Add hardware interrupt notification mechanism

This allows privileged userland drivers to get notifications of hardware
interrupts.

Initial work by Zheng Da, reworked by Damien Zammit and myself.

* Makefrag.am (libkernel_a_SOURCES): Add device/intr.c and
device/intr.h.
(include_device_HEADERS): Add include/device/notify.defs and
include/device/notify.h.
* device/dev_hdr.h (name_equal): Add declaration.
* device/ds_routines.c: Include <device/intr.h>
(ds_device_intr_register, ds_device_intr_ack): New functions.
* device/intr.c, device/intr.h: New files.
* doc/mach.texi (Device Interrupt): New section.
* i386/Makefrag.am (libkernel_a_SOURCES): Add i386/i386/irq.c and
i386/i386/irq.h.
* i386/i386/irq.c, i386/i386/irq.h: New files.
* i386/i386at/conf.c: Include <device/intr.h>.
(irqname): New macro.
(dev_name_list): Add irq device.
* include/device/device.defs (device_intr_register, device_intr_ack):
New RPCs.
* include/device/notify.defs, include/device/notify.h: New files.
* kern/startup.c: Include <device/intr.h>
(start_kernel_threads): Start intr_thread thread.
* linux/dev/arch/i386/kernel/irq.c: Include <device/intr.h>
(linux_action): Add user_intr field.
(linux_intr): Call user_intr action if any.
(mask_irq, unmask_irq): Move functions to i386/i386/pic.c
(__disable_irq, __enable_irq): Move functions to i386/i386/irq.c.
(install_user_intr_handler): New function.
(request_irq): Initialize user_intr field.
* linux/src/include/asm-i386/irq.h (__disable_irq, __enable_irq): Remove
prototypes.
* i386/i386/pic.c (mask_irq, unmask_irq): New functions.
* i386/i386/pic.h (mask_irq, unmask_irq): New prototypes.

18 files changed, 776 insertions, 81 deletions

diff --git a/Makefrag.am b/Makefrag.am
index 9cf976a5..ea612275 100644
--- a/Makefrag.am
+++ b/Makefrag.am
@@ -314,6 +314,8 @@ libkernel_a_SOURCES += \
 	device/ds_routines.h \
 	device/if_ether.h \
 	device/if_hdr.h \
+	device/intr.c \
+	device/intr.h \
 	device/io_req.h \
 	device/net_io.c \
 	device/net_io.h \
@@ -361,6 +363,8 @@ include_device_HEADERS = \
 	include/device/device_types.h \
 	include/device/disk_status.h \
 	include/device/net_status.h \
+	include/device/notify.defs \
+	include/device/notify.h \
 	include/device/tape_status.h \
 	include/device/tty_status.h
 
diff --git a/device/dev_hdr.h b/device/dev_hdr.h
index ad98e0bb..4bd12c1c 100644
--- a/device/dev_hdr.h
+++ b/device/dev_hdr.h
@@ -146,4 +146,13 @@ extern void dev_set_indirection(
     dev_ops_t   ops,
     int     	unit);
 
+/*
+ * compare device name
+ */
+extern boolean_t __attribute__ ((pure))
+name_equal(
+    const char  *src,
+    int         len,
+    const char  *target);
+
 #endif	/* _DEVICE_DEV_HDR_H_ */
diff --git a/device/ds_routines.c b/device/ds_routines.c
index fc051e8f..78ff51fe 100644
--- a/device/ds_routines.c
+++ b/device/ds_routines.c
@@ -92,6 +92,7 @@
 #include <device/device_port.h>
 #include <device/device_reply.user.h>
 #include <device/device_emul.h>
+#include <device/intr.h>
 
 #include <machine/machspl.h>
 
@@ -319,6 +320,59 @@ ds_device_map (device_t dev, vm_prot_t prot, vm_offset_t offset,
 				offset, size, pager, unmap);
 }
 
