Skip to contentSkip to conversation

vt: Make SAK run in process context.

PATCH In review Eric W. Biederman 2 participants 4 messages Dec 11, 2006 – Dec 12, 2006

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 vt: Make SAK run in process context.

This defers SAK so we can use the normal console semaphore to order operations.

This removes the xchg operations that I used to attempt to attmically update struct pid, because of the strange locking used for SAK. With SAK using the normal console semaphore nothing special is needed.

---
 drivers/char/keyboard.c |   11 +++--------
 drivers/char/sysrq.c    |    7 ++++---
 drivers/char/vt_ioctl.c |   27 +++++++++++++++++++++++++--
 include/linux/vt_kern.h |    1 +
 4 files changed, 33 insertions(+), 13 deletions(-)
drivers/char/keyboard.c +3 -8
index 7a6c1c0..bd6912d 100644
@@ -595,15 +595,10 @@ static void fn_spawn_con(struct vc_data *vc)
 
 static void fn_SAK(struct vc_data *vc)
 {
-	struct tty_struct *tty = vc->vc_tty;
+	static DECLARE_WORK(SAK_work, deferred_SAK, NULL);
+	SAK_work.data = &vc_cons[fg_console];
 
-	/*
-	 * SAK should also work in all raw modes and reset
-	 * them properly.
-	 */
-	if (tty)
-		do_SAK(tty);
-	reset_vc(vc);
+	schedule_work(&SAK_work);
 }
 
 static void fn_null(struct vc_data *vc)
drivers/char/sysrq.c +4 -3
index 05810c8..b4f67a0 100644
@@ -61,9 +61,10 @@ static struct sysrq_key_op sysrq_loglevel_op = {
 #ifdef CONFIG_VT
 static void sysrq_handle_SAK(int key, struct tty_struct *tty)
 {
-	if (tty)
-		do_SAK(tty);
-	reset_vc(vc_cons[fg_console].d);
+	static DECLARE_WORK(SAK_work, deferred_SAK, NULL);
+	SAK_work.data = &vc_cons[fg_console];
+
+	schedule_work(&SAK_work);
 }
 static struct sysrq_key_op sysrq_SAK_op = {
 	.handler	= sysrq_handle_SAK,
drivers/char/vt_ioctl.c +25 -2
index 311493e..5ecc71e 100644
@@ -672,7 +672,8 @@ int vt_ioctl(struct tty_struct *tty, struct file * file,
 		vc->vt_mode = tmp;
 		/* the frsig is ignored, so we set it to 0 */
 		vc->vt_mode.frsig = 0;
-		put_pid(xchg(&vc->vt_pid, get_pid(task_pid(current))));
+		put_pid(vc->vt_pid);
+		vc->vt_pid = get_pid(task_pid(current));
 		/* no switch is required -- [email protected] */
 		vc->vt_newvt = -1;
 		release_console_sem();
@@ -1063,12 +1064,34 @@ void reset_vc(struct vc_data *vc)
 	vc->vt_mode.relsig = 0;
 	vc->vt_mode.acqsig = 0;
 	vc->vt_mode.frsig = 0;
-	put_pid(xchg(&vc->vt_pid, NULL));
+	put_pid(vc->vt_pid);
+	vc->vt_pid = NULL;
 	vc->vt_newvt = -1;
 	if (!in_interrupt())    /* Via keyboard.c:SAK() - akpm */
 		reset_palette(vc);
 }
 
+void deferred_SAK(void *data)
+{
+	struct vc *vc_con = data;
+	struct vc_data *vc;
+	struct tty_struct *tty;
+	
+	acquire_console_sem();
+	vc = vc_con->d;
+	if (vc) {
+		tty = vc->vc_tty;
+		/*
+		 * SAK should also work in all raw modes and reset
+		 * them properly.
+		 */
+		if (tty)
+			do_SAK(tty);
+		reset_vc(vc);
+	}
+	release_console_sem();
+}
+
 /*
  * Performs the back end of a vt switch
  */
include/linux/vt_kern.h +1 -0
index 37a1a41..a727896 100644
@@ -74,6 +74,7 @@ int con_copy_unimap(struct vc_data *dst_vc, struct vc_data *src_vc);
 int vt_waitactive(int vt);
 void change_console(struct vc_data *new_vc);
 void reset_vc(struct vc_data *vc);
+void deferred_SAK(void *data);
 
 /*
  * vc_screen.c shares this temporary buffer with the console write code so that
Signature · 1 line
1.4.4.1.g278f
#1 lore raw
Quoted 8 lines · On Mon, 11 Dec 2006 06:07:03 -0700 [email protected] (Eric W. Biederman)…

This defers SAK so we can use the normal console semaphore to order operations.

