While I was in vacation, I noticed that more "tsk == current" check were
added to the signal logic because of the way signalfd fetches other task
signals.
I'm playing at the moment with this patch, that recall Ben's idea of
attaching to the sighand only during read/poll, and calling dequeue_signal()
only with "current". This simplifies the signalfd logic quite a bit.
If this patch is applied, a task calling signalfd can read its own private
signals, and its own group signals.
Comments?
- Davide
---
fs/exec.c | 3
fs/signalfd.c | 186 +++++++---------------------------------------
include/linux/init_task.h | 2
include/linux/sched.h | 2
include/linux/signalfd.h | 29 -------
kernel/exit.c | 9 --
kernel/fork.c | 2
kernel/signal.c | 8 -
8 files changed, 40 insertions(+), 201 deletions(-)
Index: linux-2.6.mod/fs/signalfd.c
===================================================================
--- linux-2.6.mod.orig/fs/signalfd.c 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/fs/signalfd.c 2007-09-01 12:33:37.000000000 -0700
@@ -11,8 +11,10 @@
* Now using anonymous inode source.
* Thanks to Oleg Nesterov for useful code review and suggestions.
* More comments and suggestions from Arnd Bergmann.
- * Sat May 19, 2007: Davi E. M. Arnaut <[email protected]>
+ * Sat May 19, 2007: Davi E. M. Arnaut <[email protected]>
* Retrieve multiple signals with one read() call
+ * Sun Jul 15, 2007: Davide Libenzi <[email protected]>
+ * Attach to the sighand only during read().
*/
#include <linux/file.h>
@@ -27,102 +29,20 @@
#include <linux/signalfd.h>
struct signalfd_ctx {
- struct list_head lnk;
- wait_queue_head_t wqh;
sigset_t sigmask;
- struct task_struct *tsk;
};
-struct signalfd_lockctx {
- struct task_struct *tsk;
- unsigned long flags;
-};
-
-/*
- * Tries to acquire the sighand lock. We do not increment the sighand
- * use count, and we do not even pin the task struct, so we need to
- * do it inside an RCU read lock, and we must be prepared for the
- * ctx->tsk going to NULL (in signalfd_deliver()), and for the sighand
- * being detached. We return 0 if the sighand has been detached, or
- * 1 if we were able to pin the sighand lock.
- */
-static int signalfd_lock(struct signalfd_ctx *ctx, struct signalfd_lockctx *lk)
-{
- struct sighand_struct *sighand = NULL;
-
- rcu_read_lock();
- lk->tsk = rcu_dereference(ctx->tsk);
- if (likely(lk->tsk != NULL))
- sighand = lock_task_sighand(lk->tsk, &lk->flags);
- rcu_read_unlock();
-
- if (!sighand)
- return 0;
-
- if (!ctx->tsk) {
- unlock_task_sighand(lk->tsk, &lk->flags);
- return 0;
- }
-
- if (lk->tsk->tgid == current->tgid)
- lk->tsk = current;
-
- return 1;
-}
-
-static void signalfd_unlock(struct signalfd_lockctx *lk)
-{
- unlock_task_sighand(lk->tsk, &lk->flags);
-}
-
/*
- * This must be called with the sighand lock held.
+ * This must be called with the "tsk" sighand lock held.
*/
void signalfd_deliver(struct task_struct *tsk, int sig)
{
- struct sighand_struct *sighand = tsk->sighand;
- struct signalfd_ctx *ctx, *tmp;
-
- BUG_ON(!sig);
- list_for_each_entry_safe(ctx, tmp, &sighand->signalfd_list, lnk) {
- /*
- * We use a negative signal value as a way to broadcast that the
- * sighand has been orphaned, so that we can notify all the
- * listeners about this. Remember the ctx->sigmask is inverted,
- * so if the user is interested in a signal, that corresponding
- * bit will be zero.
