec->ops != &arm_smmu_ops)
+               return;
+
+       arm_smmu_master_free_smes(fwspec);
        iommu_group_remove_device(dev);
+       kfree(fwspec->iommu_priv);
+       iommu_fwspec_free(dev);
 }
 
 static struct iommu_group *arm_smmu_device_group(struct device *dev)
 {
-       struct iommu_group *group;
-       int ret;
+       struct iommu_fwspec *fwspec = dev->iommu_fwspec;
+       struct arm_smmu_device *smmu = fwspec_smmu(fwspec);
+       struct iommu_group *group = NULL;
+       int i, idx;
 
-       if (dev_is_pci(dev))
-               group = pci_device_group(dev);
-       else
-               group = generic_device_group(dev);
+       for_each_cfg_sme(fwspec, i, idx) {
+               if (group && smmu->s2crs[idx].group &&
+                   group != smmu->s2crs[idx].group)
+                       return ERR_PTR(-EINVAL);
+
+               group = smmu->s2crs[idx].group;
+       }
 
-       if (IS_ERR(group))
+       if (group)
                return group;
 
        if (dev_is_pci(dev))
-               ret = arm_smmu_init_pci_device(to_pci_dev(dev), group);
+               group = pci_device_group(dev);
        else
-               ret = arm_smmu_init_platform_device(dev, group);
-
-       if (ret) {
-               iommu_group_put(group);
-               group = ERR_PTR(ret);
-       }
+               group = generic_device_group(dev);
 
        return group;
 }
@@ -1510,6 +1520,19 @@ out_unlock:
        return ret;
 }
 
+static int arm_smmu_of_xlate(struct device *dev, struct of_phandle_args *args)
+{
+       u32 fwid = 0;
+
+       if (args->args_count > 0)
+               fwid |= (u16)args->args[0];
+
+       if (args->args_count > 1)
+               fwid |= (u16)args->args[1] << SMR_MASK_SHIFT;
+
+       return iommu_fwspec_add_ids(dev, &fwid, 1);
+}
+
 static struct iommu_ops arm_smmu_ops = {
        .capable                = arm_smmu_capable,
        .domain_alloc           = arm_smmu_domain_alloc,
@@ -1524,6 +1547,7 @@ static struct iommu_ops arm_smmu_ops = {
        .device_group           = arm_smmu_device_group,
        .domain_get_attr        = arm_smmu_domain_get_attr,
        .domain_set_attr        = arm_smmu_domain_set_attr,
+       .of_xlate               = arm_smmu_of_xlate,
        .pgsize_bitmap          = -1UL, /* Restricted during device attach */
 };
 
@@ -1531,19 +1555,19 @@ static void arm_smmu_device_reset(struct arm_smmu_device *smmu)
 {
        void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
        void __iomem *cb_base;
-       int i = 0;
+       int i;
        u32 reg, major;
 
        /* clear global FSR */
        reg = readl_relaxed(ARM_SMMU_GR0_NS(smmu) + ARM_SMMU_GR0_sGFSR);
        writel(reg, ARM_SMMU_GR0_NS(smmu) + ARM_SMMU_GR0_sGFSR);
 
-       /* Mark all SMRn as invalid and all S2CRn as bypass unless overridden */
-       reg = disable_bypass ? S2CR_TYPE_FAULT : S2CR_TYPE_BYPASS;
-       for (i = 0; i < smmu->num_mapping_groups; ++i) {
-               writel_relaxed(0, gr0_base + ARM_SMMU_GR0_SMR(i));
-               writel_relaxed(reg, gr0_base + ARM_SMMU_GR0_S2CR(i));
-       }
+       /*
+        * Reset stream mapping groups: Initial values mark all SMRn as
+        * invalid and all S2CRn as bypass unless overridden.
+        */
+       for (i = 0; i < smmu->num_mapping_groups; ++i)
+               arm_smmu_write_sme(smmu, i);
 
        /*
         * Before clearing ARM_MMU500_ACTLR_CPRE, need to
@@ -1632,6 +1656,7 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
        void __iomem *gr0_base = ARM_SMMU_GR0(smmu);
        u32 id;
        bool cttw_dt, cttw_reg;
+       int i;
 
        dev_notice(smmu->dev, "probing hardware configuration...\n");
        dev_notice(smmu->dev, "SMMUv%d with:\n",
@@ -1690,39 +1715,55 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
                dev_notice(smmu->dev,
                           "\t(IDR0.CTTW overridden by dma-coherent property)\n");
 
