Merge branch 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/p…

…aulus/powerpc

* 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] sys_move_pages should be callable from an SPU
  [POWERPC] Wire up sys_epoll_pwait
  [POWERPC] Allocate syscall number for sys_getcpu
  [POWERPC] update cell_defconfig
  [POWERPC] ps3: always make sure were running on a PS3
  [POWERPC] Fix spu SLB invalidations
  [POWERPC] avoid SPU_ACTIVATE_NOWAKE optimization
  [POWERPC] spufs: fix possible memory corruption is spufs_mem_write

arch/powerpc/mm/hash_utils_64.c

@@ -685,6 +685,9 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
 				       "non-cacheable mapping\n");
 				psize = mmu_vmalloc_psize = MMU_PAGE_4K;
 			}
+#ifdef CONFIG_SPE_BASE
+			spu_flush_all_slbs(mm);
+#endif
 		}
 		if (user_region) {
 			if (psize != get_paca()->context.user_psize) {
@@ -759,6 +762,9 @@ void hash_preload(struct mm_struct *mm, unsigned long ea,
 				mmu_psize_defs[MMU_PAGE_4K].sllp;
 			get_paca()->context = mm->context;
 			slb_flush_and_rebolt();
+#ifdef CONFIG_SPE_BASE
+			spu_flush_all_slbs(mm);
+#endif
 		}
 	}
 	if (mm->context.user_psize == MMU_PAGE_64K)

arch/powerpc/mm/hugetlbpage.c

@@ -24,6 +24,7 @@
 #include <asm/machdep.h>
 #include <asm/cputable.h>
 #include <asm/tlb.h>
+#include <asm/spu.h>
 
 #include <linux/sysctl.h>
 
@@ -513,6 +514,9 @@ int prepare_hugepage_range(unsigned long addr, unsigned long len, pgoff_t pgoff)
 	if ((addr + len) > 0x100000000UL)
 		err = open_high_hpage_areas(current->mm,
 					    HTLB_AREA_MASK(addr, len));
+#ifdef CONFIG_SPE_BASE
+	spu_flush_all_slbs(current->mm);
+#endif
 	if (err) {
 		printk(KERN_DEBUG "prepare_hugepage_range(%lx, %lx)"
 		       " failed (lowmask: 0x%04hx, highmask: 0x%04hx)\n",

arch/powerpc/platforms/cell/spu_base.c

@@ -38,8 +38,61 @@
 const struct spu_management_ops *spu_management_ops;
 const struct spu_priv1_ops *spu_priv1_ops;
 
+static struct list_head spu_list[MAX_NUMNODES];
+static LIST_HEAD(spu_full_list);
+static DEFINE_MUTEX(spu_mutex);
+static spinlock_t spu_list_lock = SPIN_LOCK_UNLOCKED;
+
 EXPORT_SYMBOL_GPL(spu_priv1_ops);
 
+void spu_invalidate_slbs(struct spu *spu)
+{
+	struct spu_priv2 __iomem *priv2 = spu->priv2;
+
+	if (spu_mfc_sr1_get(spu) & MFC_STATE1_RELOCATE_MASK)
+		out_be64(&priv2->slb_invalidate_all_W, 0UL);
+}
+EXPORT_SYMBOL_GPL(spu_invalidate_slbs);
+
+/* This is called by the MM core when a segment size is changed, to
+ * request a flush of all the SPEs using a given mm
+ */
+void spu_flush_all_slbs(struct mm_struct *mm)
+{
+	struct spu *spu;
+	unsigned long flags;
+
+	spin_lock_irqsave(&spu_list_lock, flags);
+	list_for_each_entry(spu, &spu_full_list, full_list) {
+		if (spu->mm == mm)
+			spu_invalidate_slbs(spu);
+	}
+	spin_unlock_irqrestore(&spu_list_lock, flags);
+}
+
+/* The hack below stinks... try to do something better one of
+ * these days... Does it even work properly with NR_CPUS == 1 ?
+ */
+static inline void mm_needs_global_tlbie(struct mm_struct *mm)
+{
+	int nr = (NR_CPUS > 1) ? NR_CPUS : NR_CPUS + 1;
+
+	/* Global TLBIE broadcast required with SPEs. */
+	__cpus_setall(&mm->cpu_vm_mask, nr);
+}
+
+void spu_associate_mm(struct spu *spu, struct mm_struct *mm)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&spu_list_lock, flags);
+	spu->mm = mm;
+	spin_unlock_irqrestore(&spu_list_lock, flags);
+	if (mm)
+		mm_needs_global_tlbie(mm);
+}
+EXPORT_SYMBOL_GPL(spu_associate_mm);
+
 static int __spu_trap_invalid_dma(struct spu *spu)
 {
 	pr_debug("%s\n", __FUNCTION__);
@@ -74,6 +127,7 @@ static int __spu_trap_data_seg(struct spu *spu, unsigned long ea)
 	struct spu_priv2 __iomem *priv2 = spu->priv2;
 	struct mm_struct *mm = spu->mm;
 	u64 esid, vsid, llp;
+	int psize;
 
 	pr_debug("%s\n", __FUNCTION__);
 
