* Re: [2.6 RFC/PATCH] Input: possible deadlock in i8042
2003-11-30 9:00 ` Vojtech Pavlik
@ 2003-11-30 9:19 ` Dmitry Torokhov
2003-12-01 6:58 ` Dmitry Torokhov
2003-12-01 6:59 ` Dmitry Torokhov
2 siblings, 0 replies; 5+ messages in thread
From: Dmitry Torokhov @ 2003-11-30 9:19 UTC (permalink / raw)
To: Vojtech Pavlik; +Cc: linux-kernel, Vojtech Pavlik, Andrew Morton
On Sunday 30 November 2003 04:00 am, Vojtech Pavlik wrote:
> On Sun, Nov 30, 2003 at 03:03:57AM -0500, Dmitry Torokhov wrote:
> > If request_irq fails in i8042_open it will call
> > serio_unregister_port, which takes serio_sem. i8042_open may be
> > called:
> >
> > serio_register_port - serio_find_dev - dev->connect
> > serio_register_device - dev->connect
> >
> > Both serio_register_port and serio_register_device take serio_sem as
> > well.
> >
> > I think that serio_{register|unregister}_port can be converted into
> > submitting requests to kseriod thus removing deadlock on the
> > serio_sem.
> >
> > The patch below is on top of serio* patches in Andrew Morton's -mm
> > tree.
>
> It's nice to avoid the deadlock this way, but I think it's not a good
> idea to make the register/unregister asynchronous - it could be a nasty
> surprise for an unsuspecting driver writer.
>
Serio_register_port is not guaranteed to find a driver for the serio anyway
as the driver can be compiled as a module and loaded much much later so it
should not be a concern. It is somewhat asynchronous already.
Dmitry
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [2.6 RFC/PATCH] Input: possible deadlock in i8042
2003-11-30 9:00 ` Vojtech Pavlik
2003-11-30 9:19 ` Dmitry Torokhov
@ 2003-12-01 6:58 ` Dmitry Torokhov
2003-12-01 6:59 ` Dmitry Torokhov
2 siblings, 0 replies; 5+ messages in thread
From: Dmitry Torokhov @ 2003-12-01 6:58 UTC (permalink / raw)
To: Vojtech Pavlik; +Cc: linux-kernel, Vojtech Pavlik, Andrew Morton
On Sunday 30 November 2003 04:00 am, Vojtech Pavlik wrote:
> On Sun, Nov 30, 2003 at 03:03:57AM -0500, Dmitry Torokhov wrote:
> > If request_irq fails in i8042_open it will call
> > serio_unregister_port, which takes serio_sem. i8042_open may be
> > called:
> >
> > serio_register_port - serio_find_dev - dev->connect
> > serio_register_device - dev->connect
> >
> > Both serio_register_port and serio_register_device take serio_sem as
> > well.
> >
> > I think that serio_{register|unregister}_port can be converted into
> > submitting requests to kseriod thus removing deadlock on the
> > serio_sem.
> >
> > The patch below is on top of serio* patches in Andrew Morton's -mm
> > tree.
>
> It's nice to avoid the deadlock this way, but I think it's not a good
> idea to make the register/unregister asynchronous - it could be a nasty
> surprise for an unsuspecting driver writer.
>
OK, you were right, it's not a good idea to convert _all_ calls to
asynchronous versions as my patch throughly screwed up module unloading.
Still, I think it is OK to export separate asynchronous versions of the
functions to be used as needed.
The patch below implements 2 new functions, serio_register_port_delayed
and serio_unregister_port_delayed that submit appropriate to kseriod.
The patch to convert i8042 to use it will follow shortly.
Dmitry
===================================================================
ChangeSet@1.1512, 2003-11-30 23:29:06-05:00, dtor_core@ameritech.net
Input: Add serio_[un]register_port_delayed to allow delayed
execution of register/unregister code (via kseriod)
when it is not clear whether serio_sem has been taken
or not.
drivers/input/serio/serio.c | 38 +++++++++++++++++++++++++++++++++++---
include/linux/serio.h | 2 ++
2 files changed, 37 insertions(+), 3 deletions(-)
===================================================================
diff -Nru a/drivers/input/serio/serio.c b/drivers/input/serio/serio.c
--- a/drivers/input/serio/serio.c Mon Dec 1 01:15:33 2003
+++ b/drivers/input/serio/serio.c Mon Dec 1 01:15:33 2003
@@ -49,8 +49,10 @@
EXPORT_SYMBOL(serio_interrupt);
EXPORT_SYMBOL(serio_register_port);
+EXPORT_SYMBOL(serio_register_port_delayed);
EXPORT_SYMBOL(__serio_register_port);
EXPORT_SYMBOL(serio_unregister_port);
+EXPORT_SYMBOL(serio_unregister_port_delayed);
EXPORT_SYMBOL(__serio_unregister_port);
EXPORT_SYMBOL(serio_register_device);
EXPORT_SYMBOL(serio_unregister_device);
@@ -83,8 +85,10 @@
}
}
-#define SERIO_RESCAN 1
-#define SERIO_RECONNECT 2
+#define SERIO_RESCAN 1
+#define SERIO_RECONNECT 2
+#define SERIO_REGISTER_PORT 3
+#define SERIO_UNREGISTER_PORT 4
static DECLARE_WAIT_QUEUE_HEAD(serio_wait);
static DECLARE_COMPLETION(serio_exited);
@@ -111,6 +115,14 @@
goto event_done;
switch (event->type) {
+ case SERIO_REGISTER_PORT :
+ __serio_register_port(event->serio);
+ break;
+
+ case SERIO_UNREGISTER_PORT :
+ __serio_unregister_port(event->serio);
+ break;
+
case SERIO_RECONNECT :
if (event->serio->dev && event->serio->dev->reconnect)
if (event->serio->dev->reconnect(event->serio) == 0)
@@ -198,8 +210,18 @@
}
/*
+ * Submits register request to kseriod for subsequent execution.