+       /* Max. number of entries we have for stream matching/indexing */
+       size = 1 << ((id >> ID0_NUMSIDB_SHIFT) & ID0_NUMSIDB_MASK);
+       smmu->streamid_mask = size - 1;
        if (id & ID0_SMS) {
-               u32 smr, sid, mask;
+               u32 smr;
 
                smmu->features |= ARM_SMMU_FEAT_STREAM_MATCH;
-               smmu->num_mapping_groups = (id >> ID0_NUMSMRG_SHIFT) &
-                                          ID0_NUMSMRG_MASK;
-               if (smmu->num_mapping_groups == 0) {
+               size = (id >> ID0_NUMSMRG_SHIFT) & ID0_NUMSMRG_MASK;
+               if (size == 0) {
                        dev_err(smmu->dev,
                                "stream-matching supported, but no SMRs present!\n");
                        return -ENODEV;
                }
 
-               smr = SMR_MASK_MASK << SMR_MASK_SHIFT;
-               smr |= (SMR_ID_MASK << SMR_ID_SHIFT);
+               /*
+                * SMR.ID bits may not be preserved if the corresponding MASK
+                * bits are set, so check each one separately. We can reject
+                * masters later if they try to claim IDs outside these masks.
+                */
+               smr = smmu->streamid_mask << SMR_ID_SHIFT;
                writel_relaxed(smr, gr0_base + ARM_SMMU_GR0_SMR(0));
                smr = readl_relaxed(gr0_base + ARM_SMMU_GR0_SMR(0));
+               smmu->streamid_mask = smr >> SMR_ID_SHIFT;
 
-               mask = (smr >> SMR_MASK_SHIFT) & SMR_MASK_MASK;
-               sid = (smr >> SMR_ID_SHIFT) & SMR_ID_MASK;
-               if ((mask & sid) != sid) {
-                       dev_err(smmu->dev,
-                               "SMR mask bits (0x%x) insufficient for ID field (0x%x)\n",
-                               mask, sid);
-                       return -ENODEV;
-               }
+               smr = smmu->streamid_mask << SMR_MASK_SHIFT;
+               writel_relaxed(smr, gr0_base + ARM_SMMU_GR0_SMR(0));
+               smr = readl_relaxed(gr0_base + ARM_SMMU_GR0_SMR(0));
+               smmu->smr_mask_mask = smr >> SMR_MASK_SHIFT;
+
+               /* Zero-initialised to mark as invalid */
+               smmu->smrs = devm_kcalloc(smmu->dev, size, sizeof(*smmu->smrs),
+                                         GFP_KERNEL);
+               if (!smmu->smrs)
+                       return -ENOMEM;
 
                dev_notice(smmu->dev,
-                          "\tstream matching with %u register groups, mask 0x%x",
-                          smmu->num_mapping_groups, mask);
-       } else {
-               smmu->num_mapping_groups = (id >> ID0_NUMSIDB_SHIFT) &
-                                          ID0_NUMSIDB_MASK;
+                          "\tstream matching with %lu register groups, mask 0x%x",
+                          size, smmu->smr_mask_mask);
        }
+       /* s2cr->type == 0 means translation, so initialise explicitly */
+       smmu->s2crs = devm_kmalloc_array(smmu->dev, size, sizeof(*smmu->s2crs),
+                                        GFP_KERNEL);
+       if (!smmu->s2crs)
+               return -ENOMEM;
+       for (i = 0; i < size; i++)
+               smmu->s2crs[i] = s2cr_init_val;
+
+       smmu->num_mapping_groups = size;
+       mutex_init(&smmu->stream_map_mutex);
 
        if (smmu->version < ARM_SMMU_V2 || !(id & ID0_PTFS_NO_AARCH32)) {
                smmu->features |= ARM_SMMU_FEAT_FMT_AARCH32_L;
@@ -1855,15 +1896,24 @@ MODULE_DEVICE_TABLE(of, arm_smmu_of_match);
 
