1 files changed, 211 insertions, 279 deletions

diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index a2aabfdc80d..5845ffea67c 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -40,6 +40,9 @@
 #include <linux/capability.h>
 #include <linux/rcupdate.h>
 #include <linux/completion.h>
+#include <linux/tracehook.h>
+#include <linux/slab.h>
+#include <linux/cpu.h>
 
 #include <asm/errno.h>
 #include <asm/intrinsics.h>
@@ -47,7 +50,6 @@
 #include <asm/perfmon.h>
 #include <asm/processor.h>
 #include <asm/signal.h>
-#include <asm/system.h>
 #include <asm/uaccess.h>
 #include <asm/delay.h>
 
@@ -311,7 +313,7 @@ typedef struct pfm_context {
 	unsigned long		th_pmcs[PFM_NUM_PMC_REGS];	/* PMC thread save state */
 	unsigned long		th_pmds[PFM_NUM_PMD_REGS];	/* PMD thread save state */
 
-	u64			ctx_saved_psr_up;	/* only contains psr.up value */
+	unsigned long		ctx_saved_psr_up;	/* only contains psr.up value */
 
 	unsigned long		ctx_last_activation;	/* context last activation number for last_cpu */
 	unsigned int		ctx_last_cpu;		/* CPU id of current or last CPU used (SMP only) */
@@ -519,53 +521,47 @@ static pmu_config_t		*pmu_conf;
 pfm_sysctl_t pfm_sysctl;
 EXPORT_SYMBOL(pfm_sysctl);
 
-static ctl_table pfm_ctl_table[]={
+static struct ctl_table pfm_ctl_table[] = {
 	{
-		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "debug",
 		.data		= &pfm_sysctl.debug,
 		.maxlen		= sizeof(int),
 		.mode		= 0666,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
 	{
-		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "debug_ovfl",
 		.data		= &pfm_sysctl.debug_ovfl,
 		.maxlen		= sizeof(int),
 		.mode		= 0666,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
 	{
-		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "fastctxsw",
 		.data		= &pfm_sysctl.fastctxsw,
 		.maxlen		= sizeof(int),
 		.mode		= 0600,
-		.proc_handler	=  &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
 	{
-		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "expert_mode",
 		.data		= &pfm_sysctl.expert_mode,
 		.maxlen		= sizeof(int),
 		.mode		= 0600,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
 	{}
 };
-static ctl_table pfm_sysctl_dir[] = {
+static struct ctl_table pfm_sysctl_dir[] = {
 	{
-		.ctl_name	= CTL_UNNUMBERED,
 		.procname	= "perfmon",
 		.mode		= 0555,
 		.child		= pfm_ctl_table,
 	},
  	{}
 };
-static ctl_table pfm_sysctl_root[] = {
+static struct ctl_table pfm_sysctl_root[] = {
 	{
-		.ctl_name	= CTL_KERN,
 		.procname	= "kernel",
 		.mode		= 0555,
 		.child		= pfm_sysctl_dir,
@@ -609,31 +605,22 @@ pfm_unprotect_ctx_ctxsw(pfm_context_t *x, unsigned long f)
 	spin_unlock(&(x)->ctx_lock);
 }
 
-static inline unsigned int
-pfm_do_munmap(struct mm_struct *mm, unsigned long addr, size_t len, int acct)
-{
-	return do_munmap(mm, addr, len);
-}
-
-static inline unsigned long 
-pfm_get_unmapped_area(struct file *file, unsigned long addr, unsigned long len, unsigned long pgoff, unsigned long flags, unsigned long exec)
-{
-	return get_unmapped_area(file, addr, len, pgoff, flags);
-}
-
+/* forward declaration */
+static const struct dentry_operations pfmfs_dentry_operations;
 
-static int
-pfmfs_get_sb(struct file_system_type *fs_type, int flags, const char *dev_name, void *data,
-	     struct vfsmount *mnt)
+static struct dentry *
+pfmfs_mount(struct file_system_type *fs_type, int flags, const char *dev_name, void *data)
 {
-	return get_sb_pseudo(fs_type, "pfm:", NULL, PFMFS_MAGIC, mnt);
+	return mount_pseudo(fs_type, "pfm:", NULL, &pfmfs_dentry_operations,
+			PFMFS_MAGIC);
 }
 
 static struct file_system_type pfm_fs_type = {
 	.name     = "pfmfs",
-	.get_sb   = pfmfs_get_sb,
+	.mount    = pfmfs_mount,
 	.kill_sb  = kill_anon_super,
 };
+MODULE_ALIAS_FS("pfmfs");
 
 DEFINE_PER_CPU(unsigned long, pfm_syst_info);
 DEFINE_PER_CPU(struct task_struct *, pmu_owner);
@@ -834,10 +821,9 @@ pfm_rvmalloc(unsigned long size)
 	unsigned long addr;
 
 	size = PAGE_ALIGN(size);
-	mem  = vmalloc(size);
+	mem  = vzalloc(size);
 	if (mem) {
 		//printk("perfmon: CPU%d pfm_rvmalloc(%ld)=%p\n", smp_processor_id(), size, mem);
-		memset(mem, 0, size);
 		addr = (unsigned long)mem;
 		while (size > 0) {
 			pfm_reserve_page(addr);
@@ -867,7 +853,7 @@ pfm_rvfree(void *mem, unsigned long size)
 }
 
 static pfm_context_t *
-pfm_context_alloc(void)
+pfm_context_alloc(int ctx_flags)
 {
 	pfm_context_t *ctx;
 
