Skip to contentSkip to conversation

Move jffs2_gcd_mtd threads to the new kthread API

PATCH In review Gerard Lledo 3 participants 9 messages May 29, 2009 – Jun 25, 2009

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

PATCH 1/1 Move jffs2_gcd_mtd threads to the new kthread API

kernel_thread() is being depracated. This patch moves the jffs2 garbage collecting thread to the new kthread API with the minimal impact.

---
 fs/jffs2/background.c |   22 ++++++++++++++--------
 1 files changed, 14 insertions(+), 8 deletions(-)
fs/jffs2/background.c +14 -8
index e958010..077fa65 100644
@@ -15,6 +15,7 @@
 #include <linux/completion.h>
 #include <linux/sched.h>
 #include <linux/freezer.h>
+#include <linux/kthread.h>
 #include "nodelist.h"
 
 
@@ -31,7 +32,7 @@ void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c)
 /* This must only ever be called when no GC thread is currently running */
 int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 {
-	pid_t pid;
+	struct task_struct *tsk;
 	int ret = 0;
 
 	BUG_ON(c->gc_task);
@@ -39,15 +40,16 @@ int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 	init_completion(&c->gc_thread_start);
 	init_completion(&c->gc_thread_exit);
 
-	pid = kernel_thread(jffs2_garbage_collect_thread, c, CLONE_FS|CLONE_FILES);
-	if (pid < 0) {
-		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %d\n", -pid);
+	tsk = kthread_create(jffs2_garbage_collect_thread, c, "jffs2_gcd_mtd%d", c->mtd->index);
+	if (IS_ERR(tsk)) {
+		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %ld\n", -PTR_ERR(tsk));
 		complete(&c->gc_thread_exit);
-		ret = pid;
+		ret = PTR_ERR(tsk);
 	} else {
-		/* Wait for it... */
-		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", pid));
+		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", tsk->pid));
+		wake_up_process(tsk);
 		wait_for_completion(&c->gc_thread_start);
+		ret = tsk->pid;
 	}
 
 	return ret;
@@ -71,7 +73,6 @@ static int jffs2_garbage_collect_thread(void *_c)
 {
 	struct jffs2_sb_info *c = _c;
 
-	daemonize("jffs2_gcd_mtd%d", c->mtd->index);
 	allow_signal(SIGKILL);
 	allow_signal(SIGSTOP);
 	allow_signal(SIGCONT);
@@ -107,6 +108,11 @@ static int jffs2_garbage_collect_thread(void *_c)
 		 * the GC thread get there first. */
 		schedule_timeout_interruptible(msecs_to_jiffies(50));
 
+		if (kthread_should_stop()) {
+			D1(printk(KERN_DEBUG "jffs2_garbage_collect_thread():  kthread_stop() called.\n"));
+			goto die;
+		}
+
 		/* Put_super will send a SIGKILL and then wait on the sem.
 		 */
 		while (signal_pending(current) || freezing(current)) {
Signature · 1 line
1.5.6.5
#1 lore raw
Quoted 40 lines · On Fri, 29 May 2009 11:19:41 +0300 Gerard Lledo <[email protected]> wrote:

kernel_thread() is being depracated. This patch moves the jffs2 garbage collecting thread to the new kthread API with the minimal impact.

---
 fs/jffs2/background.c |   22 ++++++++++++++--------
 1 files changed, 14 insertions(+), 8 deletions(-)
fs/jffs2/background.c +5 -4
index e958010..077fa65 100644
@@ -15,6 +15,7 @@
 #include <linux/completion.h>
 #include <linux/sched.h>
 #include <linux/freezer.h>
+#include <linux/kthread.h>
 #include "nodelist.h"
 
 
@@ -31,7 +32,7 @@ void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c)
 /* This must only ever be called when no GC thread is currently running */
 int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 {
-	pid_t pid;
+	struct task_struct *tsk;
 	int ret = 0;
 
 	BUG_ON(c->gc_task);
@@ -39,15 +40,16 @@ int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 	init_completion(&c->gc_thread_start);
 	init_completion(&c->gc_thread_exit);
 
-	pid = kernel_thread(jffs2_garbage_collect_thread, c, CLONE_FS|CLONE_FILES);
-	if (pid < 0) {
-		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %d\n", -pid);
+	tsk = kthread_create(jffs2_garbage_collect_thread, c, "jffs2_gcd_mtd%d", c->mtd->index);
+	if (IS_ERR(tsk)) {
+		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %ld\n", -PTR_ERR(tsk));
 		complete(&c->gc_thread_exit);

fs/jffs2/background.c int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c) · line 53

Quoted 10 lines
fs/jffs2/background.c +4 -3
-		ret = pid;
+		ret = PTR_ERR(tsk);
 	} else {
-		/* Wait for it... */
-		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", pid));
+		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", tsk->pid));
+		wake_up_process(tsk);
 		wait_for_completion(&c->gc_thread_start);
+		ret = tsk->pid;
 	}

A simple kthread_run() would be preferable here. The code presently sort-of open-codes it.

fs/jffs2/background.c int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c) · line 55

Quoted 1 line
fs/jffs2/background.c +0 -0
 	return ret;

Also, manipulating kernel threads via their pids is atypical and inefficient. In-kernel it is better to refer to threads via their task_struct*.

Fortunately all callers of this function ignore its return value so I expect we can simply change it to be void-returning. But that's a cleanup which would best be done in a spearate patch.

