[PATCH] proc: Rewrite the proc dentry flush on exit optimization
To keep the dcache from filling up with dead /proc entries we flush them on
process exit. However over the years that code has gotten hairy with a
dentry_pointer and a lock in task_struct and misdocumented as a correctness
feature.
I have rewritten this code to look and see if we have a corresponding entry in
the dcache and if so flush it on process exit. This removes the extra fields
in the task_struct and allows me to trivially handle the case of a
/proc/<tgid>/task/<pid> entry as well as the current /proc/<pid> entries.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/fs/exec.c b/fs/exec.c
index 0b88bf6..8c51960 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -666,8 +666,6 @@
* and to assume its PID:
*/
if (!thread_group_leader(current)) {
- struct dentry *proc_dentry1, *proc_dentry2;
-
/*
* Wait for the thread group leader to be a zombie.
* It should already be zombie at this point, most
@@ -689,10 +687,6 @@
*/
current->start_time = leader->start_time;
- spin_lock(&leader->proc_lock);
- spin_lock(¤t->proc_lock);
- proc_dentry1 = proc_pid_unhash(current);
- proc_dentry2 = proc_pid_unhash(leader);
write_lock_irq(&tasklist_lock);
BUG_ON(leader->tgid != current->tgid);
@@ -729,10 +723,6 @@
leader->exit_state = EXIT_DEAD;
write_unlock_irq(&tasklist_lock);
- spin_unlock(&leader->proc_lock);
- spin_unlock(¤t->proc_lock);
- proc_pid_flush(proc_dentry1);
- proc_pid_flush(proc_dentry2);
}
/*
diff --git a/fs/proc/base.c b/fs/proc/base.c
index c863684..f435932 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1352,16 +1352,6 @@
return 0;
}
-static void pid_base_iput(struct dentry *dentry, struct inode *inode)
-{
- struct task_struct *task = proc_task(inode);
- spin_lock(&task->proc_lock);
- if (task->proc_dentry == dentry)
- task->proc_dentry = NULL;
- spin_unlock(&task->proc_lock);
- iput(inode);
-}
-
static int pid_delete_dentry(struct dentry * dentry)
{
/* Is the task we represent dead?
@@ -1383,13 +1373,6 @@
.d_delete = pid_delete_dentry,
};
-static struct dentry_operations pid_base_dentry_operations =
-{
- .d_revalidate = pid_revalidate,
- .d_iput = pid_base_iput,
- .d_delete = pid_delete_dentry,
-};
-
/* Lookups */
static unsigned name_to_int(struct dentry *dentry)
@@ -1859,57 +1842,70 @@
};
/**
- * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
- * @p: task that should be flushed.
+ * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
*
- * Drops the /proc/@pid dcache entry from the hash chains.
+ * @task: task that should be flushed.
*
- * Dropping /proc/@pid entries and detach_pid must be synchroneous,
- * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
- * if the pid value is immediately reused. This is enforced by
- * - caller must acquire spin_lock(p->proc_lock)
- * - must be called before detach_pid()
- * - proc_pid_lookup acquires proc_lock, and checks that
- * the target is not dead by looking at the attach count
- * of PIDTYPE_PID.
+ * Looks in the dcache for
+ * /proc/@pid
+ * /proc/@tgid/task/@pid
+ * if either directory is present flushes it and all of it'ts children
+ * from the dcache.
+ *
+ * It is safe and reasonable to cache /proc entries for a task until
+ * that task exits. After that they just clog up the dcache with
+ * useless entries, possibly causing useful dcache entries to be
+ * flushed instead. This routine is proved to flush those useless
+ * dcache entries at process exit time.
+ *
+ * NOTE: This routine is just an optimization so it does not guarantee
+ * that no dcache entries will exist at process exit time it
+ * just makes it very unlikely that any will persist.
*/
-
-struct dentry *proc_pid_unhash(struct task_struct *p)
+void proc_flush_task(struct task_struct *task)
{
- struct dentry *proc_dentry;
+ struct dentry *dentry, *leader, *dir;
+ char buf[30];
+ struct qstr name;
- proc_dentry = p->proc_dentry;
- if (proc_dentry != NULL) {
-
- spin_lock(&dcache_lock);
- spin_lock(&proc_dentry->d_lock);
- if (!d_unhashed(proc_dentry)) {
- dget_locked(proc_dentry);
- __d_drop(proc_dentry);
- spin_unlock(&proc_dentry->d_lock);
- } else {
- spin_unlock(&proc_dentry->d_lock);
- proc_dentry = NULL;
- }
- spin_unlock(&dcache_lock);
+ name.name = buf;
+ name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
+ dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
+ if (dentry) {
+ shrink_dcache_parent(dentry);
+ d_drop(dentry);
+ dput(dentry);
}
- return proc_dentry;
-}
-/**
- * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
- * @proc_dentry: directoy to prune.
- *
- * Shrink the /proc directory that was used by the just killed thread.