- */
- if (sig < 0) {
- if (ctx->tsk == tsk) {
- ctx->tsk = NULL;
- list_del_init(&ctx->lnk);
- wake_up(&ctx->wqh);
- }
- } else {
- if (!sigismember(&ctx->sigmask, sig))
- wake_up(&ctx->wqh);
- }
- }
-}
-
-static void signalfd_cleanup(struct signalfd_ctx *ctx)
-{
- struct signalfd_lockctx lk;
-
- /*
- * This is tricky. If the sighand is gone, we do not need to remove
- * context from the list, the list itself won't be there anymore.
- */
- if (signalfd_lock(ctx, &lk)) {
- list_del(&ctx->lnk);
- signalfd_unlock(&lk);
- }
- kfree(ctx);
+ wake_up(&tsk->sighand->signalfd_wqh);
}
static int signalfd_release(struct inode *inode, struct file *file)
{
- signalfd_cleanup(file->private_data);
+ kfree(file->private_data);
return 0;
}
@@ -130,23 +50,15 @@
{
struct signalfd_ctx *ctx = file->private_data;
unsigned int events = 0;
- struct signalfd_lockctx lk;
- poll_wait(file, &ctx->wqh, wait);
+ poll_wait(file, ¤t->sighand->signalfd_wqh, wait);
- /*
- * Let the caller get a POLLIN in this case, ala socket recv() when
- * the peer disconnects.
- */
- if (signalfd_lock(ctx, &lk)) {
- if ((lk.tsk == current &&
- next_signal(&lk.tsk->pending, &ctx->sigmask) > 0) ||
- next_signal(&lk.tsk->signal->shared_pending,
- &ctx->sigmask) > 0)
- events |= POLLIN;
- signalfd_unlock(&lk);
- } else
+ spin_lock_irq(¤t->sighand->siglock);
+ if (next_signal(¤t->pending, &ctx->sigmask) > 0 ||
+ next_signal(¤t->signal->shared_pending,
+ &ctx->sigmask) > 0)
events |= POLLIN;
+ spin_unlock_irq(¤t->sighand->siglock);
return events;
}
@@ -219,59 +131,46 @@
int nonblock)
{
ssize_t ret;
- struct signalfd_lockctx lk;
DECLARE_WAITQUEUE(wait, current);
- if (!signalfd_lock(ctx, &lk))
- return 0;
-
- ret = dequeue_signal(lk.tsk, &ctx->sigmask, info);
+ spin_lock_irq(¤t->sighand->siglock);
+ ret = dequeue_signal(current, &ctx->sigmask, info);
switch (ret) {
case 0:
if (!nonblock)
break;
ret = -EAGAIN;
default:
- signalfd_unlock(&lk);
+ spin_unlock_irq(¤t->sighand->siglock);
return ret;
}
- add_wait_queue(&ctx->wqh, &wait);
+ add_wait_queue(¤t->sighand->signalfd_wqh, &wait);
for (;;) {
set_current_state(TASK_INTERRUPTIBLE);
- ret = dequeue_signal(lk.tsk, &ctx->sigmask, info);
- signalfd_unlock(&lk);
+ ret = dequeue_signal(current, &ctx->sigmask, info);
if (ret != 0)
break;
if (signal_pending(current)) {
ret = -ERESTARTSYS;
break;
}
+ spin_unlock_irq(¤t->sighand->siglock);
schedule();
- ret = signalfd_lock(ctx, &lk);
- if (unlikely(!ret)) {
- /*
- * Let the caller read zero byte, ala socket
- * recv() when the peer disconnect. This test
- * must be done before doing a dequeue_signal(),
- * because if the sighand has been orphaned,
- * the dequeue_signal() call is going to crash
- * because ->sighand will be long gone.
- */
- break;
- }
+ spin_lock_irq(¤t->sighand->siglock);
}
+ spin_unlock_irq(¤t->sighand->siglock);
- remove_wait_queue(&ctx->wqh, &wait);
+ remove_wait_queue(¤t->sighand->signalfd_wqh, &wait);
__set_current_state(TASK_RUNNING);
return ret;
}
/*
- * Returns either the size of a "struct signalfd_siginfo", or zero if the
- * sighand we are attached to, has been orphaned. The "count" parameter
- * must be at least the size of a "struct signalfd_siginfo".