+/* TODO: missing deregister support */
+io_return_t
+ds_device_intr_register (device_t dev, int id,
+                         int flags, ipc_port_t receive_port)
+{
+  kern_return_t err;
+  mach_device_t mdev = dev->emul_data;
+
+  /* Refuse if device is dead or not completely open.  */
+  if (dev == DEVICE_NULL)
+    return D_NO_SUCH_DEVICE;
+
+  /* No flag is defined for now */
+  if (flags != 0)
+    return D_INVALID_OPERATION;
+
+  /* Must be called on the irq device only */
+  if (! name_equal(mdev->dev_ops->d_name, 3, "irq"))
+    return D_INVALID_OPERATION;
+
+  user_intr_t *e = insert_intr_entry (&irqtab, id, receive_port);
+  if (!e)
+    return D_NO_MEMORY;
+
+  // TODO The original port should be replaced
+  // when the same device driver calls it again,
+  // in order to handle the case that the device driver crashes and restarts.
+  err = install_user_intr_handler (&irqtab, id, flags, e);
+  if (err == D_SUCCESS)
+    {
+      /* If the port is installed successfully, increase its reference by 1.
+       * Thus, the port won't be destroyed after its task is terminated. */
+      ip_reference (receive_port);
+    }
+  return err;
+}
+
+kern_return_t
+ds_device_intr_ack (device_t dev, ipc_port_t receive_port)
+{
+  mach_device_t mdev = dev->emul_data;
+
+  /* Refuse if device is dead or not completely open.  */
+  if (dev == DEVICE_NULL)
+    return D_NO_SUCH_DEVICE;
+
+  /* Must be called on the irq device only */
+  if (! name_equal(mdev->dev_ops->d_name, 3, "irq"))
+    return D_INVALID_OPERATION;
+
+  return irq_acknowledge(receive_port);
+}
+
 boolean_t
 ds_notify (mach_msg_header_t *msg)
 {
diff --git a/device/intr.c b/device/intr.c
new file mode 100644
index 00000000..fbb9f495
--- /dev/null
+++ b/device/intr.c
@@ -0,0 +1,283 @@
+/*
+ * Copyright (c) 2010, 2011, 2016, 2019 Free Software Foundation, Inc.
+ *
+ * Permission to use, copy, modify and distribute this software and its
+ * documentation is hereby granted, provided that both the copyright
+ * notice and this permission notice appear in all copies of the
+ * software, derivative works or modified versions, and any portions
+ * thereof, and that both notices appear in supporting documentation.
+ *
+ * THE FREE SOFTWARE FOUNDATIONALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
+ * CONDITION.  THE FREE SOFTWARE FOUNDATION DISCLAIMS ANY LIABILITY OF ANY KIND
+ * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
+ */
+
+#include <device/intr.h>
+#include <device/device_types.h>
+#include <device/device_port.h>
+#include <device/notify.h>
+#include <kern/printf.h>
+#include <machine/spl.h>
+#include <machine/irq.h>
+#include <ipc/ipc_space.h>
+
+#ifndef MACH_XEN
+
+queue_head_t main_intr_queue;
+static boolean_t deliver_intr (int id, ipc_port_t dst_port);
+
+static user_intr_t *
+search_intr (struct irqdev *dev, ipc_port_t dst_port)
+{
+  user_intr_t *e;
+  queue_iterate (dev->intr_queue, e, user_intr_t *, chain)
+    {
+      if (e->dst_port == dst_port)
+	return e;
+    }
+  return NULL;
+}
+
+kern_return_t
+irq_acknowledge (ipc_port_t receive_port)
+{
+  user_intr_t *e;
+  kern_return_t ret = 0;
+
+  spl_t s = splhigh ();
+  e = search_intr (&irqtab, receive_port);
+
+  if (!e)
+    printf("didn't find user intr for interrupt !?\n");
+  else
+    {
+      if (!e->n_unacked)
+        ret = D_INVALID_OPERATION;
+      else
+        e->n_unacked--;
+    }
+  splx (s);
+
+  if (ret)
+    return ret;
+
+  if (irqtab.irqdev_ack)
+    (*(irqtab.irqdev_ack)) (&irqtab, e->id);
+
+  __enable_irq (irqtab.irq[e->id]);
+
+  return D_SUCCESS;
+}
+
+/* This function can only be used in the interrupt handler. */
+static void
+queue_intr (struct irqdev *dev, int id, user_intr_t *e)
+{
+  /* Until userland has handled the IRQ in the driver, we have to keep it
+   * disabled. Level-triggered interrupts would keep raising otherwise. */
+  __disable_irq (dev->irq[id]);
+
+  spl_t s = splhigh ();
+  e->n_unacked++;