This removes the xchg operations that I used to attempt to attmically update struct pid, because of the strange locking used for SAK. With SAK using the normal console semaphore nothing special is needed.

This is all a bit smelly.

drivers/char/keyboard.c static void fn_spawn_con(struct vc_data *vc) · line 599

Quoted 12 lines
drivers/char/keyboard.c +2 -1
index 7a6c1c0..bd6912d 100644
@@ -595,15 +595,10 @@ static void fn_spawn_con(struct vc_data *vc)
 
 static void fn_SAK(struct vc_data *vc)
 {
-	struct tty_struct *tty = vc->vc_tty;
+	static DECLARE_WORK(SAK_work, deferred_SAK, NULL);
+	SAK_work.data = &vc_cons[fg_console];

static storage...

drivers/char/keyboard.c static void fn_spawn_con(struct vc_data *vc) · line 602

Quoted 6 lines
drivers/char/keyboard.c +1 -3
 static void sysrq_handle_SAK(int key, struct tty_struct *tty)
 {
-	if (tty)
-		do_SAK(tty);
-	reset_vc(vc_cons[fg_console].d);
+	static DECLARE_WORK(SAK_work, deferred_SAK, NULL);

More.

Quoted 11 lines
drivers/char/keyboard.c +10 -0
+void deferred_SAK(void *data)
+{
+	struct vc *vc_con = data;
+	struct vc_data *vc;
+	struct tty_struct *tty;
+	
+	acquire_console_sem();
+	vc = vc_con->d;
+	if (vc) {
+		tty = vc->vc_tty;

drivers/char/keyboard.c static void fn_spawn_con(struct vc_data *vc)

Quoted 10 lines
drivers/char/keyboard.c +10 -0
+		/*
+		 * SAK should also work in all raw modes and reset
+		 * them properly.
+		 */
+		if (tty)
+			do_SAK(tty);
+		reset_vc(vc);
+	}
+	release_console_sem();
+}

And a workqueue callback which calls a function which immediately does another schedule_work().

I suspect you can fix all of this by passing a function pointer into do_SAK(): to either __do_SAK or to some new function which does the vc lookup then calls __do_SAK().

It probably means that you'll need to pass some payload into the workqueue callback, and dhowells just went and broke that on us. That can be fixed by adding a new `void *tty_struct.SAK_work_data'.

hmm, do_SAK() is being a bit bad, overwriting the ->SAK_work on a work_struct which might presently be scheduled. To do this safely we need a new variant on queue_work():

int queue_work_with_data(struct workqueue_struct *wq,
struct work_struct *work, void **datap, void *data

{
	int ret = 0, cpu = get_cpu();

	if (!test_and_set_bit(WORK_STRUCT_PENDING, &work->management)) {
		if (datap)
			*datap = data;
		if (unlikely(is_single_threaded(wq)))
			cpu = singlethread_cpu;
		BUG_ON(!list_empty(&work->entry));
		__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
		ret = 1;
	}
	put_cpu();
	return ret;
}

then, of course,

int queue_work(struct workqueue_struct *wq, struct work_struct *work)

{
	return queue_work_with_data(wq, work, NULL, NULL);
}

iirc, other places in the kernel need queue_work_with_data(), for removal of the *_WORK_NAR() stuff.

#2 lore raw
Quoted 7 lines · Andrew Morton
Quoted 4 lines · On Mon, 11 Dec 2006 06:07:03 -0700 [email protected] (Eric W. Biederman)…

This defers SAK so we can use the normal console semaphore to order operations.

Quoted 6 lines
Quoted 4 lines

This removes the xchg operations that I used to attempt to attmically update struct pid, because of the strange locking used for SAK. With SAK using the normal console semaphore nothing special is needed.

This is all a bit smelly.

Ok. I will take a second look, thanks for catching this.

I think I was half blind when I prepared this patch, I missed that do_SAK was scheduling work itself.

Quoted 25 lines
Quoted 11 lines
drivers/char/vt_ioctl.c +10 -0
+void deferred_SAK(void *data)
+{
+	struct vc *vc_con = data;
+	struct vc_data *vc;
+	struct tty_struct *tty;
+	
+	acquire_console_sem();
+	vc = vc_con->d;
+	if (vc) {
+		tty = vc->vc_tty;

drivers/char/vt_ioctl.c void reset_vc(struct vc_data *vc) · line 1093

Quoted 10 lines
drivers/char/vt_ioctl.c +10 -0
+		/*
+		 * SAK should also work in all raw modes and reset
+		 * them properly.
+		 */
+		if (tty)
+			do_SAK(tty);
+		reset_vc(vc);
+	}
+	release_console_sem();
+}

And a workqueue callback which calls a function which immediately does another schedule_work().