@@ -878,6 +864,46 @@ pfm_context_alloc(void)
 	ctx = kzalloc(sizeof(pfm_context_t), GFP_KERNEL);
 	if (ctx) {
 		DPRINT(("alloc ctx @%p\n", ctx));
+
+		/*
+		 * init context protection lock
+		 */
+		spin_lock_init(&ctx->ctx_lock);
+
+		/*
+		 * context is unloaded
+		 */
+		ctx->ctx_state = PFM_CTX_UNLOADED;
+
+		/*
+		 * initialization of context's flags
+		 */
+		ctx->ctx_fl_block       = (ctx_flags & PFM_FL_NOTIFY_BLOCK) ? 1 : 0;
+		ctx->ctx_fl_system      = (ctx_flags & PFM_FL_SYSTEM_WIDE) ? 1: 0;
+		ctx->ctx_fl_no_msg      = (ctx_flags & PFM_FL_OVFL_NO_MSG) ? 1: 0;
+		/*
+		 * will move to set properties
+		 * ctx->ctx_fl_excl_idle   = (ctx_flags & PFM_FL_EXCL_IDLE) ? 1: 0;
+		 */
+
+		/*
+		 * init restart semaphore to locked
+		 */
+		init_completion(&ctx->ctx_restart_done);
+
+		/*
+		 * activation is used in SMP only
+		 */
+		ctx->ctx_last_activation = PFM_INVALID_ACTIVATION;
+		SET_LAST_CPU(ctx, -1);
+
+		/*
+		 * initialize notification message queue
+		 */
+		ctx->ctx_msgq_head = ctx->ctx_msgq_tail = 0;
+		init_waitqueue_head(&ctx->ctx_msgq_wait);
+		init_waitqueue_head(&ctx->ctx_zombieq);
+
 	}
 	return ctx;
 }
@@ -1297,8 +1323,6 @@ out:
 }
 EXPORT_SYMBOL(pfm_unregister_buffer_fmt);
 
-extern void update_pal_halt_status(int);
-
 static int
 pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu)
 {
@@ -1346,9 +1370,9 @@ pfm_reserve_session(struct task_struct *task, int is_syswide, unsigned int cpu)
 		cpu));
 
 	/*
-	 * disable default_idle() to go to PAL_HALT
+	 * Force idle() into poll mode
 	 */
-	update_pal_halt_status(0);
+	cpu_idle_poll_ctrl(true);
 
 	UNLOCK_PFS(flags);
 
@@ -1405,11 +1429,8 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu)
 		is_syswide,
 		cpu));
 
-	/*
-	 * if possible, enable default_idle() to go into PAL_HALT
-	 */
-	if (pfm_sessions.pfs_task_sessions == 0 && pfm_sessions.pfs_sys_sessions == 0)
-		update_pal_halt_status(1);
+	/* Undo forced polling. Last session reenables pal_halt */
+	cpu_idle_poll_ctrl(false);
 
 	UNLOCK_PFS(flags);
 
@@ -1422,8 +1443,9 @@ pfm_unreserve_session(pfm_context_t *ctx, int is_syswide, unsigned int cpu)
  * a PROTECT_CTX() section.
  */
 static int
-pfm_remove_smpl_mapping(struct task_struct *task, void *vaddr, unsigned long size)
+pfm_remove_smpl_mapping(void *vaddr, unsigned long size)
 {
+	struct task_struct *task = current;
 	int r;
 
 	/* sanity checks */
@@ -1437,13 +1459,8 @@ pfm_remove_smpl_mapping(struct task_struct *task, void *vaddr, unsigned long siz
 	/*
 	 * does the actual unmapping
 	 */
-	down_write(&task->mm->mmap_sem);
-
-	DPRINT(("down_write done smpl_vaddr=%p size=%lu\n", vaddr, size));
+	r = vm_munmap((unsigned long)vaddr, size);
 
-	r = pfm_do_munmap(task->mm, (unsigned long)vaddr, size, 0);
-
-	up_write(&task->mm->mmap_sem);
 	if (r !=0) {
 		printk(KERN_ERR "perfmon: [%d] unable to unmap sampling buffer @%p size=%lu\n", task_pid_nr(task), vaddr, size);
 	}
@@ -1507,7 +1524,7 @@ pfm_exit_smpl_buffer(pfm_buffer_fmt_t *fmt)
  * any operations on the root directory. However, we need a non-trivial
  * d_name - pfm: will go nicely and kill the special-casing in procfs.
  */
-static struct vfsmount *pfmfs_mnt;
+static struct vfsmount *pfmfs_mnt __read_mostly;
 
 static int __init
 init_pfm_fs(void)
@@ -1537,7 +1554,7 @@ pfm_read(struct file *filp, char __user *buf, size_t size, loff_t *ppos)
 		return -EINVAL;
 	}
 
