diff options
Diffstat (limited to 'drivers/irqchip/irq-gic-v3-its.c')
| -rw-r--r-- | drivers/irqchip/irq-gic-v3-its.c | 1065 |
1 files changed, 906 insertions, 159 deletions
diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c index 83b1186ffcad..973ede0197e3 100644 --- a/drivers/irqchip/irq-gic-v3-its.c +++ b/drivers/irqchip/irq-gic-v3-its.c @@ -11,9 +11,10 @@ #include <linux/cpu.h> #include <linux/crash_dump.h> #include <linux/delay.h> -#include <linux/dma-iommu.h> #include <linux/efi.h> #include <linux/interrupt.h> +#include <linux/iommu.h> +#include <linux/iopoll.h> #include <linux/irqdomain.h> #include <linux/list.h> #include <linux/log2.h> @@ -41,11 +42,14 @@ #define ITS_FLAGS_CMDQ_NEEDS_FLUSHING (1ULL << 0) #define ITS_FLAGS_WORKAROUND_CAVIUM_22375 (1ULL << 1) #define ITS_FLAGS_WORKAROUND_CAVIUM_23144 (1ULL << 2) -#define ITS_FLAGS_SAVE_SUSPEND_STATE (1ULL << 3) #define RDIST_FLAGS_PROPBASE_NEEDS_FLUSHING (1 << 0) #define RDIST_FLAGS_RD_TABLES_PREALLOCATED (1 << 1) +#define RD_LOCAL_LPI_ENABLED BIT(0) +#define RD_LOCAL_PENDTABLE_PREALLOCATED BIT(1) +#define RD_LOCAL_MEMRESERVE_DONE BIT(2) + static u32 lpi_id_bits; /* @@ -96,6 +100,7 @@ struct its_node { struct mutex dev_alloc_lock; struct list_head entry; void __iomem *base; + void __iomem *sgir_base; phys_addr_t phys_base; struct its_cmd_block *cmd_base; struct its_cmd_block *cmd_write; @@ -172,6 +177,13 @@ static struct { int next_victim; } vpe_proxy; +struct cpu_lpi_count { + atomic_t managed; + atomic_t unmanaged; +}; + +static DEFINE_PER_CPU(struct cpu_lpi_count, cpu_lpi_count); + static LIST_HEAD(its_nodes); static DEFINE_RAW_SPINLOCK(its_lock); static struct rdists *gic_rdists; @@ -188,6 +200,15 @@ static DEFINE_IDA(its_vpeid_ida); #define gic_data_rdist_rd_base() (gic_data_rdist()->rd_base) #define gic_data_rdist_vlpi_base() (gic_data_rdist_rd_base() + SZ_128K) +/* + * Skip ITSs that have no vLPIs mapped, unless we're on GICv4.1, as we + * always have vSGIs mapped. + */ +static bool require_its_list_vmovp(struct its_vm *vm, struct its_node *its) +{ + return (gic_rdists->has_rvpeid || vm->vlpi_count[its->list_nr]); +} + static u16 get_its_list(struct its_vm *vm) { struct its_node *its; @@ -197,7 +218,7 @@ static u16 get_its_list(struct its_vm *vm) if (!is_v4(its)) continue; - if (vm->vlpi_count[its->list_nr]) + if (require_its_list_vmovp(vm, its)) __set_bit(its->list_nr, &its_list); } @@ -239,15 +260,41 @@ static struct its_vlpi_map *get_vlpi_map(struct irq_data *d) return NULL; } -static int irq_to_cpuid(struct irq_data *d) +static int vpe_to_cpuid_lock(struct its_vpe *vpe, unsigned long *flags) +{ + raw_spin_lock_irqsave(&vpe->vpe_lock, *flags); + return vpe->col_idx; +} + +static void vpe_to_cpuid_unlock(struct its_vpe *vpe, unsigned long flags) +{ + raw_spin_unlock_irqrestore(&vpe->vpe_lock, flags); +} + +static int irq_to_cpuid_lock(struct irq_data *d, unsigned long *flags) { - struct its_device *its_dev = irq_data_get_irq_chip_data(d); struct its_vlpi_map *map = get_vlpi_map(d); + int cpu; - if (map) - return map->vpe->col_idx; + if (map) { + cpu = vpe_to_cpuid_lock(map->vpe, flags); + } else { + /* Physical LPIs are already locked via the irq_desc lock */ + struct its_device *its_dev = irq_data_get_irq_chip_data(d); + cpu = its_dev->event_map.col_map[its_get_event_id(d)]; + /* Keep GCC quiet... */ + *flags = 0; + } - return its_dev->event_map.col_map[its_get_event_id(d)]; + return cpu; +} + +static void irq_to_cpuid_unlock(struct irq_data *d, unsigned long flags) +{ + struct its_vlpi_map *map = get_vlpi_map(d); + + if (map) + vpe_to_cpuid_unlock(map->vpe, flags); } static struct its_collection *valid_col(struct its_collection *col) @@ -353,6 +400,15 @@ struct its_cmd_desc { struct { struct its_vpe *vpe; } its_invdb_cmd; + + struct { + struct its_vpe *vpe; + u8 sgi; + u8 priority; + bool enable; + bool group; + bool clear; + } its_vsgi_cmd; }; }; @@ -501,6 +557,31 @@ static void its_encode_db(struct its_cmd_block *cmd, bool db) its_mask_encode(&cmd->raw_cmd[2], db, 63, 63); } +static void its_encode_sgi_intid(struct its_cmd_block *cmd, u8 sgi) +{ + its_mask_encode(&cmd->raw_cmd[0], sgi, 35, 32); +} + +static void its_encode_sgi_priority(struct its_cmd_block *cmd, u8 prio) +{ + its_mask_encode(&cmd->raw_cmd[0], prio >> 4, 23, 20); +} + +static void its_encode_sgi_group(struct its_cmd_block *cmd, bool grp) +{ + its_mask_encode(&cmd->raw_cmd[0], grp, 10, 10); +} + +static void its_encode_sgi_clear(struct its_cmd_block *cmd, bool clr) +{ + its_mask_encode(&cmd->raw_cmd[0], clr, 9, 9); +} + +static void its_encode_sgi_enable(struct its_cmd_block *cmd, bool en) +{ + its_mask_encode(&cmd->raw_cmd[0], en, 8, 8); +} + static inline void its_fixup_cmd(struct its_cmd_block *cmd) { /* Let's fixup BE commands */ @@ -665,7 +746,7 @@ static struct its_collection *its_build_invall_cmd(struct its_node *its, its_fixup_cmd(cmd); - return NULL; + return desc->its_invall_cmd.col; } static struct its_vpe *its_build_vinvall_cmd(struct its_node *its, @@ -717,8 +798,13 @@ static struct its_vpe *its_build_vmapp_cmd(struct its_node *its, its_encode_alloc(cmd, alloc); - /* We can only signal PTZ when alloc==1. Why do we have two bits? */ - its_encode_ptz(cmd, alloc); + /* + * GICv4.1 provides a way to get the VLPI state, which needs the vPE + * to be unmapped first, and in this case, we may remap the vPE + * back while the VPT is not empty. So we can't assume that the + * VPT is empty on map. This is why we never advertise PTZ. + */ + its_encode_ptz(cmd, false); its_encode_vconf_addr(cmd, vconf_addr); its_encode_vmapp_default_db(cmd, desc->its_vmapp_cmd.vpe->vpe_db_lpi); @@ -866,6 +952,26 @@ static struct its_vpe *its_build_invdb_cmd(struct its_node *its, return valid_vpe(its, desc->its_invdb_cmd.vpe); } +static struct its_vpe *its_build_vsgi_cmd(struct its_node *its, + struct its_cmd_block *cmd, + struct its_cmd_desc *desc) +{ + if (WARN_ON(!is_v4_1(its))) + return NULL; + + its_encode_cmd(cmd, GITS_CMD_VSGI); + its_encode_vpeid(cmd, desc->its_vsgi_cmd.vpe->vpe_id); + its_encode_sgi_intid(cmd, desc->its_vsgi_cmd.sgi); + its_encode_sgi_priority(cmd, desc->its_vsgi_cmd.priority); + its_encode_sgi_group(cmd, desc->its_vsgi_cmd.group); + its_encode_sgi_clear(cmd, desc->its_vsgi_cmd.clear); + its_encode_sgi_enable(cmd, desc->its_vsgi_cmd.enable); + + its_fixup_cmd(cmd); + + return