summaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2014-08-11 11:44:11 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2014-08-11 11:44:11 -0700
commitf6f993328b2abcab86a3c99d7bd9f2066ab03d36 (patch)
treeea6f3902a0fa546493731b3b52a31d98cc747a90 /kernel
parentc7a19c795b4b0a3232c157ed29eea85077e95da6 (diff)
parent12a5b5294cb1896e9a3c9fca8ff5a7e3def4e8c6 (diff)
downloadlinux-f6f993328b2abcab86a3c99d7bd9f2066ab03d36.tar.bz2
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Pull vfs updates from Al Viro: "Stuff in here: - acct.c fixes and general rework of mnt_pin mechanism. That allows to go for delayed-mntput stuff, which will permit mntput() on deep stack without worrying about stack overflows - fs shutdown will happen on shallow stack. IOW, we can do Eric's umount-on-rmdir series without introducing tons of stack overflows on new mntput() call chains it introduces. - Bruce's d_splice_alias() patches - more Miklos' rename() stuff. - a couple of regression fixes (stable fodder, in the end of branch) and a fix for API idiocy in iov_iter.c. There definitely will be another pile, maybe even two. I'd like to get Eric's series in this time, but even if we miss it, it'll go right in the beginning of for-next in the next cycle - the tricky part of prereqs is in this pile" * 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (40 commits) fix copy_tree() regression __generic_file_write_iter(): fix handling of sync error after DIO switch iov_iter_get_pages() to passing maximal number of pages fs: mark __d_obtain_alias static dcache: d_splice_alias should detect loops exportfs: update Exporting documentation dcache: d_find_alias needn't recheck IS_ROOT && DCACHE_DISCONNECTED dcache: remove unused d_find_alias parameter dcache: d_obtain_alias callers don't all want DISCONNECTED dcache: d_splice_alias should ignore DCACHE_DISCONNECTED dcache: d_splice_alias mustn't create directory aliases dcache: close d_move race in d_splice_alias dcache: move d_splice_alias namei: trivial fix to vfs_rename_dir comment VFS: allow ->d_manage() to declare -EISDIR in rcu_walk mode. cifs: support RENAME_NOREPLACE hostfs: support rename flags shmem: support RENAME_EXCHANGE shmem: support RENAME_NOREPLACE btrfs: add RENAME_NOREPLACE ...
Diffstat (limited to 'kernel')
-rw-r--r--kernel/acct.c456
1 files changed, 201 insertions, 255 deletions
diff --git a/kernel/acct.c b/kernel/acct.c
index 51793520566f..b4c667d22e79 100644
--- a/kernel/acct.c
+++ b/kernel/acct.c
@@ -59,6 +59,7 @@
#include <asm/div64.h>
#include <linux/blkdev.h> /* sector_div */
#include <linux/pid_namespace.h>
+#include <linux/fs_pin.h>
/*
* These constants control the amount of freespace that suspend and
@@ -75,172 +76,190 @@ int acct_parm[3] = {4, 2, 30};
/*
* External references and all of the globals.
*/
-static void do_acct_process(struct bsd_acct_struct *acct,
- struct pid_namespace *ns, struct file *);
+static void do_acct_process(struct bsd_acct_struct *acct);
-/*
- * This structure is used so that all the data protected by lock
- * can be placed in the same cache line as the lock. This primes
- * the cache line to have the data after getting the lock.
- */
struct bsd_acct_struct {
+ struct fs_pin pin;
+ struct mutex lock;
int active;
unsigned long needcheck;
struct file *file;
struct pid_namespace *ns;
- struct list_head list;
+ struct work_struct work;
+ struct completion done;
};
-static DEFINE_SPINLOCK(acct_lock);
-static LIST_HEAD(acct_list);
-
/*
* Check the amount of free space and suspend/resume accordingly.
*/
-static int check_free_space(struct bsd_acct_struct *acct, struct file *file)
+static int check_free_space(struct bsd_acct_struct *acct)
{
struct kstatfs sbuf;
- int res;
- int act;
- u64 resume;
- u64 suspend;
-
- spin_lock(&acct_lock);
- res = acct->active;
- if (!file || time_is_before_jiffies(acct->needcheck))
+
+ if (time_is_before_jiffies(acct->needcheck))
goto out;
- spin_unlock(&acct_lock);
/* May block */
- if (vfs_statfs(&file->f_path, &sbuf))
- return res;
- suspend = sbuf.f_blocks * SUSPEND;
- resume = sbuf.f_blocks * RESUME;
-
- do_div(suspend, 100);
- do_div(resume, 100);
-
- if (sbuf.f_bavail <= suspend)
- act = -1;
- else if (sbuf.f_bavail >= resume)
- act = 1;
- else
- act = 0;
-
- /*
- * If some joker switched acct->file under us we'ld better be
- * silent and _not_ touch anything.