-	ctx = (pfm_context_t *)filp->private_data;
+	ctx = filp->private_data;
 	if (ctx == NULL) {
 		printk(KERN_ERR "perfmon: pfm_read: NULL ctx [%d]\n", task_pid_nr(current));
 		return -EINVAL;
@@ -1637,7 +1654,7 @@ pfm_poll(struct file *filp, poll_table * wait)
 		return 0;
 	}
 
-	ctx = (pfm_context_t *)filp->private_data;
+	ctx = filp->private_data;
 	if (ctx == NULL) {
 		printk(KERN_ERR "perfmon: pfm_poll: NULL ctx [%d]\n", task_pid_nr(current));
 		return 0;
@@ -1660,8 +1677,8 @@ pfm_poll(struct file *filp, poll_table * wait)
 	return mask;
 }
 
-static int
-pfm_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg)
+static long
+pfm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	DPRINT(("pfm_ioctl called\n"));
 	return -EINVAL;
@@ -1697,7 +1714,7 @@ pfm_fasync(int fd, struct file *filp, int on)
 		return -EBADF;
 	}
 
-	ctx = (pfm_context_t *)filp->private_data;
+	ctx = filp->private_data;
 	if (ctx == NULL) {
 		printk(KERN_ERR "perfmon: pfm_fasync NULL ctx [%d]\n", task_pid_nr(current));
 		return -EBADF;
@@ -1780,7 +1797,7 @@ pfm_syswide_cleanup_other_cpu(pfm_context_t *ctx)
 	int ret;
 
 	DPRINT(("calling CPU%d for cleanup\n", ctx->ctx_cpu));
-	ret = smp_call_function_single(ctx->ctx_cpu, pfm_syswide_force_stop, ctx, 0, 1);
+	ret = smp_call_function_single(ctx->ctx_cpu, pfm_syswide_force_stop, ctx, 1);
 	DPRINT(("called CPU%d for cleanup ret=%d\n", ctx->ctx_cpu, ret));
 }
 #endif /* CONFIG_SMP */
@@ -1805,7 +1822,7 @@ pfm_flush(struct file *filp, fl_owner_t id)
 		return -EBADF;
 	}
 
-	ctx = (pfm_context_t *)filp->private_data;
+	ctx = filp->private_data;
 	if (ctx == NULL) {
 		printk(KERN_ERR "perfmon: pfm_flush: NULL ctx [%d]\n", task_pid_nr(current));
 		return -EBADF;
@@ -1824,11 +1841,6 @@ pfm_flush(struct file *filp, fl_owner_t id)
 	 * invoked after, it will find an empty queue and no
 	 * signal will be sent. In both case, we are safe
 	 */
-	if (filp->f_flags & FASYNC) {
-		DPRINT(("cleaning up async_queue=%p\n", ctx->ctx_async_queue));
-		pfm_do_fasync (-1, filp, ctx, 0);
-	}
-
 	PROTECT_CTX(ctx, flags);
 
 	state     = ctx->ctx_state;
@@ -1914,7 +1926,7 @@ pfm_flush(struct file *filp, fl_owner_t id)
 	 * because some VM function reenables interrupts.
 	 *
 	 */
-	if (smpl_buf_vaddr) pfm_remove_smpl_mapping(current, smpl_buf_vaddr, smpl_buf_size);
+	if (smpl_buf_vaddr) pfm_remove_smpl_mapping(smpl_buf_vaddr, smpl_buf_size);
 
 	return 0;
 }
@@ -1953,7 +1965,7 @@ pfm_close(struct inode *inode, struct file *filp)
 		return -EBADF;
 	}
 	
-	ctx = (pfm_context_t *)filp->private_data;
+	ctx = filp->private_data;
 	if (ctx == NULL) {
 		printk(KERN_ERR "perfmon: pfm_close: NULL ctx [%d]\n", task_pid_nr(current));
 		return -EBADF;
@@ -2143,115 +2155,72 @@ pfm_no_open(struct inode *irrelevant, struct file *dontcare)
 
 
 static const struct file_operations pfm_file_ops = {
-	.llseek   = no_llseek,
-	.read     = pfm_read,
-	.write    = pfm_write,
-	.poll     = pfm_poll,
-	.ioctl    = pfm_ioctl,
-	.open     = pfm_no_open,	/* special open code to disallow open via /proc */
-	.fasync   = pfm_fasync,
-	.release  = pfm_close,
-	.flush	  = pfm_flush
+	.llseek		= no_llseek,
+	.read		= pfm_read,
+	.write		= pfm_write,
+	.poll		= pfm_poll,
+	.unlocked_ioctl = pfm_ioctl,
+	.open		= pfm_no_open,	/* special open code to disallow open via /proc */
+	.fasync		= pfm_fasync,
+	.release	= pfm_close,
+	.flush		= pfm_flush
 };
 
-static int
-pfmfs_delete_dentry(struct dentry *dentry)
+static char *pfmfs_dname(struct dentry *dentry, char *buffer, int buflen)
 {
-	return 1;
+	return dynamic_dname(dentry, buffer, buflen, "pfm:[%lu]",
+			     dentry->d_inode->i_ino);
 }
 