valid_vpe(its, desc->its_vsgi_cmd.vpe); +} + static u64 its_cmd_ptr_to_offset(struct its_node *its, struct its_cmd_block *ptr) { @@ -1214,7 +1320,7 @@ static void its_send_vmovp(struct its_vpe *vpe) if (!is_v4(its)) continue; - if (!vpe->its_vm->vlpi_count[its->list_nr]) + if (!require_its_list_vmovp(vpe->its_vm, its)) continue; desc.its_vmovp_cmd.col = &its->collections[col_id]; @@ -1321,7 +1427,7 @@ static void lpi_write_config(struct irq_data *d, u8 clr, u8 set) static void wait_for_syncr(void __iomem *rdbase) { - while (gic_read_lpir(rdbase + GICR_SYNCR) & 1) + while (readl_relaxed(rdbase + GICR_SYNCR) & 1) cpu_relax(); } @@ -1329,7 +1435,9 @@ static void direct_lpi_inv(struct irq_data *d) { struct its_vlpi_map *map = get_vlpi_map(d); void __iomem *rdbase; + unsigned long flags; u64 val; + int cpu; if (map) { struct its_device *its_dev = irq_data_get_irq_chip_data(d); @@ -1344,10 +1452,14 @@ static void direct_lpi_inv(struct irq_data *d) } /* Target the redistributor this LPI is currently routed to */ - rdbase = per_cpu_ptr(gic_rdists->rdist, irq_to_cpuid(d))->rd_base; + cpu = irq_to_cpuid_lock(d, &flags); + raw_spin_lock(&gic_data_rdist_cpu(cpu)->rd_lock); + rdbase = per_cpu_ptr(gic_rdists->rdist, cpu)->rd_base; gic_write_lpir(val, rdbase + GICR_INVLPIR); wait_for_syncr(rdbase); + raw_spin_unlock(&gic_data_rdist_cpu(cpu)->rd_lock); + irq_to_cpuid_unlock(d, flags); } static void lpi_update_config(struct irq_data *d, u8 clr, u8 set) @@ -1389,7 +1501,7 @@ static void its_vlpi_set_doorbell(struct irq_data *d, bool enable) * * Ideally, we'd issue a VMAPTI to set the doorbell to its LPI * value or to 1023, depending on the enable bit. But that - * would be issueing a mapping for an /existing/ DevID+EventID + * would be issuing a mapping for an /existing/ DevID+EventID * pair, which is UNPREDICTABLE. Instead, let's issue a VMOVI * to the /same/ vPE, using this opportunity to adjust the * doorbell. Mouahahahaha. We loves it, Precious. @@ -1413,42 +1525,161 @@ static void its_unmask_irq(struct irq_data *d) lpi_update_config(d, 0, LPI_PROP_ENABLED); } +static __maybe_unused u32 its_read_lpi_count(struct irq_data *d, int cpu) +{ + if (irqd_affinity_is_managed(d)) + return atomic_read(&per_cpu_ptr(&cpu_lpi_count, cpu)->managed); + + return atomic_read(&per_cpu_ptr(&cpu_lpi_count, cpu)->unmanaged); +} + +static void its_inc_lpi_count(struct irq_data *d, int cpu) +{ + if (irqd_affinity_is_managed(d)) + atomic_inc(&per_cpu_ptr(&cpu_lpi_count, cpu)->managed); + else + atomic_inc(&per_cpu_ptr(&cpu_lpi_count, cpu)->unmanaged); +} + +static void its_dec_lpi_count(struct irq_data *d, int cpu) +{ + if (irqd_affinity_is_managed(d)) + atomic_dec(&per_cpu_ptr(&cpu_lpi_count, cpu)->managed); + else + atomic_dec(&per_cpu_ptr(&cpu_lpi_count, cpu)->unmanaged); +} + +static unsigned int cpumask_pick_least_loaded(struct irq_data *d, + const struct cpumask *cpu_mask) +{ + unsigned int cpu = nr_cpu_ids, tmp; + int count = S32_MAX; + + for_each_cpu(tmp, cpu_mask) { + int this_count = its_read_lpi_count(d, tmp); + if (this_count < count) { + cpu = tmp; + count = this_count; + } + } + + return cpu; +} + +/* + * As suggested by Thomas Gleixner in: + * https://lore.kernel.org/r/87h80q2aoc.fsf@nanos.tec.linutronix.de + */ +static int its_select_cpu(struct irq_data *d, + const struct cpumask *aff_mask) +{ + struct its_device *its_dev = irq_data_get_irq_chip_data(d); + static DEFINE_RAW_SPINLOCK(tmpmask_lock); + static struct cpumask __tmpmask; + struct cpumask *tmpmask; + unsigned long flags; + int cpu, node; + node = its_dev->its->numa_node; + tmpmask = &__tmpmask; + + raw_spin_lock_irqsave(&tmpmask_lock, flags); + + if (!irqd_affinity_is_managed(d)) { + /* First try the NUMA node */ + if (node != NUMA_NO_NODE) { + /* + * Try the intersection of the affinity mask and the + * node mask (and the online mask, just to be safe). + */ + cpumask_and(tmpmask, cpumask_of_node(node), aff_mask); + cpumask_and(tmpmask, tmpmask, cpu_online_mask); + + /* + * Ideally, we would check if the mask is empty, and + * try again on the full node here. + * + * But it turns out that the way ACPI describes the + * affinity for ITSs only deals about memory, and + * not target CPUs, so it cannot describe a single + * ITS placed next to two NUMA nodes. + * + * Instead, just fallback on the online mask. This + * diverges from Thomas' suggestion above. + */ + cpu = cpumask_pick_least_loaded(d, tmpmask); + if (cpu < nr_cpu_ids) + goto out; + + /* If we can't cross sockets, give up */ + if ((its_dev->its->flags & ITS_FLAGS_WORKAROUND_CAVIUM_23144)) + goto out; + + /* If the above failed, expand the search */ + } + + /* Try the intersection of the affinity and online masks */ + cpumask_and(tmpmask, aff_mask, cpu_online_mask); + + /* If that doesn't fly, the online mask is the last resort */ + if (cpumask_empty(tmpmask)) + cpumask_copy(tmpmask, cpu_online_mask); + + cpu = cpumask_pick_least_loaded(d, tmpmask); + } else { + cpumask_copy(tmpmask, aff_mask); + + /* If we cannot cross sockets, limit the search to that node */ + if ((its_dev->its->flags & ITS_FLAGS_WORKAROUND_CAVIUM_23144) && + node != NUMA_NO_NODE) + cpumask_and(tmpmask, tmpmask, cpumask_of_node(node)); + + cpu = cpumask_pick_least_loaded(d, tmpmask); + } +out: + raw_spin_unlock_irqrestore(&tmpmask_lock, flags); + + pr_debug("IRQ%d -> %*pbl CPU%d\n", d->irq, cpumask_pr_args(aff_mask), cpu); + return cpu; +} + static int its_set_affinity(struct irq_data *d, const struct cpumask *mask_val, bool force) { - unsigned int cpu; - const struct cpumask *cpu_mask = cpu_online_mask; struct its_device *its_dev = irq_data_get_irq_chip_data(d); struct its_collection *target_col; u32 id = its_get_event_id(d); + int cpu, prev_cpu; /* A forwarded interrupt should use irq_set_vcpu_affinity */ if (irqd_is_forwarded_to_vcpu(d)) return -EINVAL; - /* lpi cannot be routed to a redistributor that is on a foreign node */ - if (its_dev->its->flags & ITS_FLAGS_WORKAROUND_CAVIUM_23144) { - if (its_dev->its->numa_node >= 0) { - cpu_mask = cpumask_of_node(its_dev->its->numa_node); - if (!cpumask_intersects(mask_val, cpu_mask)) - return -EINVAL; - } - } + prev_cpu = its_dev->event_map.col_map[id]; + its_dec_lpi_count(d, prev_cpu); - cpu = cpumask_any_and(mask_val, cpu_mask); + if (!force) + cpu = its_select_cpu(d, mask_val); + else + cpu = cpumask_pick_least_loaded(d, mask_val); - if (cpu >= nr_cpu_ids) - return -EINVAL; + if (cpu < 0 || cpu >= nr_cpu_ids) + goto err; /* don't set the affinity when the target cpu is same as current one */ - if (cpu != its_dev->event_map.col_map[id]) { + if (cpu != prev_cpu) { target_col = &its_dev->its->collections[cpu]; its_send_movi(its_dev, target_col, id); its_dev->event_map.col_map[id] = cpu; irq_data_update_effective_affinity(d, cpumask_of(cpu)); } + its_inc_lpi_count(d, cpu); + return IRQ_SET_MASK_OK_DONE; + +err: + its_inc_lpi_count(d, prev_cpu); + return -EINVAL; } static u64 its_irq_get_msi_base(struct its_device *its_dev) @@ -1499,12 +1730,36 @@ static int its_irq_set_irqchip_state(struct irq_data *d, return 0; } +static int its_irq_retrigger(struct irq_data *d) +{ + return !its_irq_set_irqchip_state(d, IRQCHIP_STATE_PENDING, true); +} + +/* + * Two favourable cases: + * + * (a) Either we have a GICv4.1, and all vPEs have to be mapped at all times + * for vSGI delivery + * + * (b) Or the ITSs do not use a list map, meaning that VMOVP is cheap enough + * and we're better off mapping all VPEs always + * + * If neither (a) nor (b) is true, then we map vPEs on demand. + * + */ +static bool gic_requires_eager_mapping(void) +{ + if (!its_list_map || gic_rdists->has_rvpeid) + return true; + + return false; +} + static void its_map_vm(struct its_node *its, struct its_vm *vm) { unsigned long flags; - /* Not using the ITS list? Everything is always mapped. */ - if (!its_list_map) + if (gic_requires_eager_mapping()) return; raw_spin_lock_irqsave(&vmovp_lock, flags); @@ -1538,7 +1793,7 @@ static void its_unmap_vm(struct its_node *its, struct its_vm *vm) unsigned long flags; /* Not using the ITS list? Everything is always mapped. */ - if (!its_list_map) + if (gic_requires_eager_mapping()) return; raw_spin_lock_irqsave(&vmovp_lock, flags); @@ -1731,6 +1986,7 @@ static struct irq_chip its_irq_chip = { .irq_set_affinity = its_set_affinity, .irq_compose_msi_msg = its_irq_compose_msi_msg, .irq_set_irqchip_state = its_irq_set_irqchip_state, + .irq_retrigger = its_irq_retrigger, .irq_set_vcpu_affinity = its_irq_set_vcpu_affinity, }; @@ -1890,7 +2146,7 @@ static unsigned long *its_lpi_alloc(int nr_irqs, u32 *base, int *nr_ids) if (err) goto out; - bitmap = kcalloc(BITS_TO_LONGS(nr_irqs), sizeof (long), GFP_ATOMIC); + bitmap = bitmap_zalloc(nr_irqs, GFP_ATOMIC); if (!bitmap) goto out; @@ -1906,7 +2162,7 @@ out: static void its_lpi_free(unsigned long *bitmap, u32 base, u32 nr_ids) { WARN_ON(free_lpi_range(base, nr_ids)); - kfree(bitmap); + bitmap_free(bitmap); } static void gic_reset_prop_table(void *va) @@ -1952,7 +2208,7 @@ static bool gic_check_reserved_range(phys_addr_t addr, unsigned long size) addr_end = addr + size - 1; - for_each_reserved_mem_region(i, &start, &end) { + for_each_reserved_mem_range(i, &start, &end) { if (addr >= start && addr_end <= end) return true; } @@ -2036,18 +2292,17 @@ static void its_write_baser(struct its_node *its, struct its_baser *baser, } static int its_setup_baser(struct its_node *its, struct its_baser *baser, - u64 cache, u64 shr, u32 psz, u32 order, - bool indirect) + u64 cache, u64 shr, u32 order, bool indirect) { u64 val = its_read_baser(its, baser); u64 esz = GITS_BASER_ENTRY_SIZE(val); u64 type = GITS_BASER_TYPE(val); u64 baser_phys, tmp; - u32 alloc_pages; + u32 alloc_pages, psz; struct page *page; void *base; -retry_alloc_baser: + psz = baser->psz; alloc_pages = (PAGE_ORDER_TO_SIZE(order) / psz); if (alloc_pages > GITS_BASER_PAGES_MAX) { pr_warn("ITS@%pa: %s too large, reduce ITS pages %u->%u\n", @@ -2120,25 +2375,6 @@ retry_baser: goto retry_baser; } - if ((val ^ tmp) & GITS_BASER_PAGE_SIZE_MASK) { - /* - * Page size didn't stick. Let's try a smaller - * size and retry. If we reach 4K, then - * something is horribly wrong... - */ - free_pages((unsigned long)base, order); - baser->base = NULL; - - switch (psz) { - case SZ_16K: - psz = SZ_4K; - goto retry_alloc_baser; - case SZ_64K: - psz = SZ_16K; - goto retry_alloc_baser; - } - } - if (val != tmp) { pr_err("ITS@%pa: %s doesn't stick: %llx %llx\n", &its->phys_base, its_base_type_string[type], @@ -2164,13 +2400,14 @@ retry_baser: static bool its_parse_indirect_baser(struct its_node *its, struct its_baser *baser, - u32 psz, u32 *order, u32 ids) + u32 *order, u32 ids) { u64 tmp = its_read_baser(its, baser); u64 type = GITS_BASER_TYPE(tmp); u64 esz = GITS_BASER_ENTRY_SIZE(tmp); u64 val = GITS_BASER_InnerShareable | GITS_BASER_RaWaWb; u32 new_order = *order; + u32 psz = baser->psz; bool indirect = false; /* No need to enable Indirection if memory requirement < (psz*2)bytes */ @@ -2288,11 +2525,58 @@ static void its_free_tables(struct its_node *its) } } +static int its_probe_baser_psz(struct its_node *its, struct its_baser *baser) +{ + u64 psz = SZ_64K; + + while (psz) { + u64 val, gpsz; + + val = its_read_baser(its, baser); + val &= ~GITS_BASER_PAGE_SIZE_MASK; + + switch (psz) { + case SZ_64K: + gpsz = GITS_BASER_PAGE_SIZE_64K; + break; + case SZ_16K: + gpsz = GITS_BASER_PAGE_SIZE_16K; + break; + case SZ_4K: + default: + gpsz = GITS_BASER_PAGE_SIZE_4K; + break; + } + + gpsz >>= GITS_BASER_PAGE_SIZE_SHIFT; + + val |= FIELD_PREP(GITS_BASER_PAGE_SIZE_MASK, gpsz); + its_write_baser(its, baser, val); + + if (FIELD_GET(GITS_BASER_PAGE_SIZE_MASK, baser->val) == gpsz) + break; + + switch (psz) { + case SZ_64K: + psz = SZ_16K; + break; + case SZ_16K: + psz = SZ_4K; + break; + case SZ_4K: + default: + return -1; + } + } + + baser->psz = psz; + return 0; +} + static int its_alloc_tables(struct its_node *its) { u64 shr = GITS_BASER_InnerShareable; u64 cache = GITS_BASER_RaWaWb; - u32 psz = SZ_64K; int err, i; if (its->flags & ITS_FLAGS_WORKAROUND_CAVIUM_22375) @@ -2303,16 +2587,22 @@ static int its_alloc_tables(struct its_node *its) struct its_baser *baser = its->tables + i; u64 val = its_read_baser(its, baser); u64 type = GITS_BASER_TYPE(val); - u32 order = get_order(psz); bool indirect = false; + u32 order; - switch (type) { - case GITS_BASER_TYPE_NONE: + if (type == GITS_BASER_TYPE_NONE) continue; + if (its_probe_baser_psz(its, baser)) { + its_free_tables(its); + return -ENXIO; + } + + order = get_order(baser->psz); + + switch (type) { case GITS_BASER_TYPE_DEVICE: - indirect = its_parse_indirect_baser(its, baser, - psz, &order, + indirect = its_parse_indirect_baser(its, baser, &order, device_ids(its)); break; @@ -2328,20 +2618,18 @@ static int its_alloc_tables(struct its_node *its) } } - indirect = its_parse_indirect_baser(its, baser, - psz, &order, + indirect = its_parse_indirect_baser(its, baser, &order, ITS_MAX_VPEID_BITS); break; } - err = its_setup_baser(its, baser, cache, shr, psz, order, indirect); + err = its_setup_baser(its, baser, cache, shr, order, indirect); if (err < 0) { its_free_tables(its); return err; } /* Update settings which will be used for next BASERn */ - psz = baser->psz; cache = baser->val & GITS_BASER_CACHEABILITY_MASK; shr = baser->val & GITS_BASER_SHAREABILITY_MASK; } @@ -2452,6 +2740,10 @@ static bool allocate_vpe_l2_table(int cpu, u32 id) if (!gic_rdists->has_rvpeid) return true; + /* Skip non-present CPUs */ + if (!base) + return true; + val = gicr_read_vpropbaser(base + SZ_128K + GICR_VPROPBASER); esz = FIELD_GET(GICR_VPROPBASER_4_1_ENTRY_SIZE, val) + 1; @@ -2461,7 +2753,7 @@ static bool allocate_vpe_l2_table(int cpu, u32 id) switch (gpsz) { default: WARN_ON(1); - /* fall through */ + fallthrough; case GIC_PAGE_SIZE_4K: psz = SZ_4K; break; @@ -2538,7 +2830,7 @@ static int allocate_vpe_l1_table(void) if (val & GICR_VPROPBASER_4_1_VALID) goto out; - gic_data_rdist()->vpe_table_mask = kzalloc(sizeof(cpumask_t), GFP_KERNEL); + gic_data_rdist()->vpe_table_mask = kzalloc(sizeof(cpumask_t), GFP_ATOMIC); if (!gic_data_rdist()->vpe_table_mask) return -ENOMEM; @@ -2556,7 +2848,7 @@ static int allocate_vpe_l1_table(void) switch (gpsz) { default: gpsz = GIC_PAGE_SIZE_4K; - /* fall through */ + fallthrough; case GIC_PAGE_SIZE_4K: psz = SZ_4K; break; @@ -2605,7 +2897,7 @@ static int allocate_vpe_l1_table(void) pr_debug("np = %d, npg = %lld, psz = %d, epp = %d, esz = %d\n", np, npg, psz, epp, esz); - page = alloc_pages(GFP_KERNEL | __GFP_ZERO, get_order(np * PAGE_SIZE)); + page = alloc_pages(GFP_ATOMIC | __GFP_ZERO, get_order(np * PAGE_SIZE)); if (!page) return -ENOMEM; @@ -2721,18 +3013,12 @@ static int __init allocate_lpi_tables(void) return 0; } -static u64 its_clear_vpend_valid(void __iomem *vlpi_base, u64 clr, u64 set) +static u64 read_vpend_dirty_clear(void __iomem *vlpi_base) { u32 count = 1000000; /* 1s! */ bool clean; u64 val; - val = gicr_read_vpendbaser(vlpi_base + GICR_VPENDBASER); - val &= ~GICR_VPENDBASER_Valid; - val &= ~clr; - val |= set; - gicr_write_vpendbaser(val, vlpi_base + GICR_VPENDBASER); - do { val = gicr_read_vpendbaser(vlpi_base + GICR_VPENDBASER); clean = !(val & GICR_VPENDBASER_Dirty); @@ -2743,10 +3029,26 @@ static u64 its_clear_vpend_valid(void __iomem *vlpi_base, u64 clr, u64 set) } } while (!clean && count); - if (unlikely(val & GICR_VPENDBASER_Dirty)) { + if (unlikely(!clean)) pr_err_ratelimited("ITS virtual pending table not cleaning\n"); + + return val; +} + +static u64 its_clear_vpend_valid(void __iomem *vlpi_base, u64 clr, u64 set) +{ + u64 val; + + /* Make sure we wait until the RD is done with the initial scan */ + val = read_vpend_dirty_clear(vlpi_base); + val &= ~GICR_VPENDBASER_Valid; + val &= ~clr; + val |= set; + gicr_write_vpendbaser(val, vlpi_base + GICR_VPENDBASER); + + val = read_vpend_dirty_clear(vlpi_base); + if (unlikely(val & GICR_VPENDBASER_Dirty)) val |= GICR_VPENDBASER_PendingLast; - } return val; } @@ -2758,7 +3060,7 @@ static void its_cpu_init_lpis(void) phys_addr_t paddr; u64 val, tmp; - if (gic_data_rdist()->lpi_enabled) + if (gic_data_rdist()->flags & RD_LOCAL_LPI_ENABLED) return; val = readl_relaxed(rbase + GICR_CTLR); @@ -2777,15 +3079,13 @@ static void its_cpu_init_lpis(void) paddr &= GENMASK_ULL(51, 16); WARN_ON(!gic_check_reserved_range(paddr, LPI_PENDBASE_SZ)); - its_free_pending_table(gic_data_rdist()->pend_page); - gic_data_rdist()->pend_page = NULL; + gic_data_rdist()->flags |= RD_LOCAL_PENDTABLE_PREALLOCATED; goto out; } pend_page = gic_data_rdist()->pend_page; paddr = page_to_phys(pend_page); - WARN_ON(gic_reserve_range(paddr, LPI_PENDBASE_SZ)); /* set PROPBASE */ val = (gic_rdists->prop_table_pa | @@ -2841,7 +3141,7 @@ static void its_cpu_init_lpis(void) /* * It's possible for CPU to receive VLPIs before it is - * sheduled as a vPE, especially for the first CPU, and the + * scheduled as a vPE, especially for the first CPU, and the * VLPI with INTID larger than 2^(IDbits+1) will be considered * as out of range and dropped by GIC. * So we initialize IDbits to known value to avoid VLPI drop. @@ -2872,10 +3172,11 @@ static void its_cpu_init_lpis(void) /* Make sure the GIC has seen the above */ dsb(sy); out: - gic_data_rdist()->lpi_enabled = true; + gic_data_rdist()->flags |= RD_LOCAL_LPI_ENABLED; pr_info("GICv3: CPU%d: using %s LPI pending table @%pa\n", smp_processor_id(), - gic_data_rdist()->pend_page ? "allocated" : "reserved", + gic_data_rdist()->flags & RD_LOCAL_PENDTABLE_PREALLOCATED ? + "reserved" : "allocated", &paddr); } @@ -3101,7 +3402,7 @@ static struct its_device *its_create_device(struct its_node *its, u32 dev_id, if (!dev || !itt || !col_map || (!lpi_map && alloc_lpis)) { kfree(dev); kfree(itt); - kfree(lpi_map); + bitmap_free(lpi_map); kfree(col_map); return NULL; } @@ -3206,6 +3507,9 @@ static int its_msi_prepare(struct irq_domain *domain, struct device *dev, goto out; } + if (info->flags & MSI_ALLOC_FLAGS_PROXY_DEVICE) + its_dev->shared = true; + pr_debug("ITT %d entries, %d bits\n", nvec, ilog2(nvec)); out: mutex_unlock(&its->dev_alloc_lock); @@ -3247,6 +3551,7 @@ static int its_irq_domain_alloc(struct irq_domain *domain, unsigned int virq, msi_alloc_info_t *info = args; struct its_device *its_dev = info->scratchpad[0].ptr; struct its_node *its = its_dev->its; + struct irq_data *irqd; irq_hw_number_t hwirq; int err; int i; @@ -3266,7 +3571,9 @@ static int its_irq_domain_alloc(struct irq_domain *domain, unsigned int virq, irq_domain_set_hwirq_and_chip(domain, virq + i, hwirq + i, &its_irq_chip, its_dev); - irqd_set_single_target(irq_desc_get_irq_data(irq_to_desc(virq + i))); + irqd = irq_get_irq_data(virq + i); + irqd_set_single_target(irqd); + irqd_set_affinity_on_activate(irqd); pr_debug("ID:%d pID:%d vID:%d\n", (int)(hwirq + i - its_dev->event_map.lpi_base), (int)(hwirq + i), virq + i); @@ -3280,22 +3587,13 @@ static int its_irq_domain_activate(struct irq_domain *domain, { struct its_device *its_dev = irq_data_get_irq_chip_data(d); u32 event = its_get_event_id(d); - const struct cpumask *cpu_mask = cpu_online_mask; int cpu; - /* get the cpu_mask of local node */ - if (its_dev->its->numa_node >= 0) - cpu_mask = cpumask_of_node(its_dev->its->numa_node); - - /* Bind the LPI to the first possible CPU */ - cpu = cpumask_first_and(cpu_mask, cpu_online_mask); - if (cpu >= nr_cpu_ids) { - if (its_dev->its->flags & ITS_FLAGS_WORKAROUND_CAVIUM_23144) - return -EINVAL; - - cpu = cpumask_first(cpu_online_mask); - } + cpu = its_select_cpu(d, cpu_online_mask); + if (cpu < 0 || cpu >= nr_cpu_ids) + return -EINVAL; + its_inc_lpi_count(d, cpu); its_dev->event_map.col_map[event] = cpu; irq_data_update_effective_affinity(d, cpumask_of(cpu)); @@ -3310,6 +3608,7 @@ static void its_irq_domain_deactivate(struct irq_domain *domain, struct its_device *its_dev = irq_data_get_irq_chip_data(d); u32 event = its_get_event_id(d); + its_dec_lpi_count(d, its_dev->event_map.col_map[event]); /* Stop the delivery of interrupts */ its_send_discard(its_dev, event); } @@ -3337,7 +3636,7 @@ static void its_irq_domain_free(struct irq_domain *domain, unsigned int virq, /* * If all interrupts have been freed, start mopping the - * floor. This is conditionned on the device not being shared. + * floor. This is conditioned on the device not being shared. */ if (!its_dev->shared && bitmap_empty(its_dev->event_map.lpi_map, @@ -3482,17 +3781,25 @@ static int its_vpe_set_affinity(struct irq_data *d, { struct its_vpe *vpe = irq_data_get_irq_chip_data(d); int from, cpu = cpumask_first(mask_val); + unsigned long flags; /* * Changing affinity is mega expensive, so let's be as lazy as * we can and only do it if we really have to. Also, if mapped * into the proxy device, we need to move the doorbell * interrupt to its new location. + * + * Another thing is that changing the affinity of a vPE affects + * *other interrupts* such as all the vLPIs that are routed to + * this vPE. This means that the irq_desc lock is not enough to + * protect us, and that we must ensure nobody samples vpe->col_idx + * during the update, hence the lock below which must also be + * taken on any vLPI handling path that evaluates vpe->col_idx. */ - if (vpe->col_idx == cpu) + from = vpe_to_cpuid_lock(vpe, &flags); + if (from == cpu) goto out; - from = vpe->col_idx; vpe->col_idx = cpu; /* @@ -3508,10 +3815,25 @@ static int its_vpe_set_affinity(struct irq_data *d, out: irq_data_update_effective_affinity(d, cpumask_of(cpu)); + vpe_to_cpuid_unlock(vpe, flags); return IRQ_SET_MASK_OK_DONE; } +static void its_wait_vpt_parse_complete(void) +{ + void __iomem *vlpi_base = gic_data_rdist_vlpi_base(); + u64 val; + + if (!gic_rdists->has_vpend_valid_dirty) + return; + + WARN_ON_ONCE(readq_relaxed_poll_timeout_atomic(vlpi_base + GICR_VPENDBASER, + val, + !(val & GICR_VPENDBASER_Dirty), + 1, 500)); +} + static void its_vpe_schedule(struct its_vpe *vpe) { void __iomem *vlpi_base = gic_data_rdist_vlpi_base(); @@ -3528,7 +3850,7 @@ static void its_vpe_schedule(struct its_vpe *vpe) val = virt_to_phys(page_address(vpe->vpt_page)) & GENMASK_ULL(51, 16); val |= GICR_VPENDBASER_RaWaWb; - val |= GICR_VPENDBASER_NonShareable; + val |= GICR_VPENDBASER_InnerShareable; /* * There is no good way of finding out if the pending table is * empty as we can race against the doorbell interrupt very @@ -3589,6 +3911,10 @@ static int its_vpe_set_vcpu_affinity(struct irq_data *d, void *vcpu_info) its_vpe_deschedule(vpe); return 0; + case COMMIT_VPE: + its_wait_vpt_parse_complete(); + return 0; + case INVALL_VPE: its_vpe_invall(vpe); return 0; @@ -3619,9 +3945,11 @@ static void its_vpe_send_inv(struct irq_data *d) void __iomem *rdbase; /* Target the redistributor this VPE is currently known on */ + raw_spin_lock(&gic_data_rdist_cpu(vpe->col_idx)->rd_lock); rdbase = per_cpu_ptr(gic_rdists->rdist, vpe->col_idx)->rd_base; gic_write_lpir(d->parent_data->hwirq, rdbase + GICR_INVLPIR); wait_for_syncr(rdbase); + raw_spin_unlock(&gic_data_rdist_cpu(vpe->col_idx)->rd_lock); } else { its_vpe_send_cmd(vpe, its_send_inv); } @@ -3675,12 +4003,18 @@ static int its_vpe_set_irqchip_state(struct irq_data *d, return 0; } +static int its_vpe_retrigger(struct irq_data *d) +{ + return !its_vpe_set_irqchip_state(d, IRQCHIP_STATE_PENDING, true); +} + static struct irq_chip its_vpe_irq_chip = { .name = "GICv4-vpe", .irq_mask = its_vpe_mask_irq, .irq_unmask = its_vpe_unmask_irq, .irq_eoi = irq_chip_eoi_parent, .irq_set_affinity = its_vpe_set_affinity, + .irq_retrigger = its_vpe_retrigger, .irq_set_irqchip_state = its_vpe_set_irqchip_state, .irq_set_vcpu_affinity = its_vpe_set_vcpu_affinity, }; @@ -3751,16 +4085,24 @@ static void its_vpe_4_1_deschedule(struct its_vpe *vpe, u64 val; if (info->req_db) { + unsigned long flags; + /* * vPE is going to block: make the vPE non-resident with * PendingLast clear and DB set. The GIC guarantees that if * we read-back PendingLast clear, then a doorbell will be * delivered when an interrupt comes. + * + * Note the locking to deal with the concurrent update of + * pending_last from the doorbell interrupt handler that can + * run concurrently. */ + raw_spin_lock_irqsave(&vpe->vpe_lock, flags); val = its_clear_vpend_valid(vlpi_base, GICR_VPENDBASER_PendingLast, GICR_VPENDBASER_4_1_DB); vpe->pending_last = !!(val & GICR_VPENDBASER_PendingLast); + raw_spin_unlock_irqrestore(&vpe->vpe_lock, flags); } else { /* * We're not blocking, so just make the vPE non-resident @@ -3776,14 +4118,22 @@ static void its_vpe_4_1_deschedule(struct its_vpe *vpe, static void its_vpe_4_1_invall(struct its_vpe *vpe) { void __iomem *rdbase; + unsigned long flags; u64 val; + int cpu; val = GICR_INVALLR_V; val |= FIELD_PREP(GICR_INVALLR_VPEID, vpe->vpe_id); /* Target the redistributor this vPE is currently known on */ - rdbase = per_cpu_ptr(gic_rdists->rdist, vpe->col_idx)->rd_base; + cpu = vpe_to_cpuid_lock(vpe, &flags); + raw_spin_lock(&gic_data_rdist_cpu(cpu)->rd_lock); + rdbase = per_cpu_ptr(gic_rdists->rdist, cpu)->rd_base; gic_write_lpir(val, rdbase + GICR_INVALLR); + + wait_for_syncr(rdbase); + raw_spin_unlock(&gic_data_rdist_cpu(cpu)->rd_lock); + vpe_to_cpuid_unlock(vpe, flags); } static int its_vpe_4_1_set_vcpu_affinity(struct irq_data *d, void *vcpu_info) @@ -3800,6 +4150,10 @@ static int its_vpe_4_1_set_vcpu_affinity(struct irq_data *d, void *vcpu_info) its_vpe_4_1_deschedule(vpe, info); return 0; + case COMMIT_VPE: + its_wait_vpt_parse_complete(); + return 0; + case INVALL_VPE: its_vpe_4_1_invall(vpe); return 0; @@ -3818,6 +4172,222 @@ static struct irq_chip its_vpe_4_1_irq_chip = { .irq_set_vcpu_affinity = its_vpe_4_1_set_vcpu_affinity, }; +static void its_configure_sgi(struct irq_data *d, bool clear) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + struct its_cmd_desc desc; + + desc.its_vsgi_cmd.vpe = vpe; + desc.its_vsgi_cmd.sgi = d->hwirq; + desc.its_vsgi_cmd.priority = vpe->sgi_config[d->hwirq].priority; + desc.its_vsgi_cmd.enable = vpe->sgi_config[d->hwirq].enabled; + desc.its_vsgi_cmd.group = vpe->sgi_config[d->hwirq].group; + desc.its_vsgi_cmd.clear = clear; + + /* + * GICv4.1 allows us to send VSGI commands to any ITS as long as the + * destination VPE is mapped there. Since we map them eagerly at + * activation time, we're pretty sure the first GICv4.1 ITS will do. + */ + its_send_single_vcommand(find_4_1_its(), its_build_vsgi_cmd, &desc); +} + +static void its_sgi_mask_irq(struct irq_data *d) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + + vpe->sgi_config[d->hwirq].enabled = false; + its_configure_sgi(d, false); +} + +static void its_sgi_unmask_irq(struct irq_data *d) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + + vpe->sgi_config[d->hwirq].enabled = true; + its_configure_sgi(d, false); +} + +static int its_sgi_set_affinity(struct irq_data *d, + const struct cpumask *mask_val, + bool force) +{ + /* + * There is no notion of affinity for virtual SGIs, at least + * not on the host (since they can only be targeting a vPE). + * Tell the kernel we've done whatever it asked for. + */ + irq_data_update_effective_affinity(d, mask_val); + return IRQ_SET_MASK_OK; +} + +static int its_sgi_set_irqchip_state(struct irq_data *d, + enum irqchip_irq_state which, + bool state) +{ + if (which != IRQCHIP_STATE_PENDING) + return -EINVAL; + + if (state) { + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + struct its_node *its = find_4_1_its(); + u64 val; + + val = FIELD_PREP(GITS_SGIR_VPEID, vpe->vpe_id); + val |= FIELD_PREP(GITS_SGIR_VINTID, d->hwirq); + writeq_relaxed(val, its->sgir_base + GITS_SGIR - SZ_128K); + } else { + its_configure_sgi(d, true); + } + + return 0; +} + +static int its_sgi_get_irqchip_state(struct irq_data *d, + enum irqchip_irq_state which, bool *val) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + void __iomem *base; + unsigned long flags; + u32 count = 1000000; /* 1s! */ + u32 status; + int cpu; + + if (which != IRQCHIP_STATE_PENDING) + return -EINVAL; + + /* + * Locking galore! We can race against two different events: + * + * - Concurrent vPE affinity change: we must make sure it cannot + * happen, or we'll talk to the wrong redistributor. This is + * identical to what happens with vLPIs. + * + * - Concurrent VSGIPENDR access: As it involves accessing two + * MMIO registers, this must be made atomic one way or another. + */ + cpu = vpe_to_cpuid_lock(vpe, &flags); + raw_spin_lock(&gic_data_rdist_cpu(cpu)->rd_lock); + base = gic_data_rdist_cpu(cpu)->rd_base + SZ_128K; + writel_relaxed(vpe->vpe_id, base + GICR_VSGIR); + do { + status = readl_relaxed(base + GICR_VSGIPENDR); + if (!(status & GICR_VSGIPENDR_BUSY)) + goto out; + + count--; + if (!count) { + pr_err_ratelimited("Unable to get SGI status\n"); + goto out; + } + cpu_relax(); + udelay(1); + } while (count); + +out: + raw_spin_unlock(&gic_data_rdist_cpu(cpu)->rd_lock); + vpe_to_cpuid_unlock(vpe, flags); + + if (!count) + return -ENXIO; + + *val = !!(status & (1 << d->hwirq)); + + return 0; +} + +static int its_sgi_set_vcpu_affinity(struct irq_data *d, void *vcpu_info) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + struct its_cmd_info *info = vcpu_info; + + switch (info->cmd_type) { + case PROP_UPDATE_VSGI: + vpe->sgi_config[d->hwirq].priority = info->priority; + vpe->sgi_config[d->hwirq].group = info->group; + its_configure_sgi(d, false); + return 0; + + default: + return -EINVAL; + } +} + +static struct irq_chip its_sgi_irq_chip = { + .name = "GICv4.1-sgi", + .irq_mask = its_sgi_mask_irq, + .irq_unmask = its_sgi_unmask_irq, + .irq_set_affinity = its_sgi_set_affinity, + .irq_set_irqchip_state = its_sgi_set_irqchip_state, + .irq_get_irqchip_state = its_sgi_get_irqchip_state, + .irq_set_vcpu_affinity = its_sgi_set_vcpu_affinity, +}; + +static int its_sgi_irq_domain_alloc(struct irq_domain *domain, + unsigned int virq, unsigned int nr_irqs, + void *args) +{ + struct its_vpe *vpe = args; + int i; + + /* Yes, we do want 16 SGIs */ + WARN_ON(nr_irqs != 16); + + for (i = 0; i < 16; i++) { + vpe->sgi_config[i].priority = 0; + vpe->sgi_config[i].enabled = false; + vpe->sgi_config[i].group = false; + + irq_domain_set_hwirq_and_chip(domain, virq + i, i, + &its_sgi_irq_chip, vpe); + irq_set_status_flags(virq + i, IRQ_DISABLE_UNLAZY); + } + + return 0; +} + +static void its_sgi_irq_domain_free(struct irq_domain *domain, + unsigned int virq, + unsigned int nr_irqs) +{ + /* Nothing to do */ +} + +static int its_sgi_irq_domain_activate(struct irq_domain *domain, + struct irq_data *d, bool reserve) +{ + /* Write out the initial SGI configuration */ + its_configure_sgi(d, false); + return 0; +} + +static void its_sgi_irq_domain_deactivate(struct irq_domain *domain, + struct irq_data *d) +{ + struct its_vpe *vpe = irq_data_get_irq_chip_data(d); + + /* + * The VSGI command is awkward: + * + * - To change the configuration, CLEAR must be set to false, + * leaving the pending bit unchanged. + * - To clear the pending bit, CLEAR must be set to true, leaving + * the configuration unchanged. + * + * You just can't do both at once, hence the two commands below. + */ + vpe->sgi_config[d->hwirq].enabled = false; + its_configure_sgi(d, false); + its_configure_sgi(d, true); +} + +static const struct irq_domain_ops its_sgi_domain_ops = { + .alloc = its_sgi_irq_domain_alloc, + .free = its_sgi_irq_domain_free, + .activate = its_sgi_irq_domain_activate, + .deactivate = its_sgi_irq_domain_deactivate, +}; + static int its_vpe_id_alloc(void) { return ida_simple_get(&its_vpeid_ida, 0, ITS_MAX_VPEID, GFP_KERNEL); @@ -3851,6 +4421,7 @@ static int its_vpe_init(struct its_vpe *vpe) return -ENOMEM; } + raw_spin_lock_init(&vpe->vpe_lock); vpe->vpe_id = vpe_id; vpe->vpt_page = vpt_page; if (gic_rdists->has_rvpeid) @@ -3945,7 +4516,7 @@ static int its_vpe_irq_domain_alloc(struct irq_domain *domain, unsigned int virq if (err) { if (i > 0) - its_vpe_irq_domain_free(domain, virq, i - 1); + its_vpe_irq_domain_free(domain, virq, i); its_lpi_free(bitmap, base, nr_ids); its_free_prop_table(vprop_page); @@ -3960,8 +4531,12 @@ static int its_vpe_irq_domain_activate(struct irq_domain *domain, struct its_vpe *vpe = irq_data_get_irq_chip_data(d); struct its_node *its; - /* If we use the list map, we issue VMAPP on demand... */ - if (its_list_map) + /* + * If we use the list map, we issue VMAPP on demand... Unless + * we're on a GICv4.1 and we eagerly map the VPE on all ITSs + * so that VSGIs can work. + */ + if (!gic_requires_eager_mapping()) return 0; /* Map the VPE to the first possible CPU */ @@ -3987,10 +4562,10 @@ static void its_vpe_irq_domain_deactivate(struct irq_domain *domain, struct its_node *its; /* - * If we use the list map, we unmap the VPE once no VLPIs are - * associated with the VM. + * If we use the list map on GICv4.0, we unmap the VPE once no + * VLPIs are associated with the VM. */ - if (its_list_map) + if (!gic_requires_eager_mapping()) return; list_for_each_entry(its, &its_nodes, entry) { @@ -3999,6 +4574,15 @@ static void its_vpe_irq_domain_deactivate(struct irq_domain *domain, its_send_vmapp(its, vpe, false); } + + /* + * There may be a direct read to the VPT after unmapping the + * vPE, to guarantee the validity of this, we make the VPT + * memory coherent with the CPU caches here. + */ + if (find_4_1_its() && !atomic_read(&vpe->vmapp_count)) + gic_flush_dcache_to_poc(page_address(vpe->vpt_page), + LPI_PENDBASE_SZ); } static const struct irq_domain_ops its_vpe_domain_ops = { @@ -4192,9 +4776,6 @@ static int its_save_disable(void) list_for_each_entry(its, &its_nodes, entry) { void __iomem *base; - if (!(its->flags & ITS_FLAGS_SAVE_SUSPEND_STATE)) - continue; - base = its->base; its->ctlr_save = readl_relaxed(base + GITS_CTLR); err = its_force_quiescent(base); @@ -4213,9 +4794,6 @@ err: list_for_each_entry_continue_reverse(its, &its_nodes, entry) { void __iomem *base; - if (!(its->flags & ITS_FLAGS_SAVE_SUSPEND_STATE)) - continue; - base = its->base; writel_relaxed(its->ctlr_save, base + GITS_CTLR); } @@ -4235,9 +4813,6 @@ static void its_restore_enable(void) void __iomem *base; int i; - if (!(its->flags & ITS_FLAGS_SAVE_SUSPEND_STATE)) - continue; - base = its->base; /* @@ -4245,7 +4820,10 @@ static void its_restore_enable(void) * don't restore it since writing to CBASER or BASER<n> * registers is undefined according to the GIC v3 ITS * Specification. + * + * Firmware resuming with the ITS enabled is terminally broken. */ + WARN_ON(readl_relaxed(base + GITS_CTLR) & GITS_CTLR_ENABLE); ret = its_force_quiescent(base); if (ret) { pr_err("ITS@%pa: failed to quiesce on resume: %d\n", @@ -4290,6 +4868,38 @@ static struct syscore_ops its_syscore_ops = { .resume = its_restore_enable, }; +static void __init __iomem *its_map_one(struct resource *res, int *err) +{ + void __iomem *its_base; + u32 val; + + its_base = ioremap(res->start, SZ_64K); + if (!its_base) { + pr_warn("ITS@%pa: Unable to map ITS registers\n", &res->start); + *err = -ENOMEM; + return NULL; + } + + val = readl_relaxed(its_base + GITS_PIDR2) & GIC_PIDR2_ARCH_MASK; + if (val != 0x30 && val != 0x40) { + pr_warn("ITS@%pa: No ITS detected, giving up\n", &res->start); + *err = -ENODEV; + goto out_unmap; + } + + *err = its_force_quiescent(its_base); + if (*err) { + pr_warn("ITS@%pa: Failed to quiesce, giving up\n", &res->start); + goto out_unmap; + } + + return its_base; + +out_unmap: + iounmap(its_base); + return NULL; +} + static int its_init_domain(struct fwnode_handle *handle, struct its_node *its) { struct irq_domain *inner_domain; @@ -4332,10 +4942,8 @@ static int its_init_vpe_domain(void) entries = roundup_pow_of_two(nr_cpu_ids); vpe_proxy.vpes = kcalloc(entries, sizeof(*vpe_proxy.vpes), GFP_KERNEL); - if (!vpe_proxy.vpes) { - pr_err("ITS: Can't allocate GICv4 proxy device array\n"); + if (!vpe_proxy.vpes) return -ENOMEM; - } /* Use the last possible DevID */ devid = GENMASK(device_ids(its) - 1, 0); @@ -4399,29 +5007,14 @@ static int __init its_probe_one(struct resource *res, { struct its_node *its; void __iomem *its_base; - u32 val, ctlr; u64 baser, tmp, typer; struct page *page; + u32 ctlr; int err; - its_base = ioremap(res->start, resource_size(res)); - if (!its_base) { - pr_warn("ITS@%pa: Unable to map ITS registers\n", &res->start); - return -ENOMEM; - } - - val = readl_relaxed(its_base + GITS_PIDR2) & GIC_PIDR2_ARCH_MASK; - if (val != 0x30 && val != 0x40) { - pr_warn("ITS@%pa: No ITS detected, giving up\n", &res->start); - err = -ENODEV; - goto out_unmap; - } - - err = its_force_quiescent(its_base); - if (err) { - pr_warn("ITS@%pa: Failed to quiesce, giving up\n", &res->start); - goto out_unmap; - } + its_base = its_map_one(res, &err); + if (!its_base) + return err; pr_info("ITS %pR\n", res); @@ -4455,6 +5048,13 @@ static int __init its_probe_one(struct resource *res, if (is_v4_1(its)) { u32 svpet = FIELD_GET(GITS_TYPER_SVPET, typer); + + its->sgir_base = ioremap(res->start + SZ_128K, SZ_64K); + if (!its->sgir_base) { + err = -ENOMEM; + goto out_free_its; + } + its->mpidr = readl_relaxed(its_base + GITS_MPIDR); pr_info("ITS@%pa: Using GICv4.1 mode %08x %08x\n", @@ -4468,7 +5068,7 @@ static int __init its_probe_one(struct resource *res, get_order(ITS_CMD_QUEUE_SZ)); if (!page) { err = -ENOMEM; - goto out_free_its; + goto out_unmap_sgir; } its->cmd_base = (void *)page_address(page); its->cmd_write = its->cmd_base; @@ -4518,9 +5118,6 @@ static int __init its_probe_one(struct resource *res, ctlr |= GITS_CTLR_ImDe; writel_relaxed(ctlr, its->base + GITS_CTLR); - if (GITS_TYPER_HCC(typer)) - its->flags |= ITS_FLAGS_SAVE_SUSPEND_STATE; - err = its_init_domain(handle, its); if (err) goto out_free_tables; @@ -4535,6 +5132,9 @@ out_free_tables: its_free_tables(its); out_free_cmd: free_pages((unsigned long)its->cmd_base, get_order(ITS_CMD_QUEUE_SZ)); +out_unmap_sgir: + if (its->sgir_base) + iounmap(its->sgir_base); out_free_its: kfree(its); out_unmap: @@ -4570,7 +5170,7 @@ static int redist_disable_lpis(void) * * If running with preallocated tables, there is nothing to do. */ - if (gic_data_rdist()->lpi_enabled || + if ((gic_data_rdist()->flags & RD_LOCAL_LPI_ENABLED) || (gic_rdists->flags & RDIST_FLAGS_RD_TABLES_PREALLOCATED)) return 0; @@ -4632,6 +5232,69 @@ int its_cpu_init(void) return 0; } +static void rdist_memreserve_cpuhp_cleanup_workfn(struct work_struct *work) +{ + cpuhp_remove_state_nocalls(gic_rdists->cpuhp_memreserve_state); + gic_rdists->cpuhp_memreserve_state = CPUHP_INVALID; +} + +static DECLARE_WORK(rdist_memreserve_cpuhp_cleanup_work, + rdist_memreserve_cpuhp_cleanup_workfn); + +static int its_cpu_memreserve_lpi(unsigned int cpu) +{ + struct page *pend_page; + int ret = 0; + + /* This gets to run exactly once per CPU */ + if (gic_data_rdist()->flags & RD_LOCAL_MEMRESERVE_DONE) + return 0; + + pend_page = gic_data_rdist()->pend_page; + if (WARN_ON(!pend_page)) { + ret = -ENOMEM; + goto out; + } + /* + * If the pending table was pre-programmed, free the memory we + * preemptively allocated. Otherwise, reserve that memory for + * later kexecs. + */ + if (gic_data_rdist()->flags & RD_LOCAL_PENDTABLE_PREALLOCATED) { + its_free_pending_table(pend_page); + gic_data_rdist()->pend_page = NULL; + } else { + phys_addr_t paddr = page_to_phys(pend_page); + WARN_ON(gic_reserve_range(paddr, LPI_PENDBASE_SZ)); + } + +out: + /* Last CPU being brought up gets to issue the cleanup */ + if (!IS_ENABLED(CONFIG_SMP) || + cpumask_equal(&cpus_booted_once_mask, cpu_possible_mask)) + schedule_work(&rdist_memreserve_cpuhp_cleanup_work); + + gic_data_rdist()->flags |= RD_LOCAL_MEMRESERVE_DONE; + return ret; +} + +/* Mark all the BASER registers as invalid before they get reprogrammed */ +static int __init its_reset_one(struct resource *res) +{ + void __iomem *its_base; + int err, i; + + its_base = its_map_one(res, &err); + if (!its_base) + return err; + + for (i = 0; i < GITS_BASER_NR_REGS; i++) + gits_write_baser(0, its_base + GITS_BASER + (i << 3)); + + iounmap(its_base); + return 0; +} + static const struct of_device_id its_device_id[] = { { .compatible = "arm,gic-v3-its", }, {}, @@ -4642,6 +5305,26 @@ static int __init its_of_probe(struct device_node *node) struct device_node *np; struct resource res; + /* + * Make sure *all* the ITS are reset before we probe any, as + * they may be sharing memory. If any of the ITS fails to + * reset, don't even try to go any further, as this could + * result in something even worse. + */ + for (np = of_find_matching_node(node, its_device_id); np; + np = of_find_matching_node(np, its_device_id)) { + int err; + + if (!of_device_is_available(np) || + !of_property_read_bool(np, "msi-controller") || + of_address_to_resource(np, 0, &res)) + continue; + + err = its_reset_one(&res); + if (err) + return err; + } + for (np = of_find_matching_node(node, its_device_id); np; np = of_find_matching_node(np, its_device_id)) { if (!of_device_is_available(np)) @@ -4710,7 +5393,12 @@ static int __init gic_acpi_parse_srat_its(union acpi_subtable_headers *header, return -EINVAL; } - node = acpi_map_pxm_to_node(its_affinity->proximity_domain); + /* + * Note that in theory a new proximity node could be created by this + * entry as it is an SRAT resource allocation structure. + * We do not currently support doing so. + */ + node = pxm_to_node(its_affinity->proximity_domain); if (node == NUMA_NO_NODE || node >= MAX_NUMNODES) { pr_err("SRAT: Invalid NUMA node %d in ITS affinity\n", node); @@ -4739,10 +5427,8 @@ static void __init acpi_table_parse_srat_its(void) its_srat_maps = kmalloc_array(count, sizeof(struct its_srat_map), GFP_KERNEL); - if (!its_srat_maps) { - pr_warn("SRAT: Failed to allocate memory for its_srat_maps!\n"); + if (!its_srat_maps) return; - } acpi_table_parse_entries(ACPI_SIG_SRAT, sizeof(struct acpi_table_srat), @@ -4801,23 +5487,71 @@ dom_err: return err; } +static int __init its_acpi_reset(union acpi_subtable_headers *header, + const unsigned long end) +{ + struct acpi_madt_generic_translator *its_entry; + struct resource res; + + its_entry = (struct acpi_madt_generic_translator *)header; + res = (struct resource) { + .start = its_entry->base_address, + .end = its_entry->base_address + ACPI_GICV3_ITS_MEM_SIZE - 1, + .flags = IORESOURCE_MEM, + }; + + return its_reset_one(&res); +} + static void __init its_acpi_probe(void) { acpi_table_parse_srat_its(); - acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_TRANSLATOR, - gic_acpi_parse_madt_its, 0); + /* + * Make sure *all* the ITS are reset before we probe any, as + * they may be sharing memory. If any of the ITS fails to + * reset, don't even try to go any further, as this could + * result in something even worse. + */ + if (acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_TRANSLATOR, + its_acpi_reset, 0) > 0) + acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_TRANSLATOR, + gic_acpi_parse_madt_its, 0); acpi_its_srat_maps_free(); } #else static void __init its_acpi_probe(void) { } #endif +int __init its_lpi_memreserve_init(void) +{ + int state; + + if (!efi_enabled(EFI_CONFIG_TABLES)) + return 0; + + if (list_empty(&its_nodes)) + return 0; + + gic_rdists->cpuhp_memreserve_state = CPUHP_INVALID; + state = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, + "irqchip/arm/gicv3/memreserve:online", + its_cpu_memreserve_lpi, + NULL); + if (state < 0) + return state; + + gic_rdists->cpuhp_memreserve_state = state; + + return 0; +} + int __init its_init(struct fwnode_handle *handle, struct rdists *rdists, struct irq_domain *parent_domain) { struct device_node *of_node; struct its_node *its; bool has_v4 = false; + bool has_v4_1 = false; int err; gic_rdists = rdists; @@ -4838,12 +5572,25 @@ int __init its_init(struct fwnode_handle *handle, struct rdists *rdists, if (err) return err; - list_for_each_entry(its, &its_nodes, entry) + list_for_each_entry(its, &its_nodes, entry) { has_v4 |= is_v4(its); + has_v4_1 |= is_v4_1(its); + } + + /* Don't bother with inconsistent systems */ + if (WARN_ON(!has_v4_1 && rdists->has_rvpeid)) + rdists->has_rvpeid = false; if (has_v4 & rdists->has_vlpis) { + const struct irq_domain_ops *sgi_ops; + + if (has_v4_1) + sgi_ops = &its_sgi_domain_ops; + else + sgi_ops = NULL; + if (its_init_vpe_domain() || - its_init_v4(parent_domain, &its_vpe_domain_ops)) { + its_init_v4(parent_domain, &its_vpe_domain_ops, sgi_ops)) { rdists->has_vlpis = false; pr_err("ITS: Disabling GICv4 support\n"); } |