 static int arm_smmu_device_dt_probe(struct platform_device *pdev)
 {
-       const struct of_device_id *of_id;
        const struct arm_smmu_match_data *data;
        struct resource *res;
        struct arm_smmu_device *smmu;
        struct device *dev = &pdev->dev;
-       struct rb_node *node;
-       struct of_phandle_iterator it;
-       struct arm_smmu_phandle_args *masterspec;
        int num_irqs, i, err;
+       bool legacy_binding;
+
+       legacy_binding = of_find_property(dev->of_node, "mmu-masters", NULL);
+       if (legacy_binding && !using_generic_binding) {
+               if (!using_legacy_binding)
+                       pr_notice("deprecated \"mmu-masters\" DT property in use; DMA API support unavailable\n");
+               using_legacy_binding = true;
+       } else if (!legacy_binding && !using_legacy_binding) {
+               using_generic_binding = true;
+       } else {
+               dev_err(dev, "not probing due to mismatched DT properties\n");
+               return -ENODEV;
+       }
 
        smmu = devm_kzalloc(dev, sizeof(*smmu), GFP_KERNEL);
        if (!smmu) {
@@ -1872,8 +1922,7 @@ static int arm_smmu_device_dt_probe(struct platform_device *pdev)
        }
        smmu->dev = dev;
 
-       of_id = of_match_node(arm_smmu_of_match, dev->of_node);
-       data = of_id->data;
+       data = of_device_get_match_data(dev);
        smmu->version = data->version;
        smmu->model = data->model;
 
@@ -1923,37 +1972,6 @@ static int arm_smmu_device_dt_probe(struct platform_device *pdev)
        if (err)
                return err;
 
-       i = 0;
-       smmu->masters = RB_ROOT;
-
-       err = -ENOMEM;
-       /* No need to zero the memory for masterspec */
-       masterspec = kmalloc(sizeof(*masterspec), GFP_KERNEL);
-       if (!masterspec)
-               goto out_put_masters;
-
-       of_for_each_phandle(&it, err, dev->of_node,
-                           "mmu-masters", "#stream-id-cells", 0) {
-               int count = of_phandle_iterator_args(&it, masterspec->args,
-                                                    MAX_MASTER_STREAMIDS);
-               masterspec->np          = of_node_get(it.node);
-               masterspec->args_count  = count;
-
-               err = register_smmu_master(smmu, dev, masterspec);
-               if (err) {
-                       dev_err(dev, "failed to add master %s\n",
-                               masterspec->np->name);
-                       kfree(masterspec);
-                       goto out_put_masters;
-               }
-
-               i++;
-       }
-
-       dev_notice(dev, "registered %d master devices\n", i);
-
-       kfree(masterspec);
-
        parse_driver_options(smmu);
 
        if (smmu->version == ARM_SMMU_V2 &&
@@ -1961,8 +1979,7 @@ static int arm_smmu_device_dt_probe(struct platform_device *pdev)
                dev_err(dev,
                        "found only %d context interrupt(s) but %d required\n",
                        smmu->num_context_irqs, smmu->num_context_banks);
-               err = -ENODEV;
-               goto out_put_masters;
+               return -ENODEV;
        }
 
        for (i = 0; i < smmu->num_global_irqs; ++i) {
@@ -1974,59 +1991,39 @@ static int arm_smmu_device_dt_probe(struct platform_device *pdev)
                if (err) {
                        dev_err(dev, "failed to request global IRQ %d (%u)\n",
                                i, smmu->irqs[i]);
-                       goto out_put_masters;
+                       return err;
                }
        }
 
-       INIT_LIST_HEAD(&smmu->list);
-       spin_lock(&arm_smmu_devices_lock);
-       list_add(&smmu->list, &arm_smmu_devices);
-       spin_unlock(&arm_smmu_devices_lock);
-
+       of_iommu_set_ops(dev->of_node, &arm_smmu_ops);
+       platform_set_drvdata(pdev, smmu);
        arm_smmu_device_reset(smmu);
-       return 0;
 
-out_put_masters:
-       for (node = rb_first(&smmu->masters); node; node = rb_next(node)) {
-               struct arm_smmu_master *master
-                       = container_of(node, struct arm_smmu_master, node);
-               of_node_put(master->of_node);
+       /* Oh, for a proper bus abstraction */
+       if (!iommu_present(&platform_bus_type))
+               bus_set_iommu(&platform_bus_type, &arm_smmu_ops);
+#ifdef CONFIG_ARM_AMBA
+       if (!iommu_present(&amba_bustype))
+               bus_set_iommu(&amba_bustype, &arm_smmu_ops);
+#endif
+#ifdef CONFIG_PCI
+       if (!iommu_present(&pci_bus_type)) {
+               pci_request_acs();
+               bus_set_iommu(&pci_bus_type, &arm_smmu_ops);
        }
-
-       return err;
+#endif
+       return 0;
 }
 