-static struct dentry_operations pfmfs_dentry_operations = {
-	.d_delete = pfmfs_delete_dentry,
+static const struct dentry_operations pfmfs_dentry_operations = {
+	.d_delete = always_delete_dentry,
+	.d_dname = pfmfs_dname,
 };
 
 
-static int
-pfm_alloc_fd(struct file **cfile)
+static struct file *
+pfm_alloc_file(pfm_context_t *ctx)
 {
-	int fd, ret = 0;
-	struct file *file = NULL;
-	struct inode * inode;
-	char name[32];
-	struct qstr this;
-
-	fd = get_unused_fd();
-	if (fd < 0) return -ENFILE;
-
-	ret = -ENFILE;
-
-	file = get_empty_filp();
-	if (!file) goto out;
+	struct file *file;
+	struct inode *inode;
+	struct path path;
+	struct qstr this = { .name = "" };
 
 	/*
 	 * allocate a new inode
 	 */
 	inode = new_inode(pfmfs_mnt->mnt_sb);
-	if (!inode) goto out;
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
 
 	DPRINT(("new inode ino=%ld @%p\n", inode->i_ino, inode));
 
 	inode->i_mode = S_IFCHR|S_IRUGO;
-	inode->i_uid  = current->fsuid;
-	inode->i_gid  = current->fsgid;
-
-	sprintf(name, "[%lu]", inode->i_ino);
-	this.name = name;
-	this.len  = strlen(name);
-	this.hash = inode->i_ino;
-
-	ret = -ENOMEM;
+	inode->i_uid  = current_fsuid();
+	inode->i_gid  = current_fsgid();
 
 	/*
 	 * allocate a new dcache entry
 	 */
-	file->f_path.dentry = d_alloc(pfmfs_mnt->mnt_sb->s_root, &this);
-	if (!file->f_path.dentry) goto out;
+	path.dentry = d_alloc(pfmfs_mnt->mnt_root, &this);
+	if (!path.dentry) {
+		iput(inode);
+		return ERR_PTR(-ENOMEM);
+	}
+	path.mnt = mntget(pfmfs_mnt);
 
-	file->f_path.dentry->d_op = &pfmfs_dentry_operations;
+	d_add(path.dentry, inode);
 
-	d_add(file->f_path.dentry, inode);
-	file->f_path.mnt = mntget(pfmfs_mnt);
-	file->f_mapping = inode->i_mapping;
+	file = alloc_file(&path, FMODE_READ, &pfm_file_ops);
+	if (IS_ERR(file)) {
+		path_put(&path);
+		return file;
+	}
 
-	file->f_op    = &pfm_file_ops;
-	file->f_mode  = FMODE_READ;
 	file->f_flags = O_RDONLY;
-	file->f_pos   = 0;
+	file->private_data = ctx;
 
-	/*
-	 * may have to delay until context is attached?
-	 */
-	fd_install(fd, file);
-
-	/*
-	 * the file structure we will use
-	 */
-	*cfile = file;
-
-	return fd;
-out:
-	if (file) put_filp(file);
-	put_unused_fd(fd);
-	return ret;
-}
-
-static void
-pfm_free_fd(int fd, struct file *file)
-{
-	struct files_struct *files = current->files;
-	struct fdtable *fdt;
-
-	/* 
-	 * there ie no fd_uninstall(), so we do it here
-	 */
-	spin_lock(&files->file_lock);
-	fdt = files_fdtable(files);
-	rcu_assign_pointer(fdt->fd[fd], NULL);
-	spin_unlock(&files->file_lock);
-
-	if (file)
-		put_filp(file);
-	put_unused_fd(fd);
+	return file;
 }
 
 static int
@@ -2300,7 +2269,7 @@ pfm_smpl_buffer_alloc(struct task_struct *task, struct file *filp, pfm_context_t
 	 * if ((mm->total_vm << PAGE_SHIFT) + len> task->rlim[RLIMIT_AS].rlim_cur)
 	 * 	return -ENOMEM;
 	 */
-	if (size > task->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
+	if (size > task_rlimit(task, RLIMIT_MEMLOCK))
 		return -ENOMEM;
 
 	/*
@@ -2322,13 +2291,14 @@ pfm_smpl_buffer_alloc(struct task_struct *task, struct file *filp, pfm_context_t
 		DPRINT(("Cannot allocate vma\n"));
 		goto error_kmem;
 	}
+	INIT_LIST_HEAD(&vma->anon_vma_chain);
 
 	/*
 	 * partially initialize the vma for the sampling buffer
 	 */
 	vma->vm_mm	     = mm;
-	vma->vm_file	     = filp;
-	vma->vm_flags	     = VM_READ| VM_MAYREAD |VM_RESERVED;
+	vma->vm_file	     = get_file(filp);
+	vma->vm_flags	     = VM_READ|VM_MAYREAD|VM_DONTEXPAND|VM_DONTDUMP;
 	vma->vm_page_prot    = PAGE_READONLY; /* XXX may need to change */
 