+ * Returns a multiple of the size of a "struct signalfd_siginfo", or a negative
+ * error code. The "count" parameter must be at least the size of a
+ * "struct signalfd_siginfo".
*/
static ssize_t signalfd_read(struct file *file, char __user *buf, size_t count,
loff_t *ppos)
@@ -287,7 +186,6 @@
return -EINVAL;
siginfo = (struct signalfd_siginfo __user *) buf;
-
do {
ret = signalfd_dequeue(ctx, &info, nonblock);
if (unlikely(ret <= 0))
@@ -300,7 +198,7 @@
nonblock = 1;
} while (--count);
- return total ? total : ret;
+ return total ? total: ret;
}
static const struct file_operations signalfd_fops = {
@@ -309,20 +207,13 @@
.read = signalfd_read,
};
-/*
- * Create a file descriptor that is associated with our signal
- * state. We can pass it around to others if we want to, but
- * it will always be _our_ signal state.
- */
asmlinkage long sys_signalfd(int ufd, sigset_t __user *user_mask, size_t sizemask)
{
int error;
sigset_t sigmask;
struct signalfd_ctx *ctx;
- struct sighand_struct *sighand;
struct file *file;
struct inode *inode;
- struct signalfd_lockctx lk;
if (sizemask != sizeof(sigset_t) ||
copy_from_user(&sigmask, user_mask, sizeof(sigmask)))
@@ -335,17 +226,7 @@
if (!ctx)
return -ENOMEM;
- init_waitqueue_head(&ctx->wqh);
ctx->sigmask = sigmask;
- ctx->tsk = current->group_leader;
-
- sighand = current->sighand;
- /*
- * Add this fd to the list of signal listeners.
- */
- spin_lock_irq(&sighand->siglock);
- list_add_tail(&ctx->lnk, &sighand->signalfd_list);
- spin_unlock_irq(&sighand->siglock);
/*
* When we call this, the initialization must be complete, since
@@ -364,23 +245,18 @@
fput(file);
return -EINVAL;
}
- /*
- * We need to be prepared of the fact that the sighand this fd
- * is attached to, has been detched. In that case signalfd_lock()
- * will return 0, and we'll just skip setting the new mask.
- */
- if (signalfd_lock(ctx, &lk)) {
- ctx->sigmask = sigmask;
- signalfd_unlock(&lk);
- }
- wake_up(&ctx->wqh);
+ spin_lock_irq(¤t->sighand->siglock);
+ ctx->sigmask = sigmask;
+ spin_unlock_irq(¤t->sighand->siglock);
+
+ wake_up(¤t->sighand->signalfd_wqh);
fput(file);
}
return ufd;
err_fdalloc:
- signalfd_cleanup(ctx);
+ kfree(ctx);
return error;
}
Index: linux-2.6.mod/fs/exec.c
===================================================================
--- linux-2.6.mod.orig/fs/exec.c 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/fs/exec.c 2007-09-01 10:38:25.000000000 -0700
@@ -50,7 +50,6 @@
#include <linux/tsacct_kern.h>
#include <linux/cn_proc.h>
#include <linux/audit.h>
-#include <linux/signalfd.h>
#include <asm/uaccess.h>
#include <asm/mmu_context.h>
@@ -784,7 +783,6 @@
* and we can just re-use it all.
*/
if (atomic_read(&oldsighand->count) <= 1) {
- signalfd_detach(tsk);
exit_itimers(sig);
return 0;
}
@@ -923,7 +921,6 @@
sig->flags = 0;
no_thread_group:
- signalfd_detach(tsk);
exit_itimers(sig);
if (leader)
release_task(leader);
Index: linux-2.6.mod/kernel/exit.c
===================================================================
--- linux-2.6.mod.orig/kernel/exit.c 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/kernel/exit.c 2007-09-01 10:38:25.000000000 -0700
@@ -24,7 +24,6 @@
#include <linux/pid_namespace.h>
#include <linux/ptrace.h>
#include <linux/profile.h>
-#include <linux/signalfd.h>
#include <linux/mount.h>
#include <linux/proc_fs.h>
#include <linux/kthread.h>
@@ -86,14 +85,6 @@
sighand = rcu_dereference(tsk->sighand);
spin_lock(&sighand->siglock);
- /*
- * Notify that this sighand has been detached. This must
- * be called with the tsk->sighand lock held. Also, this
- * access tsk->sighand internally, so it must be called
- * before tsk->sighand is reset.