+  e->interrupts++;
+  dev->tot_num_intr++;
+  splx (s);
+
+  thread_wakeup ((event_t) &intr_thread);
+}
+
+int
+deliver_user_intr (struct irqdev *dev, int id, user_intr_t *e)
+{
+  /* The reference of the port was increased
+   * when the port was installed.
+   * If the reference is 1, it means the port should
+   * have been destroyed and I destroy it now. */
+  if (e->dst_port
+      && e->dst_port->ip_references == 1)
+    {
+      printf ("irq handler [%d]: release a dead delivery port %p entry %p\n", id, e->dst_port, e);
+      ipc_port_release (e->dst_port);
+      e->dst_port = MACH_PORT_NULL;
+      thread_wakeup ((event_t) &intr_thread);
+      return 0;
+    }
+  else
+    {
+      queue_intr (dev, id, e);
+      return 1;
+    }
+}
+
+/* insert an interrupt entry in the queue.
+ * This entry exists in the queue until
+ * the corresponding interrupt port is removed.*/
+user_intr_t *
+insert_intr_entry (struct irqdev *dev, int id, ipc_port_t dst_port)
+{
+  user_intr_t *e, *new, *ret;
+  int free = 0;
+
+  new = (user_intr_t *) kalloc (sizeof (*new));
+  if (new == NULL)
+    return NULL;
+
+  /* check whether the intr entry has been in the queue. */
+  spl_t s = splhigh ();
+  e = search_intr (dev, dst_port);
+  if (e)
+    {
+      printf ("the interrupt entry for irq[%d] and port %p has already been inserted\n", id, dst_port);
+      free = 1;
+      ret = NULL;
+      goto out;
+    }
+  printf("irq handler [%d]: new delivery port %p entry %p\n", id, dst_port, new);
+  ret = new;
+  new->id = id;
+  new->dst_port = dst_port;
+  new->interrupts = 0;
+  new->n_unacked = 0;
+
+  queue_enter (dev->intr_queue, new, user_intr_t *, chain);
+out:
+  splx (s);
+  if (free)
+    kfree ((vm_offset_t) new, sizeof (*new));
+  return ret;
+}
+
+void
+intr_thread (void)
+{
+  user_intr_t *e;
+  int id;
+  ipc_port_t dst_port;
+  queue_init (&main_intr_queue);
+
+  for (;;)
+    {
+      assert_wait ((event_t) &intr_thread, FALSE);
+      /* Make sure we wake up from times to times to check for aborted processes */
+      thread_set_timeout (hz);
+      spl_t s = splhigh ();
+
+      /* Check for aborted processes */
+      queue_iterate (&main_intr_queue, e, user_intr_t *, chain)
+	{
+	  if ((!e->dst_port || e->dst_port->ip_references == 1) && e->n_unacked)
+	    {
+	      printf ("irq handler [%d]: release dead delivery %d unacked irqs port %p entry %p\n", e->id, e->n_unacked, e->dst_port, e);
+	      /* The reference of the port was increased
+	       * when the port was installed.
+	       * If the reference is 1, it means the port should
+	       * have been destroyed and I clear unacked irqs now, so the Linux
+	       * handling can trigger, and we will cleanup later after the Linux
+	       * handler is cleared. */
+	      /* TODO: rather immediately remove from Linux handler */
+	      while (e->n_unacked)
+	      {
+		__enable_irq (irqtab.irq[e->id]);
+		e->n_unacked--;
+	      }
+	    }
+	}
+
+      /* Now check for interrupts */
+      while (irqtab.tot_num_intr)
+	{
+	  int del = 0;
+
+	  queue_iterate (&main_intr_queue, e, user_intr_t *, chain)
+	    {
+	      /* if an entry doesn't have dest port,
+	       * we should remove it. */
+	      if (e->dst_port == MACH_PORT_NULL)
+		{
+		  clear_wait (current_thread (), 0, 0);
+		  del = 1;
+		  break;
+		}
+
+	      if (e->interrupts)
+		{
+		  clear_wait (current_thread (), 0, 0);
+		  id = e->id;
+		  dst_port = e->dst_port;
+		  e->interrupts--;
+		  irqtab.tot_num_intr--;
+
+		  splx (s);
+		  deliver_intr (id, dst_port);
+		  s = splhigh ();
+		}
+	    }
+
+	  /* remove the entry without dest port from the queue and free it. */
+	  if (del)
+	    {
+	      assert (!queue_empty (&main_intr_queue));
+	      queue_remove (&main_intr_queue, e, user_intr_t *, chain);
+	      if (e->n_unacked)
+		printf("irq handler [%d]: still %d unacked irqs in entry %p\n", e->id, e->n_unacked, e);
+	      while (e->n_unacked)
+	      {
