2004-11-17 23:57:40

by Ian Pratt

Subject: [patch 3] Xen core patch : runtime VT console disable

This patch enables the VT console to be disabled at runtime even if it
is built into the kernel. arch-xen needs this to avoid trying to
initialise a VT in virtual machine that doesn't have access to the
console hardware.

Signed-off-by: [email protected]

---

diff -Nurp pristine-linux-2.6.9/drivers/char/tty_io.c linux-2.6.9-xen-sparse/drivers/char/tty_io.c
--- pristine-linux-2.6.9/drivers/char/tty_io.c 2004-10-18 22:54:32.000000000 +0100
+++ linux-2.6.9-xen-sparse/drivers/char/tty_io.c 2004-11-17 01:51:48.000000000 +0000
@@ -131,6 +131,8 @@ LIST_HEAD(tty_drivers); /* linked list
vt.c for deeply disgusting hack reasons */
DECLARE_MUTEX(tty_sem);

+int console_use_vt = 1;
+
#ifdef CONFIG_UNIX98_PTYS
extern struct tty_driver *ptm_driver; /* Unix98 pty masters; for /dev/ptmx */
extern int pty_limit; /* Config limit on Unix98 ptys */
@@ -1739,7 +1741,7 @@ retry_open:
goto got_driver;
}
#ifdef CONFIG_VT
- if (device == MKDEV(TTY_MAJOR,0)) {
+ if (console_use_vt && device == MKDEV(TTY_MAJOR,0)) {
extern int fg_console;
extern struct tty_driver *console_driver;
driver = console_driver;
@@ -1959,7 +1961,7 @@ static int tiocswinsz(struct tty_struct
if (!memcmp(&tmp_ws, &tty->winsize, sizeof(*arg)))
return 0;
#ifdef CONFIG_VT
- if (tty->driver->type == TTY_DRIVER_TYPE_CONSOLE) {
+ if (console_use_vt && tty->driver->type == TTY_DRIVER_TYPE_CONSOLE) {
unsigned int currcons = tty->index;
int rc;

@@ -2933,14 +2935,19 @@ static int __init tty_init(void)
#endif

#ifdef CONFIG_VT
- cdev_init(&vc0_cdev, &console_fops);
- if (cdev_add(&vc0_cdev, MKDEV(TTY_MAJOR, 0), 1) ||
- register_chrdev_region(MKDEV(TTY_MAJOR, 0), 1, "/dev/vc/0") < 0)
- panic("Couldn't register /dev/tty0 driver\n");
- devfs_mk_cdev(MKDEV(TTY_MAJOR, 0), S_IFCHR|S_IRUSR|S_IWUSR, "vc/0");
- class_simple_device_add(tty_class, MKDEV(TTY_MAJOR, 0), NULL, "tty0");
+ if (console_use_vt) {
+ cdev_init(&vc0_cdev, &console_fops);
+ if (cdev_add(&vc0_cdev, MKDEV(TTY_MAJOR, 0), 1) ||
+ register_chrdev_region(MKDEV(TTY_MAJOR, 0), 1,
+ "/dev/vc/0") < 0)
+ panic("Couldn't register /dev/tty0 driver\n");
+ devfs_mk_cdev(MKDEV(TTY_MAJOR, 0), S_IFCHR|S_IRUSR|S_IWUSR,
+ "vc/0");
+ class_simple_device_add(tty_class, MKDEV(TTY_MAJOR, 0), NULL,
+ "tty0");

- vty_init();
+ vty_init();
+ }
#endif
return 0;
}

2004-11-18 03:00:36

by Andrew Morton

Subject: Re: [patch 3] Xen core patch : runtime VT console disable

Ian Pratt <[email protected]> wrote:
>
> --- pristine-linux-2.6.9/drivers/char/tty_io.c 2004-10-18 22:54:32.000000000 +0100
> +++ linux-2.6.9-xen-sparse/drivers/char/tty_io.c 2004-11-17 01:51:48.000000000 +0000
> @@ -131,6 +131,8 @@ LIST_HEAD(tty_drivers); /* linked list
> vt.c for deeply disgusting hack reasons */
> DECLARE_MUTEX(tty_sem);

This patch has had its tabs replaced by spaces.


> +int console_use_vt = 1;

Should this not have static scope?

2004-11-18 10:04:41

by Gerd Knorr

Subject: Re: [patch 3] Xen core patch : runtime VT console disable

Andrew Morton <[email protected]> writes:

> > +int console_use_vt = 1;
>
> Should this not have static scope?

I'd like to have that one globally visible, so that code somewhere in
arch/{xen|um} can enable/disable that at boot time depending on the
virtual machine configuration.

Gerd

--
#define printk(args...) fprintf(stderr, ## args)

2004-11-18 10:13:23

by Ian Pratt

Subject: Re: [patch 3] Xen core patch : runtime VT console disable

> Andrew Morton <[email protected]> writes:
>
> > > +int console_use_vt = 1;
> >
> > Should this not have static scope?
>
> I'd like to have that one globally visible, so that code somewhere in
> arch/{xen|um} can enable/disable that at boot time depending on the
> virtual machine configuration.

Yep, that's why we need it global. In
arch/xen/i386/kernel/setup.c we set it to zero if the virtual
machine doesn't have access to the console hardware.

Ian

2004-11-18 10:23:36

by Christoph Hellwig

Subject: Re: [patch 3] Xen core patch : runtime VT console disable

On Wed, Nov 17, 2004 at 11:51:14PM +0000, Ian Pratt wrote:
>
> This patch enables the VT console to be disabled at runtime even if it
> is built into the kernel. arch-xen needs this to avoid trying to
> initialise a VT in virtual machine that doesn't have access to the
> console hardware.

You should only need the conditional initialization - all the runtime
checks couldn't be reached anymore if the device isn't actually registered.