 	/*
@@ -2348,8 +2318,8 @@ pfm_smpl_buffer_alloc(struct task_struct *task, struct file *filp, pfm_context_t
 	down_write(&task->mm->mmap_sem);
 
 	/* find some free area in address space, must have mmap sem held */
-	vma->vm_start = pfm_get_unmapped_area(NULL, 0, size, 0, MAP_PRIVATE|MAP_ANONYMOUS, 0);
-	if (vma->vm_start == 0UL) {
+	vma->vm_start = get_unmapped_area(NULL, 0, size, 0, MAP_PRIVATE|MAP_ANONYMOUS);
+	if (IS_ERR_VALUE(vma->vm_start)) {
 		DPRINT(("Cannot find unmapped area for size %ld\n", size));
 		up_write(&task->mm->mmap_sem);
 		goto error;
@@ -2366,15 +2336,12 @@ pfm_smpl_buffer_alloc(struct task_struct *task, struct file *filp, pfm_context_t
 		goto error;
 	}
 
-	get_file(filp);
-
 	/*
 	 * now insert the vma in the vm list for the process, must be
 	 * done with mmap lock held
 	 */
 	insert_vm_struct(mm, vma);
 
-	mm->total_vm  += size >> PAGE_SHIFT;
 	vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file,
 							vma_pages(vma));
 	up_write(&task->mm->mmap_sem);
@@ -2401,22 +2368,33 @@ error_kmem:
 static int
 pfm_bad_permissions(struct task_struct *task)
 {
+	const struct cred *tcred;
+	kuid_t uid = current_uid();
+	kgid_t gid = current_gid();
+	int ret;
+
+	rcu_read_lock();
+	tcred = __task_cred(task);
+
 	/* inspired by ptrace_attach() */
 	DPRINT(("cur: uid=%d gid=%d task: euid=%d suid=%d uid=%d egid=%d sgid=%d\n",
-		current->uid,
-		current->gid,
-		task->euid,
-		task->suid,
-		task->uid,
-		task->egid,
-		task->sgid));
-
-	return ((current->uid != task->euid)
-	    || (current->uid != task->suid)
-	    || (current->uid != task->uid)
-	    || (current->gid != task->egid)
-	    || (current->gid != task->sgid)
-	    || (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE);
+		from_kuid(&init_user_ns, uid),
+		from_kgid(&init_user_ns, gid),
+		from_kuid(&init_user_ns, tcred->euid),
+		from_kuid(&init_user_ns, tcred->suid),
+		from_kuid(&init_user_ns, tcred->uid),
+		from_kgid(&init_user_ns, tcred->egid),
+		from_kgid(&init_user_ns, tcred->sgid)));
+
+	ret = ((!uid_eq(uid, tcred->euid))
+	       || (!uid_eq(uid, tcred->suid))
+	       || (!uid_eq(uid, tcred->uid))
+	       || (!gid_eq(gid, tcred->egid))
+	       || (!gid_eq(gid, tcred->sgid))
+	       || (!gid_eq(gid, tcred->gid))) && !capable(CAP_SYS_PTRACE);
+
+	rcu_read_unlock();
+	return ret;
 }
 
 static int
@@ -2475,6 +2453,7 @@ pfm_setup_buffer_fmt(struct task_struct *task, struct file *filp, pfm_context_t
 
 	/* link buffer format and context */
 	ctx->ctx_buf_fmt = fmt;
+	ctx->ctx_fl_is_sampling = 1; /* assume record() is defined */
 
 	/*
 	 * check if buffer format wants to use perfmon buffer allocation/mapping service
@@ -2623,7 +2602,7 @@ pfm_task_incompatible(pfm_context_t *ctx, struct task_struct *task)
 	/*
 	 * make sure the task is off any CPU
 	 */
-	wait_task_inactive(task);
+	wait_task_inactive(task, 0);
 
 	/* more to come... */
 
@@ -2669,79 +2648,46 @@ pfm_context_create(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg
 {
 	pfarg_context_t *req = (pfarg_context_t *)arg;
 	struct file *filp;
+	struct path path;
 	int ctx_flags;
+	int fd;
 	int ret;
 
 	/* let's check the arguments first */
 	ret = pfarg_is_sane(current, req);
-	if (ret < 0) return ret;
+	if (ret < 0)
+		return ret;
 
 	ctx_flags = req->ctx_flags;
 
 	ret = -ENOMEM;
 
-	ctx = pfm_context_alloc();
-	if (!ctx) goto error;
+	fd = get_unused_fd();
+	if (fd < 0)
+		return fd;
 
-	ret = pfm_alloc_fd(&filp);
-	if (ret < 0) goto error_file;
+	ctx = pfm_context_alloc(ctx_flags);
+	if (!ctx)
+		goto error;
 
-	req->ctx_fd = ctx->ctx_fd = ret;
+	filp = pfm_alloc_file(ctx);
+	if (IS_ERR(filp)) {
+		ret = PTR_ERR(filp);
+		goto error_file;
+	}
 
-	/*
-	 * attach context to file
-	 */
-	filp->private_data = ctx;
+	req->ctx_fd = ctx->ctx_fd = fd;
 