+		__enable_irq (irqtab.irq[e->id]);
+		e->n_unacked--;
+	      }
+	      printf("irq handler [%d]: removed entry %p\n", e->id, e);
+	      splx (s);
+	      kfree ((vm_offset_t) e, sizeof (*e));
+	      s = splhigh ();
+	    }
+	}
+      splx (s);
+      thread_block (NULL);
+    }
+}
+
+static boolean_t
+deliver_intr (int id, ipc_port_t dst_port)
+{
+  ipc_kmsg_t kmsg;
+  device_intr_notification_t *n;
+  mach_port_t dest = (mach_port_t) dst_port;
+
+  if (dest == MACH_PORT_NULL)
+    return FALSE;
+
+  kmsg = ikm_alloc(sizeof *n);
+  if (kmsg == IKM_NULL)
+    return FALSE;
+
+  ikm_init(kmsg, sizeof *n);
+  n = (device_intr_notification_t *) &kmsg->ikm_header;
+
+  mach_msg_header_t *m = &n->intr_header;
+  mach_msg_type_t *t = &n->intr_type;
+
+  m->msgh_bits = MACH_MSGH_BITS(MACH_MSG_TYPE_PORT_SEND, 0);
+  m->msgh_size = sizeof *n;
+  m->msgh_seqno = DEVICE_NOTIFY_MSGH_SEQNO;
+  m->msgh_local_port = MACH_PORT_NULL;
+  m->msgh_remote_port = MACH_PORT_NULL;
+  m->msgh_id = DEVICE_INTR_NOTIFY;
+
+  t->msgt_name = MACH_MSG_TYPE_INTEGER_32;
+  t->msgt_size = 32;
+  t->msgt_number = 1;
+  t->msgt_inline = TRUE;
+  t->msgt_longform = FALSE;
+  t->msgt_deallocate = FALSE;
+  t->msgt_unused = 0;
+
+  n->intr_header.msgh_remote_port = dest;
+  n->id = id;
+
+  ipc_port_copy_send (dst_port);
+  ipc_mqueue_send_always(kmsg);
+
+  return TRUE;
+}
+
+#endif	/* MACH_XEN */
diff --git a/device/intr.h b/device/intr.h
new file mode 100644
index 00000000..cd3e0bce
--- /dev/null
+++ b/device/intr.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (c) 2010, 2011, 2019 Free Software Foundation, Inc.
+ *
+ * Permission to use, copy, modify and distribute this software and its
+ * documentation is hereby granted, provided that both the copyright
+ * notice and this permission notice appear in all copies of the
+ * software, derivative works or modified versions, and any portions
+ * thereof, and that both notices appear in supporting documentation.
+ *
+ * THE FREE SOFTWARE FOUNDATIONALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
+ * CONDITION.  THE FREE SOFTWARE FOUNDATION DISCLAIMS ANY LIABILITY OF ANY KIND
+ * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
+ */
+
+#ifndef __INTR_H__
+#define __INTR_H__
+
+#ifndef MACH_XEN
+
+#include <mach/kern_return.h>
+#include <mach/port.h>
+#include <kern/queue.h>
+#include <ipc/ipc_port.h>
+#include <device/conf.h>
+
+#define DEVICE_NOTIFY_MSGH_SEQNO 0
+
+#include <sys/types.h>
+
+struct irqdev;
+#include <machine/irq.h>
+
+typedef struct {
+  queue_chain_t chain;
+  int interrupts; /* Number of interrupts occurred since last run of intr_thread */
+  int n_unacked;  /* Number of times irqs were disabled for this */
+  ipc_port_t dst_port; /* Notification port */
+  int id; /* Mapping to machine dependent irq_t array elem */
+} user_intr_t;
+
+struct irqdev {
+  char *name;
+  void (*irqdev_ack)(struct irqdev *dev, int id);
+
+  queue_head_t *intr_queue;
+  int tot_num_intr; /* Total number of unprocessed interrupts */
+
+  /* Machine dependent */
+  irq_t irq[NINTR];
+};
+
+extern queue_head_t main_intr_queue;
+extern int install_user_intr_handler (struct irqdev *dev, int id, unsigned long flags, user_intr_t *e);
+extern int deliver_user_intr (struct irqdev *dev, int id, user_intr_t *e);
+extern user_intr_t *insert_intr_entry (struct irqdev *dev, int id, ipc_port_t receive_port);
+
+void intr_thread (void);
+kern_return_t irq_acknowledge (ipc_port_t receive_port);
+
+#endif /* MACH_XEN */
+
+#endif
diff --git a/doc/mach.texi b/doc/mach.texi
index 91ec96ee..11b0a066 100644
--- a/doc/mach.texi
+++ b/doc/mach.texi
@@ -286,6 +286,7 @@ Device Interface
 * Device Map::                    Mapping devices into virtual memory.
 * Device Status::                 Querying and manipulating a device.
 * Device Filter::                 Filtering packets arriving on a device.
+* Device Interrupt::              Getting hardware interrupt notifications.
 