- */
- spin_lock(&acct_lock);
- if (file != acct->file) {
- if (act)
- res = act > 0;
+ if (vfs_statfs(&acct->file->f_path, &sbuf))
goto out;
- }
if (acct->active) {
- if (act < 0) {
+ u64 suspend = sbuf.f_blocks * SUSPEND;
+ do_div(suspend, 100);
+ if (sbuf.f_bavail <= suspend) {
acct->active = 0;
pr_info("Process accounting paused\n");
}
} else {
- if (act > 0) {
+ u64 resume = sbuf.f_blocks * RESUME;
+ do_div(resume, 100);
+ if (sbuf.f_bavail >= resume) {
acct->active = 1;
pr_info("Process accounting resumed\n");
}
}
acct->needcheck = jiffies + ACCT_TIMEOUT*HZ;
- res = acct->active;
out:
- spin_unlock(&acct_lock);
+ return acct->active;
+}
+
+static struct bsd_acct_struct *acct_get(struct pid_namespace *ns)
+{
+ struct bsd_acct_struct *res;
+again:
+ smp_rmb();
+ rcu_read_lock();
+ res = ACCESS_ONCE(ns->bacct);
+ if (!res) {
+ rcu_read_unlock();
+ return NULL;
+ }
+ if (!atomic_long_inc_not_zero(&res->pin.count)) {
+ rcu_read_unlock();
+ cpu_relax();
+ goto again;
+ }
+ rcu_read_unlock();
+ mutex_lock(&res->lock);
+ if (!res->ns) {
+ mutex_unlock(&res->lock);
+ pin_put(&res->pin);
+ goto again;
+ }
return res;
}
-/*
- * Close the old accounting file (if currently open) and then replace
- * it with file (if non-NULL).
- *
- * NOTE: acct_lock MUST be held on entry and exit.
- */
-static void acct_file_reopen(struct bsd_acct_struct *acct, struct file *file,
- struct pid_namespace *ns)
+static void close_work(struct work_struct *work)
{
- struct file *old_acct = NULL;
- struct pid_namespace *old_ns = NULL;
-
- if (acct->file) {
- old_acct = acct->file;
- old_ns = acct->ns;
- acct->active = 0;
- acct->file = NULL;
+ struct bsd_acct_struct *acct = container_of(work, struct bsd_acct_struct, work);
+ struct file *file = acct->file;
+ if (file->f_op->flush)
+ file->f_op->flush(file, NULL);
+ __fput_sync(file);
+ complete(&acct->done);
+}
+
+static void acct_kill(struct bsd_acct_struct *acct,
+ struct bsd_acct_struct *new)
+{
+ if (acct) {
+ struct pid_namespace *ns = acct->ns;
+ do_acct_process(acct);
+ INIT_WORK(&acct->work, close_work);
+ init_completion(&acct->done);
+ schedule_work(&acct->work);
+ wait_for_completion(&acct->done);
+ pin_remove(&acct->pin);
+ ns->bacct = new;
acct->ns = NULL;
- list_del(&acct->list);
- }
- if (file) {
- acct->file = file;
- acct->ns = ns;
- acct->needcheck = jiffies + ACCT_TIMEOUT*HZ;
- acct->active = 1;
- list_add(&acct->list, &acct_list);
+ atomic_long_dec(&acct->pin.count);
+ mutex_unlock(&acct->lock);
+ pin_put(&acct->pin);
}
- if (old_acct) {
- mnt_unpin(old_acct->f_path.mnt);
- spin_unlock(&acct_lock);
- do_acct_process(acct, old_ns, old_acct);
- filp_close(old_acct, NULL);
- spin_lock(&acct_lock);
+}
+
+static void acct_pin_kill(struct fs_pin *pin)
+{
+ struct bsd_acct_struct *acct;
+ acct = container_of(pin, struct bsd_acct_struct, pin);
+ mutex_lock(&acct->lock);
+ if (!acct->ns) {
+ mutex_unlock(&acct->lock);
+ pin_put(pin);
+ acct = NULL;
}
+ acct_kill(acct, NULL);
}
static int acct_on(struct filename *pathname)
{
struct file *file;
- struct vfsmount *mnt;
- struct pid_namespace *ns;
- struct bsd_acct_struct *acct = NULL;
+ struct vfsmount *mnt, *internal;
+ struct pid_namespace *ns = task_active_pid_ns(current);