 	/*
 	 * does the user want to sample?
 	 */
 	if (pfm_uuid_cmp(req->ctx_smpl_buf_id, pfm_null_uuid)) {
 		ret = pfm_setup_buffer_fmt(current, filp, ctx, ctx_flags, 0, req);
-		if (ret) goto buffer_error;
+		if (ret)
+			goto buffer_error;
 	}
 
-	/*
-	 * init context protection lock
-	 */
-	spin_lock_init(&ctx->ctx_lock);
-
-	/*
-	 * context is unloaded
-	 */
-	ctx->ctx_state = PFM_CTX_UNLOADED;
-
-	/*
-	 * initialization of context's flags
-	 */
-	ctx->ctx_fl_block       = (ctx_flags & PFM_FL_NOTIFY_BLOCK) ? 1 : 0;
-	ctx->ctx_fl_system      = (ctx_flags & PFM_FL_SYSTEM_WIDE) ? 1: 0;
-	ctx->ctx_fl_is_sampling = ctx->ctx_buf_fmt ? 1 : 0; /* assume record() is defined */
-	ctx->ctx_fl_no_msg      = (ctx_flags & PFM_FL_OVFL_NO_MSG) ? 1: 0;
-	/*
-	 * will move to set properties
-	 * ctx->ctx_fl_excl_idle   = (ctx_flags & PFM_FL_EXCL_IDLE) ? 1: 0;
-	 */
-
-	/*
-	 * init restart semaphore to locked
-	 */
-	init_completion(&ctx->ctx_restart_done);
-
-	/*
-	 * activation is used in SMP only
-	 */
-	ctx->ctx_last_activation = PFM_INVALID_ACTIVATION;
-	SET_LAST_CPU(ctx, -1);
-
-	/*
-	 * initialize notification message queue
-	 */
-	ctx->ctx_msgq_head = ctx->ctx_msgq_tail = 0;
-	init_waitqueue_head(&ctx->ctx_msgq_wait);
-	init_waitqueue_head(&ctx->ctx_zombieq);
-
-	DPRINT(("ctx=%p flags=0x%x system=%d notify_block=%d excl_idle=%d no_msg=%d ctx_fd=%d \n",
+	DPRINT(("ctx=%p flags=0x%x system=%d notify_block=%d excl_idle=%d no_msg=%d ctx_fd=%d\n",
 		ctx,
 		ctx_flags,
 		ctx->ctx_fl_system,
@@ -2755,10 +2701,14 @@ pfm_context_create(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg
 	 */
 	pfm_reset_pmu_state(ctx);
 
+	fd_install(fd, filp);
+
 	return 0;
 
 buffer_error:
-	pfm_free_fd(ctx->ctx_fd, filp);
+	path = filp->f_path;
+	put_filp(filp);
+	path_put(&path);
 
 	if (ctx->ctx_buf_fmt) {
 		pfm_buf_fmt_exit(ctx->ctx_buf_fmt, current, NULL, regs);
@@ -2767,6 +2717,7 @@ error_file:
 	pfm_context_free(ctx);
 
 error:
+	put_unused_fd(fd);
 	return ret;
 }
 
@@ -3541,7 +3492,7 @@ pfm_use_debug_registers(struct task_struct *task)
  * IA64_THREAD_DBG_VALID set. This indicates a task which was
  * able to use the debug registers for debugging purposes via
  * ptrace(). Therefore we know it was not using them for
- * perfmormance monitoring, so we only decrement the number
+ * performance monitoring, so we only decrement the number
  * of "ptraced" debug register users to keep the count up to date
  */
 int
@@ -3700,7 +3651,7 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 	 * "self-monitoring".
 	 */
 	if (CTX_OVFL_NOBLOCK(ctx) == 0 && state == PFM_CTX_MASKED) {
-		DPRINT(("unblocking [%d] \n", task_pid_nr(task)));
+		DPRINT(("unblocking [%d]\n", task_pid_nr(task)));
 		complete(&ctx->ctx_restart_done);
 	} else {
 		DPRINT(("[%d] armed exit trap\n", task_pid_nr(task)));
@@ -3709,7 +3660,7 @@ pfm_restart(pfm_context_t *ctx, void *arg, int count, struct pt_regs *regs)
 
 		PFM_SET_WORK_PENDING(task, 1);
 
-		tsk_set_notify_resume(task);
+		set_notify_resume(task);
 
 		/*
 		 * XXX: send reschedule if task runs on another CPU
@@ -4204,10 +4155,10 @@ pfm_check_task_exist(pfm_context_t *ctx)
 	do_each_thread (g, t) {
 		if (t->thread.pfm_context == ctx) {
 			ret = 0;
-			break;
+			goto out;
 		}
 	} while_each_thread (g, t);
-
+out:
 	read_unlock(&tasklist_lock);
 
 	DPRINT(("pfm_check_task_exist: ret=%d ctx=%p\n", ret, ctx));
@@ -4799,7 +4750,7 @@ recheck:
 
 		UNPROTECT_CTX(ctx, flags);
 