@@ -90,22 +144,25 @@ static int __spu_trap_data_seg(struct spu *spu, unsigned long ea)
 	case USER_REGION_ID:
 #ifdef CONFIG_HUGETLB_PAGE
 		if (in_hugepage_area(mm->context, ea))
-			llp = mmu_psize_defs[mmu_huge_psize].sllp;
+			psize = mmu_huge_psize;
 		else
 #endif
-			llp = mmu_psize_defs[mmu_virtual_psize].sllp;
+			psize = mm->context.user_psize;
 		vsid = (get_vsid(mm->context.id, ea) << SLB_VSID_SHIFT) |
-				SLB_VSID_USER | llp;
+				SLB_VSID_USER;
 		break;
 	case VMALLOC_REGION_ID:
-		llp = mmu_psize_defs[mmu_virtual_psize].sllp;
+		if (ea < VMALLOC_END)
+			psize = mmu_vmalloc_psize;
+		else
+			psize = mmu_io_psize;
 		vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
-			SLB_VSID_KERNEL | llp;
+			SLB_VSID_KERNEL;
 		break;
 	case KERNEL_REGION_ID:
-		llp = mmu_psize_defs[mmu_linear_psize].sllp;
+		psize = mmu_linear_psize;
 		vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
-			SLB_VSID_KERNEL | llp;
+			SLB_VSID_KERNEL;
 		break;
 	default:
 		/* Future: support kernel segments so that drivers
@@ -114,9 +171,10 @@ static int __spu_trap_data_seg(struct spu *spu, unsigned long ea)
 		pr_debug("invalid region access at %016lx\n", ea);
 		return 1;
 	}
+	llp = mmu_psize_defs[psize].sllp;
 
 	out_be64(&priv2->slb_index_W, spu->slb_replace);
-	out_be64(&priv2->slb_vsid_RW, vsid);
+	out_be64(&priv2->slb_vsid_RW, vsid | llp);
 	out_be64(&priv2->slb_esid_RW, esid);
 
 	spu->slb_replace++;
@@ -330,10 +388,6 @@ static void spu_free_irqs(struct spu *spu)
 		free_irq(spu->irqs[2], spu);
 }
 
-static struct list_head spu_list[MAX_NUMNODES];
-static LIST_HEAD(spu_full_list);
-static DEFINE_MUTEX(spu_mutex);
-
 static void spu_init_channels(struct spu *spu)
 {
 	static const struct {
@@ -593,6 +647,7 @@ static int __init create_spu(void *data)
 	struct spu *spu;
 	int ret;
 	static int number;
+	unsigned long flags;
 
 	ret = -ENOMEM;
 	spu = kzalloc(sizeof (*spu), GFP_KERNEL);
@@ -620,8 +675,10 @@ static int __init create_spu(void *data)
 		goto out_free_irqs;
 
 	mutex_lock(&spu_mutex);
+	spin_lock_irqsave(&spu_list_lock, flags);
 	list_add(&spu->list, &spu_list[spu->node]);
 	list_add(&spu->full_list, &spu_full_list);
+	spin_unlock_irqrestore(&spu_list_lock, flags);
 	mutex_unlock(&spu_mutex);
 
 	goto out;

arch/powerpc/platforms/cell/spufs/file.c

@@ -63,8 +63,8 @@ static ssize_t
 spufs_mem_read(struct file *file, char __user *buffer,
 				size_t size, loff_t *pos)
 {
-	int ret;
 	struct spu_context *ctx = file->private_data;
+	ssize_t ret;
 
 	spu_acquire(ctx);
 	ret = __spufs_mem_read(ctx, buffer, size, pos);
@@ -74,25 +74,29 @@ spufs_mem_read(struct file *file, char __user *buffer,
 
 static ssize_t
 spufs_mem_write(struct file *file, const char __user *buffer,
-					size_t size, loff_t *pos)
+					size_t size, loff_t *ppos)
 {
 	struct spu_context *ctx = file->private_data;
 	char *local_store;
+	loff_t pos = *ppos;
 	int ret;
 
-	size = min_t(ssize_t, LS_SIZE - *pos, size);
-	if (size <= 0)
+	if (pos < 0)
+		return -EINVAL;
+	if (pos > LS_SIZE)
 		return -EFBIG;
-	*pos += size;
+	if (size > LS_SIZE - pos)
+		size = LS_SIZE - pos;
 
 	spu_acquire(ctx);
-
 	local_store = ctx->ops->get_ls(ctx);
-	ret = copy_from_user(local_store + *pos - size,
-			     buffer, size) ? -EFAULT : size;
-
+	ret = copy_from_user(local_store + pos, buffer, size);
 	spu_release(ctx);
-	return ret;
+
+	if (ret)
+		return -EFAULT;
+	*ppos = pos + size;
+	return size;
 }
 
 static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct *vma,

arch/powerpc/platforms/cell/spufs/run.c

@@ -143,7 +143,7 @@ static inline int spu_run_init(struct spu_context *ctx, u32 * npc)
 	int ret;
 	unsigned long runcntl = SPU_RUNCNTL_RUNNABLE;
 