 Kernel Debugger
 
@@ -6305,6 +6306,7 @@ All constants and functions in this chapter are defined in
 * Device Map::                    Mapping devices into virtual memory.
 * Device Status::                 Querying and manipulating a device.
 * Device Filter::                 Filtering packets arriving on a device.
+* Device Interrupt::              Getting hardware interrupt notifications.
 @end menu
 
 
@@ -6699,6 +6701,26 @@ a device port or the device is dead or not completely open.
 @end deftypefun
 
 
+@node Device Interrupt
+@section Device Interrupt
+
+@deftypefun kern_return_t device_intr_register (@w{device_t @var{device}}, @w{int @var{id}}, @w{int @var{flags}}, @w{mach_port_t @var{receive_port}})
+The function @code{device_intr_register} registers for receiving hardware
+interrupt events through @var{device_intr_notify} notifications.  The hardware
+interrupt identifier is specified by @var{id}.  @var{flags} must be set to 0.  The
+notifications will be sent on the @var{receive_port} send right.
+@code{device_intr_register} is only available on the dedicated @code{irq} device.
+@end deftypefun
+
+@deftypefun kern_return_t device_intr_ack (@w{device_t @var{device}}, @w{mach_port_t @var{receive_port}})
+On a hardware interrupt, the kernel disables the interrupt line before sending
+notifications.  To prevent from interrupt losses, the interrupt is kept disabled
+until @code{device_intr_ack} is called to acknowledge the interrupt.
+@var{receive_port} is the send right on which the interrupt notification was
+received.
+@end deftypefun
+
+
 @node Kernel Debugger
 @chapter Kernel Debugger
 
diff --git a/i386/Makefrag.am b/i386/Makefrag.am
index f38c0785..59571416 100644
--- a/i386/Makefrag.am
+++ b/i386/Makefrag.am
@@ -102,6 +102,8 @@ libkernel_a_SOURCES += \
 	i386/i386/io_perm.c \
 	i386/i386/io_perm.h \
 	i386/i386/ipl.h \
+	i386/i386/irq.c \
+	i386/i386/irq.h \
 	i386/i386/ktss.c \
 	i386/i386/ktss.h \
 	i386/i386/kttd_interface.c \
diff --git a/i386/i386/irq.c b/i386/i386/irq.c
new file mode 100644
index 00000000..c65d2ea2
--- /dev/null
+++ b/i386/i386/irq.c
@@ -0,0 +1,66 @@
+/*
+ * Copyright (C) 1995 Shantanu Goel
+ * Copyright (C) 2020 Free Software Foundation, Inc
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <i386/irq.h>
+#include <device/intr.h>
+#include <mach/kern_return.h>
+#include <kern/queue.h>
+#include <machine/machspl.h>
+
+extern queue_head_t main_intr_queue;
+
+static void
+irq_eoi (struct irqdev *dev, int id)
+{
+  /* TODO EOI(dev->irq[id]) */
+}
+
+static unsigned int ndisabled_irq[NINTR];
+
+void
+__disable_irq (irq_t irq_nr)
+{
+  assert (irq_nr < NINTR);
+
+  spl_t s = splhigh();
+  ndisabled_irq[irq_nr]++;
+  assert (ndisabled_irq[irq_nr] > 0);
+  if (ndisabled_irq[irq_nr] == 1)
+    mask_irq (irq_nr);
+  splx(s);
+}
+
+void
+__enable_irq (irq_t irq_nr)
+{
+  assert (irq_nr < NINTR);
+
+  spl_t s = splhigh();
+  assert (ndisabled_irq[irq_nr] > 0);
+  ndisabled_irq[irq_nr]--;
+  if (ndisabled_irq[irq_nr] == 0)
+    unmask_irq (irq_nr);
+  splx(s);
+}
+
+struct irqdev irqtab = {
+  "irq", irq_eoi, &main_intr_queue, 0,
+  {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
+};
+
diff --git a/i386/i386/irq.h b/i386/i386/irq.h
new file mode 100644
index 00000000..d48a8e92
--- /dev/null
+++ b/i386/i386/irq.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright (C) 2020 Free Software Foundation, Inc.
+ *
+ * Permission to use, copy, modify and distribute this software and its
+ * documentation is hereby granted, provided that both the copyright
+ * notice and this permission notice appear in all copies of the
+ * software, derivative works or modified versions, and any portions
+ * thereof, and that both notices appear in supporting documentation.
+ *
+ * THE FREE SOFTWARE FOUNDATION ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
+ * CONDITION.  THE FREE SOFTWARE FOUNDATION DISCLAIMS ANY LIABILITY OF ANY KIND
+ * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
+ */
+
+#ifndef _I386_IRQ_H
+#define _I386_IRQ_H
+
+#include <i386/pic.h>
+
+typedef unsigned int irq_t;
+
+void __enable_irq (irq_t irq);
+void __disable_irq (irq_t irq);
+
+extern struct irqdev irqtab;
+
+#endif
diff --git a/i386/i386/pic.c b/i386/i386/pic.c
index 0feebc6f..62ed9ed1 100644
--- a/i386/i386/pic.c
+++ b/i386/i386/pic.c
@@ -49,6 +49,24 @@ OTHER TORTIOUS ACTION, ARISING OUR OF OR IN CONNECTION
 WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */
 
+/*
+ * Copyright (C) 1995 Shantanu Goel.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
 #include <sys/types.h>
 #include <kern/printf.h>
 #include <i386/ipl.h>
@@ -185,3 +203,57 @@ intnull(int unit_dev)
 	}
 
