Have EVFILT_TIMERS allocate their callouts via malloc() instead of using

the static callout list allocated by the system.

Change malloc type from M_TEMP to M_KQUEUE to better track memory.

Add a kern.kq_calloutmax to globally limit the amount of kernel memory
that can be allocated by callouts.

Submitted by: iedowse  (items 1, 2)

1 files changed, 30 insertions, 14 deletions

diff --git a/sys/kern/kern_event.c b/sys/kern/kern_event.c
index 44a1401..191010d 100644
--- a/sys/kern/kern_event.c
+++ b/sys/kern/kern_event.c
@@ -45,11 +45,14 @@
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/stat.h>
+#include <sys/sysctl.h>
 #include <sys/sysproto.h>
 #include <sys/uio.h>
 
 #include <vm/vm_zone.h>
 
+MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system");
+
 static int	kqueue_scan(struct file *fp, int maxevents,
 		    struct kevent *ulistp, const struct timespec *timeout,
 		    struct thread *td);
@@ -105,6 +108,10 @@ static struct filterops timer_filtops =
 	{ 0, filt_timerattach, filt_timerdetach, filt_timer };
 
 static vm_zone_t	knote_zone;
+static int 		kq_ncallouts = 0;
+static int 		kq_calloutmax = (4 * 1024);
+SYSCTL_INT(_kern, OID_AUTO, kq_calloutmax, CTLFLAG_RW,
+    &kq_calloutmax, 0, "Maximum number of callouts allocated for kqueue");
 
 #define KNOTE_ACTIVATE(kn) do { 					\
 	kn->kn_status |= KN_ACTIVE;					\
@@ -278,7 +285,7 @@ static void
 filt_timerexpire(void *knx)
 {
 	struct knote *kn = knx;
-	struct callout_handle ch;
+	struct callout *calloutp;
 	struct timeval tv;
 	int tticks;
 
@@ -289,8 +296,8 @@ filt_timerexpire(void *knx)
 		tv.tv_sec = kn->kn_sdata / 1000;
 		tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
 		tticks = tvtohz(&tv);
-		ch = timeout(filt_timerexpire, kn, tticks);
-		kn->kn_hook = (caddr_t)ch.callout;
+		calloutp = (struct callout *)kn->kn_hook;
+		callout_reset(calloutp, tticks, filt_timerexpire, kn);
 	}
 }
 
@@ -300,17 +307,24 @@ filt_timerexpire(void *knx)
 static int
 filt_timerattach(struct knote *kn)
 {
-	struct callout_handle ch;
+	struct callout *calloutp;
 	struct timeval tv;
 	int tticks;
 
+	if (kq_ncallouts >= kq_calloutmax)
+		return (ENOMEM);
+	kq_ncallouts++;
+
 	tv.tv_sec = kn->kn_sdata / 1000;
 	tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
 	tticks = tvtohz(&tv);
 
 	kn->kn_flags |= EV_CLEAR;		/* automatically set */
-	ch = timeout(filt_timerexpire, kn, tticks);
-	kn->kn_hook = (caddr_t)ch.callout;
+	MALLOC(calloutp, struct callout *, sizeof(*calloutp),
+	    M_KQUEUE, M_WAITOK);
+	callout_init(calloutp, 0);
+	callout_reset(calloutp, tticks, filt_timerexpire, kn);
+	kn->kn_hook = (caddr_t)calloutp;
 
 	return (0);
 }
@@ -318,10 +332,12 @@ filt_timerattach(struct knote *kn)
 static void
 filt_timerdetach(struct knote *kn)
 {
-	struct callout_handle ch;
+	struct callout *calloutp;
 
-	ch.callout = (struct callout *)kn->kn_hook;
-	untimeout(filt_timerexpire, kn, ch);
+	calloutp = (struct callout *)kn->kn_hook;
+	callout_stop(calloutp);
+	FREE(calloutp, M_KQUEUE);
+	kq_ncallouts--;
 }
 
 static int
@@ -350,7 +366,7 @@ kqueue(struct thread *td, struct kqueue_args *uap)
 	fp->f_flag = FREAD | FWRITE;
 	fp->f_type = DTYPE_KQUEUE;
 	fp->f_ops = &kqueueops;
-	kq = malloc(sizeof(struct kqueue), M_TEMP, M_WAITOK | M_ZERO);
+	kq = malloc(sizeof(struct kqueue), M_KQUEUE, M_WAITOK | M_ZERO);
 	TAILQ_INIT(&kq->kq_head);
 	fp->f_data = (caddr_t)kq;
 	td->td_retval[0] = fd;
@@ -809,7 +825,7 @@ kqueue_close(struct file *fp, struct thread *td)
 			}
 		}
 	}
-	free(kq, M_TEMP);
+	free(kq, M_KQUEUE);
 	fp->f_data = NULL;
 
 	return (0);
@@ -877,7 +893,7 @@ knote_attach(struct knote *kn, struct filedesc *fdp)
 
 	if (! kn->kn_fop->f_isfd) {
 		if (fdp->fd_knhashmask == 0)
-			fdp->fd_knhash = hashinit(KN_HASHSIZE, M_TEMP,
+			fdp->fd_knhash = hashinit(KN_HASHSIZE, M_KQUEUE,
 			    &fdp->fd_knhashmask);
 		list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)];
 		goto done;
@@ -888,14 +904,14 @@ knote_attach(struct knote *kn, struct filedesc *fdp)
 		while (size <= kn->kn_id)
 			size += KQEXTENT;
 		MALLOC(list, struct klist *,
-		    size * sizeof(struct klist *), M_TEMP, M_WAITOK);
+		    size * sizeof(struct klist *), M_KQUEUE, M_WAITOK);
 		bcopy((caddr_t)fdp->fd_knlist, (caddr_t)list,
 		    fdp->fd_knlistsize * sizeof(struct klist *));
 		bzero((caddr_t)list +
 		    fdp->fd_knlistsize * sizeof(struct klist *),
 		    (size - fdp->fd_knlistsize) * sizeof(struct klist *));
 		if (fdp->fd_knlist != NULL)
-			FREE(fdp->fd_knlist, M_TEMP);
+			FREE(fdp->fd_knlist, M_KQUEUE);
 		fdp->fd_knlistsize = size;
 		fdp->fd_knlist = list;
 	}