 static int arm_smmu_device_remove(struct platform_device *pdev)
 {
-       int i;
-       struct device *dev = &pdev->dev;
-       struct arm_smmu_device *curr, *smmu = NULL;
-       struct rb_node *node;
-
-       spin_lock(&arm_smmu_devices_lock);
-       list_for_each_entry(curr, &arm_smmu_devices, list) {
-               if (curr->dev == dev) {
-                       smmu = curr;
-                       list_del(&smmu->list);
-                       break;
-               }
-       }
-       spin_unlock(&arm_smmu_devices_lock);
+       struct arm_smmu_device *smmu = platform_get_drvdata(pdev);
 
        if (!smmu)
                return -ENODEV;
 
-       for (node = rb_first(&smmu->masters); node; node = rb_next(node)) {
-               struct arm_smmu_master *master
-                       = container_of(node, struct arm_smmu_master, node);
-               of_node_put(master->of_node);
-       }
-
        if (!bitmap_empty(smmu->context_map, ARM_SMMU_MAX_CBS))
-               dev_err(dev, "removing device with active domains!\n");
-
-       for (i = 0; i < smmu->num_global_irqs; ++i)
-               devm_free_irq(smmu->dev, smmu->irqs[i], smmu);
+               dev_err(&pdev->dev, "removing device with active domains!\n");
 
        /* Turn the thing off */
        writel(sCR0_CLIENTPD, ARM_SMMU_GR0_NS(smmu) + ARM_SMMU_GR0_sCR0);
@@ -2044,41 +2041,14 @@ static struct platform_driver arm_smmu_driver = {
 
 static int __init arm_smmu_init(void)
 {
-       struct device_node *np;
-       int ret;
-
-       /*
-        * Play nice with systems that don't have an ARM SMMU by checking that
-        * an ARM SMMU exists in the system before proceeding with the driver
-        * and IOMMU bus operation registration.
-        */
-       np = of_find_matching_node(NULL, arm_smmu_of_match);
-       if (!np)
-               return 0;
-
-       of_node_put(np);
-
-       ret = platform_driver_register(&arm_smmu_driver);
-       if (ret)
-               return ret;
-
-       /* Oh, for a proper bus abstraction */
-       if (!iommu_present(&platform_bus_type))
-               bus_set_iommu(&platform_bus_type, &arm_smmu_ops);
-
-#ifdef CONFIG_ARM_AMBA
-       if (!iommu_present(&amba_bustype))
-               bus_set_iommu(&amba_bustype, &arm_smmu_ops);
-#endif
+       static bool registered;
+       int ret = 0;
 
-#ifdef CONFIG_PCI
-       if (!iommu_present(&pci_bus_type)) {
-               pci_request_acs();
-               bus_set_iommu(&pci_bus_type, &arm_smmu_ops);
+       if (!registered) {
+               ret = platform_driver_register(&arm_smmu_driver);
+               registered = !ret;
        }
-#endif
-
-       return 0;
+       return ret;
 }
 
 static void __exit arm_smmu_exit(void)
@@ -2089,6 +2059,25 @@ static void __exit arm_smmu_exit(void)
 subsys_initcall(arm_smmu_init);
 module_exit(arm_smmu_exit);
 
+static int __init arm_smmu_of_init(struct device_node *np)
+{
+       int ret = arm_smmu_init();
+
+       if (ret)
+               return ret;
+
+       if (!of_platform_device_create(np, NULL, platform_bus_type.dev_root))
+               return -ENODEV;
+
+       return 0;
+}
+IOMMU_OF_DECLARE(arm_smmuv1, "arm,smmu-v1", arm_smmu_of_init);
+IOMMU_OF_DECLARE(arm_smmuv2, "arm,smmu-v2", arm_smmu_of_init);
+IOMMU_OF_DECLARE(arm_mmu400, "arm,mmu-400", arm_smmu_of_init);
+IOMMU_OF_DECLARE(arm_mmu401, "arm,mmu-401", arm_smmu_of_init);
+IOMMU_OF_DECLARE(arm_mmu500, "arm,mmu-500", arm_smmu_of_init);
+IOMMU_OF_DECLARE(cavium_smmuv2, "cavium,smmu-v2", arm_smmu_of_init);
+
 MODULE_DESCRIPTION("IOMMU API for ARM architected SMMU implementations");
 MODULE_AUTHOR("Will Deacon <will.deacon@arm.com>");
 MODULE_LICENSE("GPL v2");