 }
+
+/*
+ * Mask a PIC IRQ.
+ */
+inline void
+mask_irq (unsigned int irq_nr)
+{
+	int new_pic_mask = curr_pic_mask | 1 << irq_nr;
+
+	if (curr_pic_mask != new_pic_mask)
+	{
+		curr_pic_mask = new_pic_mask;
+		if (irq_nr < 8)
+		{
+			outb (PIC_MASTER_OCW, curr_pic_mask & 0xff);
+		}
+		else
+		{
+			outb (PIC_SLAVE_OCW, curr_pic_mask >> 8);
+		}
+	}
+}
+
+/*
+ * Unmask a PIC IRQ.
+ */
+inline void
+unmask_irq (unsigned int irq_nr)
+{
+	int mask;
+	int new_pic_mask;
+
+	mask = 1 << irq_nr;
+	if (irq_nr >= 8)
+	{
+		mask |= 1 << 2;
+	}
+
+	new_pic_mask = curr_pic_mask & ~mask;
+
+	if (curr_pic_mask != new_pic_mask)
+	{
+		curr_pic_mask = new_pic_mask;
+		if (irq_nr < 8)
+		{
+			outb (PIC_MASTER_OCW, curr_pic_mask & 0xff);
+		}
+		else
+		{
+			outb (PIC_SLAVE_OCW, curr_pic_mask >> 8);
+		}
+	}
+}
+
diff --git a/i386/i386/pic.h b/i386/i386/pic.h
index f492de5e..6434bf08 100644
--- a/i386/i386/pic.h
+++ b/i386/i386/pic.h
@@ -180,6 +180,8 @@ WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 extern void picinit (void);
 extern int curr_pic_mask;
 extern void intnull(int unit);
+extern inline void mask_irq (unsigned int irq_nr);
+extern inline void unmask_irq (unsigned int irq_nr);
 #endif /* __ASSEMBLER__ */
 
 #endif	/* _I386_PIC_H_ */
diff --git a/i386/i386at/conf.c b/i386/i386at/conf.c
index fe7c7c09..ca5d0dfb 100644
--- a/i386/i386at/conf.c
+++ b/i386/i386at/conf.c
@@ -68,6 +68,9 @@
 #define hypcnname		"hyp"
 #endif	/* MACH_HYP */
 
+#include <device/intr.h>
+#define irqname			"irq"
+
 /*
  * List of devices - console must be at slot 0
  */
@@ -149,6 +152,11 @@ struct dev_ops	dev_name_list[] =
 	  nodev },
 #endif	/* MACH_HYP */
 
+        { irqname,      nulldev_open,   nulldev_close,    nulldev_read,
+          nulldev_write,nulldev_getstat,nulldev_setstat,  nomap,
+          nodev,        nulldev,        nulldev_portdeath,0,
+          nodev },
+
 };
 int	dev_name_count = sizeof(dev_name_list)/sizeof(dev_name_list[0]);
 
diff --git a/include/device/device.defs b/include/device/device.defs
index 409146f5..ec4b5bf8 100644
--- a/include/device/device.defs
+++ b/include/device/device.defs
@@ -142,3 +142,21 @@ routine device_set_filter(
 	in	filter		: filter_array_t
 	);
 