+ * Can be used when it is not obvious whether the serio_sem is
+ * taken or not and when delayed execution is feasible.
+ */
+void serio_register_port_delayed(struct serio *serio)
+{
+ serio_queue_event(serio, SERIO_REGISTER_PORT);
+}
+
+/*
* Should only be called directly if serio_sem has already been taken,
- * for example when unregistering a serio from other input device's
+ * for example when registering a serio from other input device's
* connect() function.
*/
void __serio_register_port(struct serio *serio)
@@ -213,6 +235,16 @@
down(&serio_sem);
__serio_unregister_port(serio);
up(&serio_sem);
+}
+
+/*
+ * Submits unregister request to kseriod for subsequent execution.
+ * Can be used when it is not obvious whether the serio_sem is
+ * taken or not and when delayed execution is feasible.
+ */
+void serio_unregister_port_delayed(struct serio *serio)
+{
+ serio_queue_event(serio, SERIO_UNREGISTER_PORT);
}
/*
diff -Nru a/include/linux/serio.h b/include/linux/serio.h
--- a/include/linux/serio.h Mon Dec 1 01:15:33 2003
+++ b/include/linux/serio.h Mon Dec 1 01:15:33 2003
@@ -63,8 +63,10 @@
irqreturn_t serio_interrupt(struct serio *serio, unsigned char data, unsigned int flags, struct pt_regs *regs);
void serio_register_port(struct serio *serio);
+void serio_register_port_delayed(struct serio *serio);
void __serio_register_port(struct serio *serio);
void serio_unregister_port(struct serio *serio);
+void serio_unregister_port_delayed(struct serio *serio);
void __serio_unregister_port(struct serio *serio);
void serio_register_device(struct serio_dev *dev);
void serio_unregister_device(struct serio_dev *dev);
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [2.6 RFC/PATCH] Input: possible deadlock in i8042
2003-11-30 9:00 ` Vojtech Pavlik
2003-11-30 9:19 ` Dmitry Torokhov
2003-12-01 6:58 ` Dmitry Torokhov
@ 2003-12-01 6:59 ` Dmitry Torokhov
2 siblings, 0 replies; 5+ messages in thread
From: Dmitry Torokhov @ 2003-12-01 6:59 UTC (permalink / raw)
To: Vojtech Pavlik; +Cc: linux-kernel, Vojtech Pavlik, Andrew Morton
===================================================================
ChangeSet@1.1513, 2003-11-30 23:38:00-05:00, dtor_core@ameritech.net
Input: Unregister port not only when there us no free IRQ
available but also when write to control register
failed.
Use serio_unregister_port_delayed in i8042_open to
avoid deadlock on serio_sem.
Also logging moved a bit
i8042.c | 22 +++++++++++++++-------
1 files changed, 15 insertions(+), 7 deletions(-)
===================================================================
diff -Nru a/drivers/input/serio/i8042.c b/drivers/input/serio/i8042.c
--- a/drivers/input/serio/i8042.c Mon Dec 1 01:15:56 2003
+++ b/drivers/input/serio/i8042.c Mon Dec 1 01:15:56 2003
@@ -231,21 +231,29 @@
if (request_irq(values->irq, i8042_interrupt,
SA_SHIRQ, "i8042", i8042_request_irq_cookie)) {
printk(KERN_ERR "i8042.c: Can't get irq %d for %s, unregistering the port.\n", values->irq, values->name);
- values->exists = 0;
- serio_unregister_port(port);
- return -1;
+ goto irq_fail;
}
i8042_ctr |= values->irqen;
if (i8042_command(&i8042_ctr, I8042_CMD_CTL_WCTR)) {
- printk(KERN_ERR "i8042.c: Can't write CTR while opening %s.\n", values->name);
- return -1;
+ printk(KERN_ERR "i8042.c: Can't write CTR while opening %s, unregistering the port\n", values->name);
+ goto ctr_fail;
}
i8042_interrupt(0, NULL, NULL);
return 0;
+
+ctr_fail:
+ i8042_ctr &= ~values->irqen;
+ free_irq(values->irq, i8042_request_irq_cookie);
+
+irq_fail:
+ values->exists = 0;
+ serio_unregister_port_delayed(port);
+
+ return -1;
}
/*
@@ -691,13 +699,13 @@
return -1;
}
- serio_register_port(port);
-
printk(KERN_INFO "serio: i8042 %s port at %#lx,%#lx irq %d\n",
values->name,
(unsigned long) I8042_DATA_REG,
(unsigned long) I8042_COMMAND_REG,
values->irq);
+
+ serio_register_port(port);
return 0;
}
^ permalink raw reply [flat|nested] 5+ messages in thread