-	ret = spu_acquire_runnable(ctx, SPU_ACTIVATE_NOWAKE);
+	ret = spu_acquire_runnable(ctx, 0);
 	if (ret)
 		return ret;
 
@@ -155,7 +155,7 @@ static inline int spu_run_init(struct spu_context *ctx, u32 * npc)
 			spu_release(ctx);
 			ret = spu_setup_isolated(ctx);
 			if (!ret)
-				ret = spu_acquire_runnable(ctx, SPU_ACTIVATE_NOWAKE);
+				ret = spu_acquire_runnable(ctx, 0);
 		}
 
 		/* if userspace has set the runcntrl register (eg, to issue an

arch/powerpc/platforms/cell/spufs/sched.c

@@ -127,14 +127,6 @@ static void spu_remove_from_active_list(struct spu *spu)
 	mutex_unlock(&spu_prio->active_mutex[node]);
 }
 
-static inline void mm_needs_global_tlbie(struct mm_struct *mm)
-{
-	int nr = (NR_CPUS > 1) ? NR_CPUS : NR_CPUS + 1;
-
-	/* Global TLBIE broadcast required with SPEs. */
-	__cpus_setall(&mm->cpu_vm_mask, nr);
-}
-
 static BLOCKING_NOTIFIER_HEAD(spu_switch_notifier);
 
 static void spu_switch_notify(struct spu *spu, struct spu_context *ctx)
@@ -167,8 +159,7 @@ static void spu_bind_context(struct spu *spu, struct spu_context *ctx)
 	ctx->spu = spu;
 	ctx->ops = &spu_hw_ops;
 	spu->pid = current->pid;
-	spu->mm = ctx->owner;
-	mm_needs_global_tlbie(spu->mm);
+	spu_associate_mm(spu, ctx->owner);
 	spu->ibox_callback = spufs_ibox_callback;
 	spu->wbox_callback = spufs_wbox_callback;
 	spu->stop_callback = spufs_stop_callback;
@@ -205,7 +196,7 @@ static void spu_unbind_context(struct spu *spu, struct spu_context *ctx)
 	spu->stop_callback = NULL;
 	spu->mfc_callback = NULL;
 	spu->dma_callback = NULL;
-	spu->mm = NULL;
+	spu_associate_mm(spu, NULL);
 	spu->pid = 0;
 	ctx->ops = &spu_backing_ops;
 	ctx->spu = NULL;
@@ -263,7 +254,6 @@ static void spu_prio_wait(struct spu_context *ctx)
 {
 	DEFINE_WAIT(wait);
 
-	set_bit(SPU_SCHED_WAKE, &ctx->sched_flags);
 	prepare_to_wait_exclusive(&ctx->stop_wq, &wait, TASK_INTERRUPTIBLE);
 	if (!signal_pending(current)) {
 		mutex_unlock(&ctx->state_mutex);
@@ -272,7 +262,6 @@ static void spu_prio_wait(struct spu_context *ctx)
 	}
 	__set_current_state(TASK_RUNNING);
 	remove_wait_queue(&ctx->stop_wq, &wait);
-	clear_bit(SPU_SCHED_WAKE, &ctx->sched_flags);
 }
 
 /**
@@ -292,7 +281,7 @@ static void spu_reschedule(struct spu *spu)
 	best = sched_find_first_bit(spu_prio->bitmap);
 	if (best < MAX_PRIO) {
 		struct spu_context *ctx = spu_grab_context(best);
-		if (ctx && test_bit(SPU_SCHED_WAKE, &ctx->sched_flags))
+		if (ctx)
 			wake_up(&ctx->stop_wq);
 	}
 	spin_unlock(&spu_prio->runq_lock);
@@ -414,8 +403,7 @@ int spu_activate(struct spu_context *ctx, unsigned long flags)
 		}
 
 		spu_add_to_rq(ctx);
-		if (!(flags & SPU_ACTIVATE_NOWAKE))
-			spu_prio_wait(ctx);
+		spu_prio_wait(ctx);
 		spu_del_from_rq(ctx);
 	} while (!signal_pending(current));
 

arch/powerpc/platforms/cell/spufs/spufs.h

@@ -41,7 +41,7 @@ struct spu_gang;
 
 /* ctx->sched_flags */
 enum {
-	SPU_SCHED_WAKE = 0,
+	SPU_SCHED_WAKE = 0, /* currently unused */
 };
 
 struct spu_context {
@@ -191,9 +191,7 @@ void spu_forget(struct spu_context *ctx);
 int spu_acquire_runnable(struct spu_context *ctx, unsigned long flags);
 void spu_acquire_saved(struct spu_context *ctx);
 int spu_acquire_exclusive(struct spu_context *ctx);
-enum {
-	SPU_ACTIVATE_NOWAKE = 1,
-};
+
 int spu_activate(struct spu_context *ctx, unsigned long flags);
 void spu_deactivate(struct spu_context *ctx);
 void spu_yield(struct spu_context *ctx);