+routine device_intr_register(
+		device		: device_t;
+	in	id		: int;
+	in	flags		: int;
+	in	receive_port	: mach_port_send_t
+	);
+
+/*
+ *	Acknowledge the specified interrupt notification.
+ */
+/*
+ * When an IRQ happens and an intr notification is thus sent, the IRQ line
+ * is kept disabled until the notification is acknowledged with this RPC
+ */
+routine device_intr_ack(
+		device		: device_t;
+	in	receive_port	: mach_port_send_t);
+
diff --git a/include/device/notify.defs b/include/device/notify.defs
new file mode 100644
index 00000000..7919b339
--- /dev/null
+++ b/include/device/notify.defs
@@ -0,0 +1,36 @@
+/* 
+ * Mach Operating System
+ * Copyright (c) 1991,1990,1989 Carnegie Mellon University
+ * All Rights Reserved.
+ * 
+ * Permission to use, copy, modify and distribute this software and its
+ * documentation is hereby granted, provided that both the copyright
+ * notice and this permission notice appear in all copies of the
+ * software, derivative works or modified versions, and any portions
+ * thereof, and that both notices appear in supporting documentation.
+ * 
+ * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
+ * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
+ * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
+ * 
+ * Carnegie Mellon requests users of this software to return to
+ * 
+ *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
+ *  School of Computer Science
+ *  Carnegie Mellon University
+ *  Pittsburgh PA 15213-3890
+ * 
+ * any improvements or extensions that they make and grant Carnegie Mellon
+ * the rights to redistribute these changes.
+ */
+
+subsystem notify 100;
+
+#include <mach/std_types.defs>
+
+serverprefix do_;
+serverdemux device_intr_notify_server;
+
+simpleroutine device_intr_notify(
+		notify	: notify_port_t;
+		id	: int);
diff --git a/include/device/notify.h b/include/device/notify.h
new file mode 100644
index 00000000..addf9114
--- /dev/null
+++ b/include/device/notify.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright (c) 2010 Free Software Foundation, Inc.
+ *
+ * Permission to use, copy, modify and distribute this software and its
+ * documentation is hereby granted, provided that both the copyright
+ * notice and this permission notice appear in all copies of the
+ * software, derivative works or modified versions, and any portions
+ * thereof, and that both notices appear in supporting documentation.
+ *
+ * THE FREE SOFTWARE FOUNDATIONALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
+ * CONDITION.  THE FREE SOFTWARE FOUNDATION DISCLAIMS ANY LIABILITY OF ANY KIND
+ * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
+ */
+
+/*
+ * Device notification definitions.
+ */
+
+#ifndef	_MACH_DEVICE_NOTIFY_H_
+#define _MACH_DEVICE_NOTIFY_H_
+
+#include <mach/port.h>
+#include <mach/message.h>
+
+typedef struct
+{
+  mach_msg_header_t intr_header;
+  mach_msg_type_t   intr_type;
+  int		    id;
+} device_intr_notification_t;
+
+#define DEVICE_INTR_NOTIFY 100
+
+#endif	/* _MACH_DEVICE_NOTIFY_H_ */
diff --git a/kern/startup.c b/kern/startup.c
index 56e05b3f..1f873192 100644
--- a/kern/startup.c
+++ b/kern/startup.c
@@ -63,6 +63,7 @@
 #include <machine/model_dep.h>
 #include <mach/version.h>
 #include <device/device_init.h>
+#include <device/intr.h>
 
 #if MACH_KDB
 #include <device/cons.h>
@@ -228,6 +229,9 @@ void start_kernel_threads(void)
 	(void) kernel_thread(kernel_task, reaper_thread, (char *) 0);
 	(void) kernel_thread(kernel_task, swapin_thread, (char *) 0);
 	(void) kernel_thread(kernel_task, sched_thread, (char *) 0);
+#ifndef MACH_XEN
+	(void) kernel_thread(kernel_task, intr_thread, (char *)0);
+#endif	/* MACH_XEN */
 
 #if	NCPUS > 1
 	/*
diff --git a/linux/dev/arch/i386/kernel/irq.c b/linux/dev/arch/i386/kernel/irq.c
index 18448638..1e911f33 100644
--- a/linux/dev/arch/i386/kernel/irq.c
+++ b/linux/dev/arch/i386/kernel/irq.c
@@ -50,6 +50,8 @@
 #include <linux/dev/glue/glue.h>
 #include <machine/machspl.h>
 
+#include <device/intr.h>
+
 #if 0
 /* XXX: This is the way it's done in linux 2.2. GNU Mach currently uses intr_count. It should be made using local_{bh/irq}_count instead (through hardirq_enter/exit) for SMP support. */
 unsigned int local_bh_count[NR_CPUS];
@@ -76,6 +78,7 @@ struct linux_action
   void *dev_id;
   struct linux_action *next;
   unsigned long flags;
+  user_intr_t *user_intr;
 };
 
 static struct linux_action *irq_action[16] =
@@ -95,6 +98,7 @@ linux_intr (int irq)
 {
   struct pt_regs regs;
   struct linux_action *action = *(irq_action + irq);
+  struct linux_action **prev = &irq_action[irq];
   unsigned long flags;
 
   kstat.interrupts[irq]++;
@@ -106,93 +110,35 @@ linux_intr (int irq)
 
   while (action)
     {
-      action->handler (irq, action->dev_id, &regs);
+      // TODO I might need to check whether the interrupt belongs to
+      // the current device. But I don't do it for now.
+      if (action->user_intr)
+	{
+	  if (!deliver_user_intr(&irqtab, irq, action->user_intr))
+	  {
+	    *prev = action->next;
+	    linux_kfree(action);
+	    action = *prev;
+	    continue;
+	  }
+	}
+      else if (action->handler)
+	action->handler (irq, action->dev_id, &regs);
+      prev = &action->next;
       action = action->next;
     }
 