Quoted 3 lines

I suspect you can fix all of this by passing a function pointer into do_SAK(): to either __do_SAK or to some new function which does the vc lookup then calls __do_SAK().

Yes. It looks like all I need is an appropriate factor of __do_SAK() that I can call immediately.

Quoted 8 lines

It probably means that you'll need to pass some payload into the workqueue callback, and dhowells just went and broke that on us. That can be fixed by adding a new `void *tty_struct.SAK_work_data'.

hmm, do_SAK() is being a bit bad, overwriting the ->SAK_work on a work_struct which might presently be scheduled. To do this safely we need a new variant on queue_work():

And of course there is the truly silly issue that X spells uses Ctrl-Alt-Backspace instead of the kernel provided SAK to implement this.

Regardless that looks right. Unless there is some locking on the tty we can exploit.

Quoted 20 lines

int queue_work_with_data(struct workqueue_struct *wq,
struct work_struct *work, void **datap, void *data

{
	int ret = 0, cpu = get_cpu();

	if (!test_and_set_bit(WORK_STRUCT_PENDING, &work->management)) {
		if (datap)
			*datap = data;
		if (unlikely(is_single_threaded(wq)))
			cpu = singlethread_cpu;
		BUG_ON(!list_empty(&work->entry));
		__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
		ret = 1;
	}
	put_cpu();
	return ret;
}

then, of course,

Quoted 7 lines

int queue_work(struct workqueue_struct *wq, struct work_struct *work)

{
	return queue_work_with_data(wq, work, NULL, NULL);
}

iirc, other places in the kernel need queue_work_with_data(), for removal of the *_WORK_NAR() stuff.

Wow. The intersection of the clean ups.

Eric

#3 lore raw

PATCH 1/1 vt: Refactor console SAK processing.

In reply to Andrew Morton

This does several things.
- It moves looking up of the current foreground console into process context where we can safely take the semaphore that protects this operation.
- It uses the new flavor of work queue processing.
- This generates a factor of do_SAK, __do_SAK that runs immediately.
- This calls __do_SAK with the console semaphore held ensuring nothing else happens to the console while we process the SAK operation.
- With the console SAK processing moved into process context this patch removes the xchg operations that I used to attempt to attomically update struct pid, because of the strange locking used in the SAK processing. With SAK using the normal console semaphore nothing special is needed.

---
 drivers/char/keyboard.c        |   12 +++---------
 drivers/char/sysrq.c           |    6 +++---
 drivers/char/tty_io.c          |   13 +++++++++----
 drivers/char/vt_ioctl.c        |   28 ++++++++++++++++++++++++++--
 include/linux/console_struct.h |    3 +++
 include/linux/tty.h            |    1 +
 6 files changed, 45 insertions(+), 18 deletions(-)
drivers/char/keyboard.c +3 -9
index 7a6c1c0..c654a3e 100644
@@ -595,15 +595,9 @@ static void fn_spawn_con(struct vc_data *vc)
 
 static void fn_SAK(struct vc_data *vc)
 {
-	struct tty_struct *tty = vc->vc_tty;
-
-	/*
-	 * SAK should also work in all raw modes and reset
-	 * them properly.
-	 */
-	if (tty)
-		do_SAK(tty);
-	reset_vc(vc);
+	struct work_struct *SAK_work = &vc_cons[fg_console].SAK_work;
+	PREPARE_WORK(SAK_work, vc_SAK);
+	schedule_work(SAK_work);
 }
 
 static void fn_null(struct vc_data *vc)
drivers/char/sysrq.c +3 -3
index 05810c8..6cf877f 100644
@@ -61,9 +61,9 @@ static struct sysrq_key_op sysrq_loglevel_op = {
 #ifdef CONFIG_VT
 static void sysrq_handle_SAK(int key, struct tty_struct *tty)
 {
-	if (tty)
-		do_SAK(tty);
-	reset_vc(vc_cons[fg_console].d);
+	struct work_struct *SAK_work = &vc_cons[fg_console].SAK_work;
+	PREPARE_WORK(SAK_work, vc_SAK);
+	schedule_work(SAK_work);
 }
 static struct sysrq_key_op sysrq_SAK_op = {
 	.handler	= sysrq_handle_SAK,
drivers/char/tty_io.c +9 -4
index 4044c86..a333881 100644
@@ -3324,10 +3324,8 @@ int tty_ioctl(struct inode * inode, struct file * file,
  * Nasty bug: do_SAK is being called in interrupt context.  This can
  * deadlock.  We punt it up to process context.  AKPM - 16Mar2001
  */
-static void __do_SAK(struct work_struct *work)
+void __do_SAK(struct tty_struct *tty)
 {
-	struct tty_struct *tty =
-		container_of(work, struct tty_struct, SAK_work);
 #ifdef TTY_SOFT_SAK
 	tty_hangup(tty);
 #else
@@ -3399,6 +3397,13 @@ static void __do_SAK(struct work_struct *work)
 #endif
 }
 