+ struct bsd_acct_struct *acct, *old;
+ int err;
+
+ acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
+ if (!acct)
+ return -ENOMEM;
/* Difference from BSD - they don't do O_APPEND */
file = file_open_name(pathname, O_WRONLY|O_APPEND|O_LARGEFILE, 0);
- if (IS_ERR(file))
+ if (IS_ERR(file)) {
+ kfree(acct);
return PTR_ERR(file);
+ }
if (!S_ISREG(file_inode(file)->i_mode)) {
+ kfree(acct);
filp_close(file, NULL);
return -EACCES;
}
if (!file->f_op->write) {
+ kfree(acct);
filp_close(file, NULL);
return -EIO;
}
-
- ns = task_active_pid_ns(current);
- if (ns->bacct == NULL) {
- acct = kzalloc(sizeof(struct bsd_acct_struct), GFP_KERNEL);
- if (acct == NULL) {
- filp_close(file, NULL);
- return -ENOMEM;
- }
+ internal = mnt_clone_internal(&file->f_path);
+ if (IS_ERR(internal)) {
+ kfree(acct);
+ filp_close(file, NULL);
+ return PTR_ERR(internal);
}
-
- spin_lock(&acct_lock);
- if (ns->bacct == NULL) {
- ns->bacct = acct;
- acct = NULL;
+ err = mnt_want_write(internal);
+ if (err) {
+ mntput(internal);
+ kfree(acct);
+ filp_close(file, NULL);
+ return err;
}
-
mnt = file->f_path.mnt;
- mnt_pin(mnt);
- acct_file_reopen(ns->bacct, file, ns);
- spin_unlock(&acct_lock);
-
- mntput(mnt); /* it's pinned, now give up active reference */
- kfree(acct);
-
+ file->f_path.mnt = internal;
+
+ atomic_long_set(&acct->pin.count, 1);
+ acct->pin.kill = acct_pin_kill;
+ acct->file = file;
+ acct->needcheck = jiffies;
+ acct->ns = ns;
+ mutex_init(&acct->lock);
+ mutex_lock_nested(&acct->lock, 1); /* nobody has seen it yet */
+ pin_insert(&acct->pin, mnt);
+
+ old = acct_get(ns);
+ if (old)
+ acct_kill(old, acct);
+ else
+ ns->bacct = acct;
+ mutex_unlock(&acct->lock);
+ mnt_drop_write(mnt);
+ mntput(mnt);
return 0;
}
+static DEFINE_MUTEX(acct_on_mutex);
+
/**
* sys_acct - enable/disable process accounting
* @name: file name for accounting records or NULL to shutdown accounting
@@ -264,78 +283,20 @@ SYSCALL_DEFINE1(acct, const char __user *, name)
if (IS_ERR(tmp))
return PTR_ERR(tmp);
+ mutex_lock(&acct_on_mutex);
error = acct_on(tmp);
+ mutex_unlock(&acct_on_mutex);
putname(tmp);
} else {
- struct bsd_acct_struct *acct;
-
- acct = task_active_pid_ns(current)->bacct;
- if (acct == NULL)
- return 0;
-
- spin_lock(&acct_lock);
- acct_file_reopen(acct, NULL, NULL);
- spin_unlock(&acct_lock);
+ acct_kill(acct_get(task_active_pid_ns(current)), NULL);
}
return error;
}
-/**
- * acct_auto_close - turn off a filesystem's accounting if it is on
- * @m: vfsmount being shut down
- *
- * If the accounting is turned on for a file in the subtree pointed to
- * to by m, turn accounting off. Done when m is about to die.
- */
-void acct_auto_close_mnt(struct vfsmount *m)
-{
- struct bsd_acct_struct *acct;
-
- spin_lock(&acct_lock);
-restart:
- list_for_each_entry(acct, &acct_list, list)
- if (acct->file && acct->file->f_path.mnt == m) {
- acct_file_reopen(acct, NULL, NULL);
- goto restart;
- }
- spin_unlock(&acct_lock);
-}
-
-/**
- * acct_auto_close - turn off a filesystem's accounting if it is on
- * @sb: super block for the filesystem
- *
- * If the accounting is turned on for a file in the filesystem pointed
- * to by sb, turn accounting off.