-  restore_flags (flags);
-
-  intr_count--;
-}
-
-/*
- * Mask an IRQ.
- */
-static inline void
-mask_irq (unsigned int irq_nr)
-{
-  int new_pic_mask = curr_pic_mask | 1 << irq_nr;
-
-  if (curr_pic_mask != new_pic_mask)
-    {
-      curr_pic_mask = new_pic_mask;
-      if (irq_nr < 8)
-       outb (curr_pic_mask & 0xff, PIC_MASTER_OCW);
-      else
-       outb (curr_pic_mask >> 8, PIC_SLAVE_OCW);
-    }
-}
-
-/*
- * Unmask an IRQ.
- */
-static inline void
-unmask_irq (unsigned int irq_nr)
-{
-  int mask;
-  int new_pic_mask;
-
-  mask = 1 << irq_nr;
-  if (irq_nr >= 8)
-    mask |= 1 << 2;
-
-  new_pic_mask = curr_pic_mask & ~mask;
-
-  if (curr_pic_mask != new_pic_mask)
+  if (!irq_action[irq])
     {
-      curr_pic_mask = new_pic_mask;
-      if (irq_nr < 8)
-	outb (curr_pic_mask & 0xff, PIC_MASTER_OCW);
-      else
-	outb (curr_pic_mask >> 8, PIC_SLAVE_OCW);
+      /* No handler any more, disable interrupt */
+      mask_irq (irq);
+      ivect[irq] = intnull;
+      iunit[irq] = irq;
     }
-}
-
-/* Count how many subsystems requested to disable each IRQ */
-static unsigned ndisabled_irq[NR_IRQS];
-
-/* These disable/enable IRQs for real after counting how many subsystems
- * requested that */
-void
-__disable_irq (unsigned int irq_nr)
-{
-  unsigned long flags;
-
-  assert (irq_nr < NR_IRQS);
 
-  save_flags (flags);
-  cli ();
-  ndisabled_irq[irq_nr]++;
-  assert (ndisabled_irq[irq_nr] > 0);
-  if (ndisabled_irq[irq_nr] == 1)
-    mask_irq (irq_nr);
   restore_flags (flags);
-}
 
-void
-__enable_irq (unsigned int irq_nr)
-{
-  unsigned long flags;
-
-  assert (irq_nr < NR_IRQS);
-
-  save_flags (flags);
-  cli ();
-  assert (ndisabled_irq[irq_nr] > 0);
-  ndisabled_irq[irq_nr]--;
-  if (ndisabled_irq[irq_nr] == 0)
-    unmask_irq (irq_nr);
-  restore_flags (flags);
+  intr_count--;
 }
 
 /* IRQ mask according to Linux drivers */
@@ -273,6 +219,53 @@ setup_x86_irq (int irq, struct linux_action *new)
   return 0;
 }
 
+int
+install_user_intr_handler (struct irqdev *dev, int id, unsigned long flags,
+			  user_intr_t *user_intr)
+{
+  struct linux_action *action;
+  struct linux_action *old;
+  int retval;
+
+  unsigned int irq = dev->irq[id];
+
+  assert (irq < 16);
+
+  /* Test whether the irq handler has been set */
+  // TODO I need to protect the array when iterating it.
+  old = irq_action[irq];
+  while (old)
+    {
+      if (old->user_intr && old->user_intr->dst_port == user_intr->dst_port)
+	{
+	  printk ("The interrupt handler has already been installed on line %d", irq);
+	  return linux_to_mach_error (-EAGAIN);
+	}
+      old = old->next;
+    }
+
+  /*
+   * Hmm... Should I use `kalloc()' ?
+   * By OKUJI Yoshinori.
+   */
+  action = (struct linux_action *)
+    linux_kmalloc (sizeof (struct linux_action), GFP_KERNEL);
+  if (action == NULL)
+    return linux_to_mach_error (-ENOMEM);
+  
+  action->handler = NULL;
+  action->next = NULL;
+  action->dev_id = NULL;
+  action->flags = SA_SHIRQ;
+  action->user_intr = user_intr;
+  
+  retval = setup_x86_irq (irq, action);
+  if (retval)
+    linux_kfree (action);
+  
+  return linux_to_mach_error (retval);
+}
+
 /*
  * Attach a handler to an IRQ.
  */
@@ -301,6 +294,7 @@ request_irq (unsigned int irq, void (*handler) (int, void *, struct pt_regs *),
   action->next = NULL;
   action->dev_id = dev_id;
   action->flags = flags;
+  action->user_intr = NULL;
   
   retval = setup_x86_irq (irq, action);
   if (retval)
diff --git a/linux/src/include/asm-i386/irq.h b/linux/src/include/asm-i386/irq.h
index d7d1e3c5..c75744a5 100644
--- a/linux/src/include/asm-i386/irq.h
+++ b/linux/src/include/asm-i386/irq.h
@@ -16,8 +16,6 @@
 
 #define TIMER_IRQ 0
 
-extern void __disable_irq(unsigned int);
-extern void __enable_irq(unsigned int);
 extern void disable_irq(unsigned int);
 extern void enable_irq(unsigned int);