+static void do_SAK_work(struct work_struct *work)
+{
+	struct tty_struct *tty =
+		container_of(work, struct tty_struct, SAK_work);
+	__do_SAK(tty);
+}
+
 /*
  * The tq handling here is a little racy - tty->SAK_work may already be queued.
  * Fortunately we don't need to worry, because if ->SAK_work is already queued,
@@ -3409,7 +3414,7 @@ void do_SAK(struct tty_struct *tty)
 {
 	if (!tty)
 		return;
-	PREPARE_WORK(&tty->SAK_work, __do_SAK);
+	PREPARE_WORK(&tty->SAK_work, do_SAK_work);
 	schedule_work(&tty->SAK_work);
 }
 
drivers/char/vt_ioctl.c +26 -2
index 311493e..8384846 100644
@@ -672,7 +672,8 @@ int vt_ioctl(struct tty_struct *tty, struct file * file,
 		vc->vt_mode = tmp;
 		/* the frsig is ignored, so we set it to 0 */
 		vc->vt_mode.frsig = 0;
-		put_pid(xchg(&vc->vt_pid, get_pid(task_pid(current))));
+		put_pid(vc->vt_pid);
+		vc->vt_pid = get_pid(task_pid(current));
 		/* no switch is required -- [email protected] */
 		vc->vt_newvt = -1;
 		release_console_sem();
@@ -1063,12 +1064,35 @@ void reset_vc(struct vc_data *vc)
 	vc->vt_mode.relsig = 0;
 	vc->vt_mode.acqsig = 0;
 	vc->vt_mode.frsig = 0;
-	put_pid(xchg(&vc->vt_pid, NULL));
+	put_pid(vc->vt_pid);
+	vc->vt_pid = NULL;
 	vc->vt_newvt = -1;
 	if (!in_interrupt())    /* Via keyboard.c:SAK() - akpm */
 		reset_palette(vc);
 }
 
+void vc_SAK(struct work_struct *work)
+{
+	struct vc *vc_con =
+		container_of(work, struct vc, SAK_work);
+	struct vc_data *vc;
+	struct tty_struct *tty;
+	
+	acquire_console_sem();
+	vc = vc_con->d;
+	if (vc) {
+		tty = vc->vc_tty;
+		/*
+		 * SAK should also work in all raw modes and reset
+		 * them properly.
+		 */
+		if (tty)
+			__do_SAK(tty);
+		reset_vc(vc);
+	}
+	release_console_sem();
+}
+
 /*
  * Performs the back end of a vt switch
  */
include/linux/console_struct.h +3 -0
index ed6c0fe..a86162b 100644
@@ -11,6 +11,7 @@
 
 #include <linux/wait.h>
 #include <linux/vt.h>
+#include <linux/workqueue.h>
 
 struct vt_struct;
 
@@ -103,6 +104,7 @@ struct vc_data {
 
 struct vc {
 	struct vc_data *d;
+	struct work_struct SAK_work;
 
 	/* might add  scrmem, vt_struct, kbd  at some time,
 	   to have everything in one place - the disadvantage
@@ -110,6 +112,7 @@ struct vc {
 };
 
 extern struct vc vc_cons [MAX_NR_CONSOLES];
+extern void vc_SAK(struct work_struct *work);
 
 #define CUR_DEF		0
 #define CUR_NONE	1
include/linux/tty.h +1 -0
index 65cbcf2..0161a8c 100644
@@ -291,6 +291,7 @@ extern void tty_vhangup(struct tty_struct * tty);
 extern void tty_unhangup(struct file *filp);
 extern int tty_hung_up_p(struct file * filp);
 extern void do_SAK(struct tty_struct *tty);
+extern void __do_SAK(struct tty_struct *tty);
 extern void disassociate_ctty(int priv);
 extern void tty_flip_buffer_push(struct tty_struct *tty);
 extern speed_t tty_get_baud_rate(struct tty_struct *tty);
Signature · 1 line
1.4.4.1.g278f