- */
-
-void proc_pid_flush(struct dentry *proc_dentry)
-{
- might_sleep();
- if(proc_dentry != NULL) {
- shrink_dcache_parent(proc_dentry);
- dput(proc_dentry);
+ if (thread_group_leader(task))
+ goto out;
+
+ name.name = buf;
+ name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
+ leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
+ if (!leader)
+ goto out;
+
+ name.name = "task";
+ name.len = strlen(name.name);
+ dir = d_hash_and_lookup(leader, &name);
+ if (!dir)
+ goto out_put_leader;
+
+ name.name = buf;
+ name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
+ dentry = d_hash_and_lookup(dir, &name);
+ if (dentry) {
+ shrink_dcache_parent(dentry);
+ d_drop(dentry);
+ dput(dentry);
}
+
+ dput(dir);
+out_put_leader:
+ dput(leader);
+out:
+ return;
}
/* SMP-safe */
@@ -1919,7 +1915,6 @@
struct inode *inode;
struct proc_inode *ei;
unsigned tgid;
- int died;
if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
inode = new_inode(dir->i_sb);
@@ -1965,23 +1960,16 @@
inode->i_nlink = 4;
#endif
- dentry->d_op = &pid_base_dentry_operations;
+ dentry->d_op = &pid_dentry_operations;
- died = 0;
d_add(dentry, inode);
- spin_lock(&task->proc_lock);
- task->proc_dentry = dentry;
if (!pid_alive(task)) {
- dentry = proc_pid_unhash(task);
- died = 1;
- }
- spin_unlock(&task->proc_lock);
-
- put_task_struct(task);
- if (died) {
- proc_pid_flush(dentry);
+ d_drop(dentry);
+ shrink_dcache_parent(dentry);
goto out;
}
+
+ put_task_struct(task);
return NULL;
out:
return ERR_PTR(-ENOENT);
@@ -2024,7 +2012,7 @@
inode->i_nlink = 3;
#endif
- dentry->d_op = &pid_base_dentry_operations;
+ dentry->d_op = &pid_dentry_operations;
d_add(dentry, inode);
diff --git a/include/linux/init_task.h b/include/linux/init_task.h
index 41ecbb8..e127ef7 100644
--- a/include/linux/init_task.h
+++ b/include/linux/init_task.h
@@ -119,7 +119,6 @@
.signal = {{0}}}, \
.blocked = {{0}}, \
.alloc_lock = SPIN_LOCK_UNLOCKED, \
- .proc_lock = SPIN_LOCK_UNLOCKED, \
.journal_info = NULL, \
.cpu_timers = INIT_CPU_TIMERS(tsk.cpu_timers), \
.fs_excl = ATOMIC_INIT(0), \
diff --git a/include/linux/proc_fs.h b/include/linux/proc_fs.h
index 9dd8488..d4d2081 100644
--- a/include/linux/proc_fs.h
+++ b/include/linux/proc_fs.h
@@ -99,9 +99,8 @@
struct mm_struct;
+void proc_flush_task(struct task_struct *task);
struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *);
-struct dentry *proc_pid_unhash(struct task_struct *p);
-void proc_pid_flush(struct dentry *proc_dentry);
int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir);
unsigned long task_vsize(struct mm_struct *);
int task_statm(struct mm_struct *, int *, int *, int *, int *);
@@ -211,8 +210,7 @@
#define proc_net_create(name, mode, info) ({ (void)(mode), NULL; })
static inline void proc_net_remove(const char *name) {}
-static inline struct dentry *proc_pid_unhash(struct task_struct *p) { return NULL; }
-static inline void proc_pid_flush(struct dentry *proc_dentry) { }
+static inline void proc_flush_task(struct task_struct *task) { }
static inline struct proc_dir_entry *create_proc_entry(const char *name,
mode_t mode, struct proc_dir_entry *parent) { return NULL; }
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 8d11d93..122a25c 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -842,8 +842,6 @@
u32 self_exec_id;
/* Protection of (de-)allocation: mm, files, fs, tty, keyrings */
spinlock_t alloc_lock;
-/* Protection of proc_dentry: nesting proc_lock, dcache_lock, write_lock_irq(&tasklist_lock); */
- spinlock_t proc_lock;
#ifdef CONFIG_DEBUG_MUTEXES
/* mutex deadlock detection */
@@ -856,7 +854,6 @@
/* VM state */
struct reclaim_state *reclaim_state;
- struct dentry *proc_dentry;
struct backing_dev_info *backing_dev_info;
struct io_context *io_context;
diff --git a/kernel/exit.c b/kernel/exit.c
index e76bd02..304ef63 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -137,12 +137,8 @@
{
int zap_leader;
task_t *leader;
- struct dentry *proc_dentry;
-
repeat:
atomic_dec(&p->user->processes);
- spin_lock(&p->proc_lock);
- proc_dentry = proc_pid_unhash(p);
write_lock_irq(&tasklist_lock);
ptrace_unlink(p);
BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
@@ -171,8 +167,7 @@
sched_exit(p);
write_unlock_irq(&tasklist_lock);
- spin_unlock(&p->proc_lock);
- proc_pid_flush(proc_dentry);
+ proc_flush_task(p);
release_thread(p);
call_rcu(&p->rcu, delayed_put_task_struct);
diff --git a/kernel/fork.c b/kernel/fork.c
index dfd10cb..79e9104 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -993,13 +993,10 @@
if (put_user(p->pid, parent_tidptr))
goto bad_fork_cleanup;
- p->proc_dentry = NULL;
-
INIT_LIST_HEAD(&p->children);
INIT_LIST_HEAD(&p->sibling);
p->vfork_done = NULL;
spin_lock_init(&p->alloc_lock);
- spin_lock_init(&p->proc_lock);
clear_tsk_thread_flag(p, TIF_SIGPENDING);
init_sigpending(&p->pending);