-		wait_task_inactive(task);
+		wait_task_inactive(task, 0);
 
 		PROTECT_CTX(ctx, flags);
 
@@ -4820,7 +4771,7 @@ recheck:
 asmlinkage long
 sys_perfmonctl (int fd, int cmd, void __user *arg, int count)
 {
-	struct file *file = NULL;
+	struct fd f = {NULL, 0};
 	pfm_context_t *ctx = NULL;
 	unsigned long flags = 0UL;
 	void *args_k = NULL;
@@ -4917,17 +4868,17 @@ restart_args:
 
 	ret = -EBADF;
 
-	file = fget(fd);
-	if (unlikely(file == NULL)) {
+	f = fdget(fd);
+	if (unlikely(f.file == NULL)) {
 		DPRINT(("invalid fd %d\n", fd));
 		goto error_args;
 	}
-	if (unlikely(PFM_IS_FILE(file) == 0)) {
+	if (unlikely(PFM_IS_FILE(f.file) == 0)) {
 		DPRINT(("fd %d not related to perfmon\n", fd));
 		goto error_args;
 	}
 
-	ctx = (pfm_context_t *)file->private_data;
+	ctx = f.file->private_data;
 	if (unlikely(ctx == NULL)) {
 		DPRINT(("no context for fd %d\n", fd));
 		goto error_args;
@@ -4957,8 +4908,8 @@ abort_locked:
 	if (call_made && PFM_CMD_RW_ARG(cmd) && copy_to_user(arg, args_k, base_sz*count)) ret = -EFAULT;
 
 error_args:
-	if (file)
-		fput(file);
+	if (f.file)
+		fdput(f);
 
 	kfree(args_k);
 
@@ -5038,12 +4989,13 @@ pfm_context_force_terminate(pfm_context_t *ctx, struct pt_regs *regs)
 }
 
 static int pfm_ovfl_notify_user(pfm_context_t *ctx, unsigned long ovfl_pmds);
+
  /*
   * pfm_handle_work() can be called with interrupts enabled
   * (TIF_NEED_RESCHED) or disabled. The down_interruptible
   * call may sleep, therefore we must re-enable interrupts
   * to avoid deadlocks. It is safe to do so because this function
-  * is called ONLY when returning to user level (PUStk=1), in which case
+  * is called ONLY when returning to user level (pUStk=1), in which case
   * there is no risk of kernel stack overflow due to deep
   * interrupt nesting.
   */
@@ -5059,7 +5011,8 @@ pfm_handle_work(void)
 
 	ctx = PFM_GET_CTX(current);
 	if (ctx == NULL) {
-		printk(KERN_ERR "perfmon: [%d] has no PFM context\n", task_pid_nr(current));
+		printk(KERN_ERR "perfmon: [%d] has no PFM context\n",
+			task_pid_nr(current));
 		return;
 	}
 
@@ -5067,8 +5020,6 @@ pfm_handle_work(void)
 
 	PFM_SET_WORK_PENDING(current, 0);
 
-	tsk_clear_notify_resume(current);
-
 	regs = task_pt_regs(current);
 
 	/*
@@ -5083,11 +5034,12 @@ pfm_handle_work(void)
 	/*
 	 * must be done before we check for simple-reset mode
 	 */
-	if (ctx->ctx_fl_going_zombie || ctx->ctx_state == PFM_CTX_ZOMBIE) goto do_zombie;
-
+	if (ctx->ctx_fl_going_zombie || ctx->ctx_state == PFM_CTX_ZOMBIE)
+		goto do_zombie;
 
 	//if (CTX_OVFL_NOBLOCK(ctx)) goto skip_blocking;
-	if (reason == PFM_TRAP_REASON_RESET) goto skip_blocking;
+	if (reason == PFM_TRAP_REASON_RESET)
+		goto skip_blocking;
 
 	/*
 	 * restore interrupt mask to what it was on entry.
@@ -5135,7 +5087,8 @@ do_zombie:
 	/*
 	 * in case of interruption of down() we don't restart anything
 	 */
-	if (ret < 0) goto nothing_to_do;
+	if (ret < 0)
+		goto nothing_to_do;
 
 skip_blocking:
 	pfm_resume_after_ovfl(ctx, ovfl_regs, regs);
@@ -5229,8 +5182,8 @@ pfm_end_notify_user(pfm_context_t *ctx)
  * main overflow processing routine.
  * it can be called from the interrupt path or explicitly during the context switch code
  */
-static void
-pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, struct pt_regs *regs)
+static void pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx,
+				unsigned long pmc0, struct pt_regs *regs)
 {
 	pfm_ovfl_arg_t *ovfl_arg;
 	unsigned long mask;
@@ -5435,7 +5388,7 @@ pfm_overflow_handler(struct task_struct *task, pfm_context_t *ctx, u64 pmc0, str
 			 * when coming from ctxsw, current still points to the
 			 * previous task, therefore we must work with task and not current.
 			 */
-			tsk_set_notify_resume(task);
+			set_notify_resume(task);
 		}
 		/*
 		 * defer until state is changed (shorten spin window). the context is locked
@@ -5511,7 +5464,7 @@ stop_monitoring:
 }
 
 static int
-pfm_do_interrupt_handler(int irq, void *arg, struct pt_regs *regs)
+pfm_do_interrupt_handler(void *arg, struct pt_regs *regs)
 {
 	struct task_struct *task;
 	pfm_context_t *ctx;
@@ -5591,7 +5544,7 @@ pfm_interrupt_handler(int irq, void *arg)
 
 		start_cycles = ia64_get_itc();
 