- */
- signalfd_detach_locked(tsk);
-
posix_cpu_timers_exit(tsk);
if (atomic_dec_and_test(&sig->count))
posix_cpu_timers_exit_group(tsk);
Index: linux-2.6.mod/include/linux/signalfd.h
===================================================================
--- linux-2.6.mod.orig/include/linux/signalfd.h 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/include/linux/signalfd.h 2007-09-01 12:36:20.000000000 -0700
@@ -57,37 +57,14 @@
*/
static inline void signalfd_notify(struct task_struct *tsk, int sig)
{
- if (unlikely(!list_empty(&tsk->sighand->signalfd_list)))
+ if (unlikely(waitqueue_active(&tsk->sighand->signalfd_wqh)))
signalfd_deliver(tsk, sig);
}
-/*
- * The signal -1 is used to notify the signalfd that the sighand
- * is on its way to be detached.
- */
-static inline void signalfd_detach_locked(struct task_struct *tsk)
-{
- if (unlikely(!list_empty(&tsk->sighand->signalfd_list)))
- signalfd_deliver(tsk, -1);
-}
-
-static inline void signalfd_detach(struct task_struct *tsk)
-{
- struct sighand_struct *sighand = tsk->sighand;
-
- if (unlikely(!list_empty(&sighand->signalfd_list))) {
- spin_lock_irq(&sighand->siglock);
- signalfd_deliver(tsk, -1);
- spin_unlock_irq(&sighand->siglock);
- }
-}
-
#else /* CONFIG_SIGNALFD */
-#define signalfd_deliver(t, s) do { } while (0)
-#define signalfd_notify(t, s) do { } while (0)
-#define signalfd_detach_locked(t) do { } while (0)
-#define signalfd_detach(t) do { } while (0)
+static void signalfd_deliver(struct task_struct *tsk, int sig) { }
+static inline void signalfd_notify(struct task_struct *tsk, int sig) { }
#endif /* CONFIG_SIGNALFD */
Index: linux-2.6.mod/kernel/signal.c
===================================================================
--- linux-2.6.mod.orig/kernel/signal.c 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/kernel/signal.c 2007-09-01 10:38:25.000000000 -0700
@@ -378,8 +378,7 @@
/* We only dequeue private signals from ourselves, we don't let
* signalfd steal them
*/
- if (likely(tsk == current))
- signr = __dequeue_signal(&tsk->pending, mask, info);
+ signr = __dequeue_signal(&tsk->pending, mask, info);
if (!signr) {
signr = __dequeue_signal(&tsk->signal->shared_pending,
mask, info);
@@ -407,8 +406,7 @@
}
}
}
- if (likely(tsk == current))
- recalc_sigpending();
+ recalc_sigpending();
if (signr && unlikely(sig_kernel_stop(signr))) {
/*
* Set a marker that we have dequeued a stop signal. Our
@@ -425,7 +423,7 @@
if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT))
tsk->signal->flags |= SIGNAL_STOP_DEQUEUED;
}
- if (signr && likely(tsk == current) &&
+ if (signr &&
((info->si_code & __SI_MASK) == __SI_TIMER) &&
info->si_sys_private){
/*
Index: linux-2.6.mod/include/linux/sched.h
===================================================================
--- linux-2.6.mod.orig/include/linux/sched.h 2007-09-01 10:37:39.000000000 -0700
+++ linux-2.6.mod/include/linux/sched.h 2007-09-01 12:25:03.000000000 -0700
@@ -438,7 +438,7 @@
atomic_t count;
struct k_sigaction action[_NSIG];
spinlock_t siglock;
- struct list_head signalfd_list;
+ wait_queue_head_t signalfd_wqh;
};
struct pacct_struct {
Index: linux-2.6.mod/kernel/fork.c
===================================================================
--- linux-2.6.mod.orig/kernel/fork.c 2007-09-01 12:27:47.000000000 -0700