- */
-void acct_auto_close(struct super_block *sb)
-{
- struct bsd_acct_struct *acct;
-
- spin_lock(&acct_lock);
-restart:
- list_for_each_entry(acct, &acct_list, list)
- if (acct->file && acct->file->f_path.dentry->d_sb == sb) {
- acct_file_reopen(acct, NULL, NULL);
- goto restart;
- }
- spin_unlock(&acct_lock);
-}
-
void acct_exit_ns(struct pid_namespace *ns)
{
- struct bsd_acct_struct *acct = ns->bacct;
-
- if (acct == NULL)
- return;
-
- spin_lock(&acct_lock);
- if (acct->file != NULL)
- acct_file_reopen(acct, NULL, NULL);
- spin_unlock(&acct_lock);
-
- kfree(acct);
+ acct_kill(acct_get(ns), NULL);
}
/*
@@ -450,38 +411,20 @@ static u32 encode_float(u64 value)
* do_exit() or when switching to a different output file.
*/
-/*
- * do_acct_process does all actual work. Caller holds the reference to file.
- */
-static void do_acct_process(struct bsd_acct_struct *acct,
- struct pid_namespace *ns, struct file *file)
+static void fill_ac(acct_t *ac)
{
struct pacct_struct *pacct = &current->signal->pacct;
- acct_t ac;
- mm_segment_t fs;
- unsigned long flim;
u64 elapsed, run_time;
struct tty_struct *tty;
- const struct cred *orig_cred;
-
- /* Perform file operations on behalf of whoever enabled accounting */
- orig_cred = override_creds(file->f_cred);
-
- /*
- * First check to see if there is enough free_space to continue
- * the process accounting system.
- */
- if (!check_free_space(acct, file))
- goto out;
/*
* Fill the accounting struct with the needed info as recorded
* by the different kernel functions.
*/
- memset(&ac, 0, sizeof(acct_t));
+ memset(ac, 0, sizeof(acct_t));
- ac.ac_version = ACCT_VERSION | ACCT_BYTEORDER;
- strlcpy(ac.ac_comm, current->comm, sizeof(ac.ac_comm));
+ ac->ac_version = ACCT_VERSION | ACCT_BYTEORDER;
+ strlcpy(ac->ac_comm, current->comm, sizeof(ac->ac_comm));
/* calculate run_time in nsec*/
run_time = ktime_get_ns();
@@ -489,9 +432,9 @@ static void do_acct_process(struct bsd_acct_struct *acct,
/* convert nsec -> AHZ */
elapsed = nsec_to_AHZ(run_time);
#if ACCT_VERSION == 3
- ac.ac_etime = encode_float(elapsed);
+ ac->ac_etime = encode_float(elapsed);
#else
- ac.ac_etime = encode_comp_t(elapsed < (unsigned long) -1l ?
+ ac->ac_etime = encode_comp_t(elapsed < (unsigned long) -1l ?
(unsigned long) elapsed : (unsigned long) -1l);
#endif
#if ACCT_VERSION == 1 || ACCT_VERSION == 2
@@ -499,18 +442,58 @@ static void do_acct_process(struct bsd_acct_struct *acct,
/* new enlarged etime field */
comp2_t etime = encode_comp2_t(elapsed);
- ac.ac_etime_hi = etime >> 16;
- ac.ac_etime_lo = (u16) etime;
+ ac->ac_etime_hi = etime >> 16;
+ ac->ac_etime_lo = (u16) etime;
}
#endif
do_div(elapsed, AHZ);
- ac.ac_btime = get_seconds() - elapsed;
+ ac->ac_btime = get_seconds() - elapsed;
+#if ACCT_VERSION==2
+ ac->ac_ahz = AHZ;
+#endif
+
+ spin_lock_irq(&current->sighand->siglock);
+ tty = current->signal->tty; /* Safe as we hold the siglock */
+ ac->ac_tty = tty ? old_encode_dev(tty_devnum(tty)) : 0;
+ ac->ac_utime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_utime)));
+ ac->ac_stime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_stime)));
+ ac->ac_flag = pacct->ac_flag;
+ ac->ac_mem = encode_comp_t(pacct->ac_mem);
+ ac->ac_minflt = encode_comp_t(pacct->ac_minflt);
+ ac->ac_majflt = encode_comp_t(pacct->ac_majflt);
+ ac->ac_exitcode = pacct->ac_exitcode;
+ spin_unlock_irq(&current->sighand->siglock);
+}
+/*
+ * do_acct_process does all actual work. Caller holds the reference to file.
+ */
+static void do_acct_process(struct bsd_acct_struct *acct)
+{
+ acct_t ac;
+ unsigned long flim;
+ const struct cred *orig_cred;
+ struct pid_namespace *ns = acct->ns;
+ struct file *file = acct->file;
+
+ /*
+ * Accounting records are not subject to resource limits.
+ */
+ flim = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
+ current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
+ /* Perform file operations on behalf of whoever enabled accounting */
+ orig_cred = override_creds(file->f_cred);
+
+ /*
+ * First check to see if there is enough free_space to continue
+ * the process accounting system.