-		ret = pfm_do_interrupt_handler(irq, arg, regs);
+		ret = pfm_do_interrupt_handler(arg, regs);
 
 		total_cycles = ia64_get_itc();
 
@@ -5611,7 +5564,7 @@ pfm_interrupt_handler(int irq, void *arg)
 		(*pfm_alt_intr_handler->handler)(irq, arg, regs);
 	}
 
-	put_cpu_no_resched();
+	put_cpu();
 	return IRQ_HANDLED;
 }
 
@@ -5619,7 +5572,7 @@ pfm_interrupt_handler(int irq, void *arg)
  * /proc/perfmon interface, for debug only
  */
 
-#define PFM_PROC_SHOW_HEADER	((void *)NR_CPUS+1)
+#define PFM_PROC_SHOW_HEADER	((void *)(long)nr_cpu_ids+1)
 
 static void *
 pfm_proc_start(struct seq_file *m, loff_t *pos)
@@ -5628,7 +5581,7 @@ pfm_proc_start(struct seq_file *m, loff_t *pos)
 		return PFM_PROC_SHOW_HEADER;
 	}
 
-	while (*pos <= NR_CPUS) {
+	while (*pos <= nr_cpu_ids) {
 		if (cpu_online(*pos - 1)) {
 			return (void *)*pos;
 		}
@@ -5688,24 +5641,8 @@ pfm_proc_show_header(struct seq_file *m)
 
 	list_for_each(pos, &pfm_buffer_fmt_list) {
 		entry = list_entry(pos, pfm_buffer_fmt_t, fmt_list);
-		seq_printf(m, "format                    : %02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x %s\n",
-			entry->fmt_uuid[0],
-			entry->fmt_uuid[1],
-			entry->fmt_uuid[2],
-			entry->fmt_uuid[3],
-			entry->fmt_uuid[4],
-			entry->fmt_uuid[5],
-			entry->fmt_uuid[6],
-			entry->fmt_uuid[7],
-			entry->fmt_uuid[8],
-			entry->fmt_uuid[9],
-			entry->fmt_uuid[10],
-			entry->fmt_uuid[11],
-			entry->fmt_uuid[12],
-			entry->fmt_uuid[13],
-			entry->fmt_uuid[14],
-			entry->fmt_uuid[15],
-			entry->fmt_name);
+		seq_printf(m, "format                    : %16phD %s\n",
+			   entry->fmt_uuid, entry->fmt_name);
 	}
 	spin_unlock(&pfm_buffer_fmt_lock);
 
@@ -6450,7 +6387,6 @@ pfm_flush_pmds(struct task_struct *task, pfm_context_t *ctx)
 
 static struct irqaction perfmon_irqaction = {
 	.handler = pfm_interrupt_handler,
-	.flags   = IRQF_DISABLED,
 	.name    = "perfmon"
 };
 
@@ -6529,7 +6465,7 @@ pfm_install_alt_pmu_interrupt(pfm_intr_handler_desc_t *hdl)
 	}
 
 	/* save the current system wide pmu states */
-	ret = on_each_cpu(pfm_alt_save_pmu_state, NULL, 0, 1);
+	ret = on_each_cpu(pfm_alt_save_pmu_state, NULL, 1);
 	if (ret) {
 		DPRINT(("on_each_cpu() failed: %d\n", ret));
 		goto cleanup_reserve;
@@ -6574,7 +6510,7 @@ pfm_remove_alt_pmu_interrupt(pfm_intr_handler_desc_t *hdl)
 
 	pfm_alt_intr_handler = NULL;
 
-	ret = on_each_cpu(pfm_alt_restore_pmu_state, NULL, 0, 1);
+	ret = on_each_cpu(pfm_alt_restore_pmu_state, NULL, 1);
 	if (ret) {
 		DPRINT(("on_each_cpu() failed: %d\n", ret));
 	}
@@ -6695,16 +6631,12 @@ pfm_init(void)
 	/*
 	 * create /proc/perfmon (mostly for debugging purposes)
 	 */
- 	perfmon_dir = create_proc_entry("perfmon", S_IRUGO, NULL);
+	perfmon_dir = proc_create("perfmon", S_IRUGO, NULL, &pfm_proc_fops);
 	if (perfmon_dir == NULL) {
 		printk(KERN_ERR "perfmon: cannot create /proc entry, perfmon disabled\n");
 		pmu_conf = NULL;
 		return -1;
 	}
-  	/*
- 	 * install customized file operations for /proc/perfmon entry
- 	 */
- 	perfmon_dir->proc_fops = &pfm_proc_fops;
 
 	/*
 	 * create /proc/sys/kernel/perfmon (for debugging purposes)