+++ linux-2.6.mod/kernel/fork.c 2007-09-01 12:27:55.000000000 -0700
@@ -1438,7 +1438,7 @@
struct sighand_struct *sighand = data;
spin_lock_init(&sighand->siglock);
- INIT_LIST_HEAD(&sighand->signalfd_list);
+ init_waitqueue_head(&sighand->signalfd_wqh);
}
void __init proc_caches_init(void)
Index: linux-2.6.mod/include/linux/init_task.h
===================================================================
--- linux-2.6.mod.orig/include/linux/init_task.h 2007-09-01 12:29:17.000000000 -0700
+++ linux-2.6.mod/include/linux/init_task.h 2007-09-01 12:29:24.000000000 -0700
@@ -86,7 +86,7 @@
.count = ATOMIC_INIT(1), \
.action = { { { .sa_handler = NULL, } }, }, \
.siglock = __SPIN_LOCK_UNLOCKED(sighand.siglock), \
- .signalfd_list = LIST_HEAD_INIT(sighand.signalfd_list), \
+ .signalfd_wqh = __WAIT_QUEUE_HEAD_INITIALIZER(sighand.signalfd_wqh), \
}
extern struct group_info init_groups;
On 09/01, Davide Libenzi wrote:
>
> I'm playing at the moment with this patch, that recall Ben's idea of
> attaching to the sighand only during read/poll, and calling dequeue_signal()
> only with "current". This simplifies the signalfd logic quite a bit.
> If this patch is applied, a task calling signalfd can read its own private
> signals, and its own group signals.
>
> fs/exec.c | 3
> fs/signalfd.c | 186 +++++++---------------------------------------
> include/linux/init_task.h | 2
> include/linux/sched.h | 2
> include/linux/signalfd.h | 29 -------
> kernel/exit.c | 9 --
> kernel/fork.c | 2
> kernel/signal.c | 8 -
> 8 files changed, 40 insertions(+), 201 deletions(-)
Imho, very very nice. We lose the ability to read the cross-process signals,
but I doubt very much we should regret about that.
I cc'ed Michael, because it makes sense to document a user-visible change.
With this patch, the forked child reads its own signals (not parent's) via
the inherited signalfd (or if it was passed with unix socket).
Small problem: unless I missed something, signalfd_deliver() and sys_signalfd()
should use wake_up_all(), not wake_up() which implies nr_exclusive == 1.
It is possible that we have multiple threads waiting on ->signalfd_wqh with
the the different ->sigmask. In this case, the first woken thread can ignore
the signal, we should wake up all of them.
We can optimize this later, using a "clever" wait_queue_func_t if needed.
> + spin_lock_irq(¤t->sighand->siglock);
> + if (next_signal(¤t->pending, &ctx->sigmask) > 0 ||
> + next_signal(¤t->signal->shared_pending,
> + &ctx->sigmask) > 0)
Very minor nit: next_signal() always returns the value >= 0, imho the "> 0"
check looks a bit confusing.
Oleg.
On 09/02, Oleg Nesterov wrote:
>
> Small problem: unless I missed something, signalfd_deliver() and sys_signalfd()
> should use wake_up_all(), not wake_up() which implies nr_exclusive == 1.
>
> It is possible that we have multiple threads waiting on ->signalfd_wqh with
> the the different ->sigmask. In this case, the first woken thread can ignore
> the signal, we should wake up all of them.
Oops, sorry for noise. I forgot about WQ_FLAG_EXCLUSIVE, we don't set it by
default.
Oleg.
On Sun, 2 Sep 2007, Oleg Nesterov wrote:
> We can optimize this later, using a "clever" wait_queue_func_t if needed.
Good idea. Will do ...
- Davide