+ */
+ if (!check_free_space(acct))
+ goto out;
+
+ fill_ac(&ac);
/* we really need to bite the bullet and change layout */
ac.ac_uid = from_kuid_munged(file->f_cred->user_ns, orig_cred->uid);
ac.ac_gid = from_kgid_munged(file->f_cred->user_ns, orig_cred->gid);
-#if ACCT_VERSION == 2
- ac.ac_ahz = AHZ;
-#endif
#if ACCT_VERSION == 1 || ACCT_VERSION == 2
/* backward-compatible 16 bit fields */
ac.ac_uid16 = ac.ac_uid;
@@ -522,45 +505,18 @@ static void do_acct_process(struct bsd_acct_struct *acct,
ac.ac_ppid = task_tgid_nr_ns(rcu_dereference(current->real_parent), ns);
rcu_read_unlock();
#endif
-
- spin_lock_irq(&current->sighand->siglock);
- tty = current->signal->tty; /* Safe as we hold the siglock */
- ac.ac_tty = tty ? old_encode_dev(tty_devnum(tty)) : 0;
- ac.ac_utime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_utime)));
- ac.ac_stime = encode_comp_t(jiffies_to_AHZ(cputime_to_jiffies(pacct->ac_stime)));
- ac.ac_flag = pacct->ac_flag;
- ac.ac_mem = encode_comp_t(pacct->ac_mem);
- ac.ac_minflt = encode_comp_t(pacct->ac_minflt);
- ac.ac_majflt = encode_comp_t(pacct->ac_majflt);
- ac.ac_exitcode = pacct->ac_exitcode;
- spin_unlock_irq(&current->sighand->siglock);
- ac.ac_io = encode_comp_t(0 /* current->io_usage */); /* %% */
- ac.ac_rw = encode_comp_t(ac.ac_io / 1024);
- ac.ac_swaps = encode_comp_t(0);
-
/*
* Get freeze protection. If the fs is frozen, just skip the write
* as we could deadlock the system otherwise.
*/
- if (!file_start_write_trylock(file))
- goto out;
- /*
- * Kernel segment override to datasegment and write it
- * to the accounting file.
- */
- fs = get_fs();
- set_fs(KERNEL_DS);
- /*
- * Accounting records are not subject to resource limits.
- */
- flim = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
- current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
- file->f_op->write(file, (char *)&ac,
- sizeof(acct_t), &file->f_pos);
- current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim;
- set_fs(fs);
- file_end_write(file);
+ if (file_start_write_trylock(file)) {
+ /* it's been opened O_APPEND, so position is irrelevant */
+ loff_t pos = 0;
+ __kernel_write(file, (char *)&ac, sizeof(acct_t), &pos);
+ file_end_write(file);
+ }
out:
+ current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim;
revert_creds(orig_cred);
}
@@ -609,34 +565,20 @@ void acct_collect(long exitcode, int group_dead)
spin_unlock_irq(&current->sighand->siglock);
}
-static void acct_process_in_ns(struct pid_namespace *ns)
+static void slow_acct_process(struct pid_namespace *ns)
{
- struct file *file = NULL;
- struct bsd_acct_struct *acct;
-
- acct = ns->bacct;
- /*
- * accelerate the common fastpath:
- */
- if (!acct || !acct->file)
- return;
-
- spin_lock(&acct_lock);
- file = acct->file;
- if (unlikely(!file)) {
- spin_unlock(&acct_lock);
- return;
+ for ( ; ns; ns = ns->parent) {
+ struct bsd_acct_struct *acct = acct_get(ns);
+ if (acct) {
+ do_acct_process(acct);
+ mutex_unlock(&acct->lock);
+ pin_put(&acct->pin);
+ }
}
- get_file(file);
- spin_unlock(&acct_lock);
-
- do_acct_process(acct, ns, file);
- fput(file);
}
/**
- * acct_process - now just a wrapper around acct_process_in_ns,
- * which in turn is a wrapper around do_acct_process.
+ * acct_process
*
* handles process accounting for an exiting task
*/
@@ -649,6 +591,10 @@ void acct_process(void)
* alive and holds its namespace, which in turn holds
* its parent.
*/
- for (ns = task_active_pid_ns(current); ns != NULL; ns = ns->parent)
- acct_process_in_ns(ns);
+ for (ns = task_active_pid_ns(current); ns != NULL; ns = ns->parent) {
+ if (ns->bacct)
+ break;
+ }
+ if (unlikely(ns))
+ slow_acct_process(ns);
}