Quoted 20 lines
fs/jffs2/background.c +5 -1
@@ -71,7 +73,6 @@ static int jffs2_garbage_collect_thread(void *_c)
 {
 	struct jffs2_sb_info *c = _c;
 
-	daemonize("jffs2_gcd_mtd%d", c->mtd->index);
 	allow_signal(SIGKILL);
 	allow_signal(SIGSTOP);
 	allow_signal(SIGCONT);
@@ -107,6 +108,11 @@ static int jffs2_garbage_collect_thread(void *_c)
 		 * the GC thread get there first. */
 		schedule_timeout_interruptible(msecs_to_jiffies(50));
 
+		if (kthread_should_stop()) {
+			D1(printk(KERN_DEBUG "jffs2_garbage_collect_thread():  kthread_stop() called.\n"));
+			goto die;
+		}
+
 		/* Put_super will send a SIGKILL and then wait on the sem.
 		 */
 		while (signal_pending(current) || freezing(current)) {
#2 lore raw
Quoted 59 lines · Andrew Morton
Quoted 40 lines · On Fri, 29 May 2009 11:19:41 +0300 Gerard Lledo <[email protected]> wrote:

kernel_thread() is being depracated. This patch moves the jffs2 garbage collecting thread to the new kthread API with the minimal impact.

---
 fs/jffs2/background.c |   22 ++++++++++++++--------
 1 files changed, 14 insertions(+), 8 deletions(-)
fs/jffs2/background.c +5 -4
index e958010..077fa65 100644
@@ -15,6 +15,7 @@
 #include <linux/completion.h>
 #include <linux/sched.h>
 #include <linux/freezer.h>
+#include <linux/kthread.h>
 #include "nodelist.h"
 
 
@@ -31,7 +32,7 @@ void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c)
 /* This must only ever be called when no GC thread is currently running */
 int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 {
-	pid_t pid;
+	struct task_struct *tsk;
 	int ret = 0;
 
 	BUG_ON(c->gc_task);
@@ -39,15 +40,16 @@ int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 	init_completion(&c->gc_thread_start);
 	init_completion(&c->gc_thread_exit);
 
-	pid = kernel_thread(jffs2_garbage_collect_thread, c, CLONE_FS|CLONE_FILES);
-	if (pid < 0) {
-		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %d\n", -pid);
+	tsk = kthread_create(jffs2_garbage_collect_thread, c, "jffs2_gcd_mtd%d", c->mtd->index);
+	if (IS_ERR(tsk)) {
+		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %ld\n", -PTR_ERR(tsk));
 		complete(&c->gc_thread_exit);

fs/jffs2/background.c int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c) · line 53

Quoted 10 lines
fs/jffs2/background.c +4 -3
-		ret = pid;
+		ret = PTR_ERR(tsk);
 	} else {
-		/* Wait for it... */
-		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", pid));
+		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", tsk->pid));
+		wake_up_process(tsk);
 		wait_for_completion(&c->gc_thread_start);
+		ret = tsk->pid;
 	}

A simple kthread_run() would be preferable here. The code presently sort-of open-codes it.

fs/jffs2/background.c int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c) · line 55

Quoted 1 line
fs/jffs2/background.c +0 -0
 	return ret;
Quoted 8 lines

Also, manipulating kernel threads via their pids is atypical and inefficient. In-kernel it is better to refer to threads via their task_struct*.

Fortunately all callers of this function ignore its return value so I expect we can simply change it to be void-returning. But that's a cleanup which would best be done in a spearate patch.

OK, I'll redo the previous patch with kthread_run() instead. I will also post a different patch to cleanup the return value and use void.

#3 lore raw

PATCH 1/1 Move jffs2_gcd_mtd threads to the new kthread API

kernel_thread() is being depracated. This patch moves the jffs2 garbage collecting thread to the new kthread API with the minimal impact.

---
 fs/jffs2/background.c |   20 +++++++++++++-------
 1 files changed, 13 insertions(+), 7 deletions(-)
fs/jffs2/background.c +13 -7
index e958010..3ff50da 100644
@@ -15,6 +15,7 @@
 #include <linux/completion.h>
 #include <linux/sched.h>
 #include <linux/freezer.h>
+#include <linux/kthread.h>
 #include "nodelist.h"
 
 
@@ -31,7 +32,7 @@ void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c)
 /* This must only ever be called when no GC thread is currently running */
 int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 {
-	pid_t pid;
+	struct task_struct *tsk;
 	int ret = 0;
 
 	BUG_ON(c->gc_task);
@@ -39,15 +40,16 @@ int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 	init_completion(&c->gc_thread_start);
 	init_completion(&c->gc_thread_exit);
 
-	pid = kernel_thread(jffs2_garbage_collect_thread, c, CLONE_FS|CLONE_FILES);
-	if (pid < 0) {
-		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %d\n", -pid);
+	tsk = kthread_run(jffs2_garbage_collect_thread, c, "jffs2_gcd_mtd%d", c->mtd->index);
+	if (IS_ERR(tsk)) {
+		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %ld\n", -PTR_ERR(tsk));
 		complete(&c->gc_thread_exit);
-		ret = pid;
+		ret = PTR_ERR(tsk);
 	} else {
 		/* Wait for it... */
-		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", pid));
+		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", tsk->pid));
 		wait_for_completion(&c->gc_thread_start);
+		ret = tsk->pid;
 	}
 
 	return ret;
@@ -71,7 +73,6 @@ static int jffs2_garbage_collect_thread(void *_c)
 {
 	struct jffs2_sb_info *c = _c;
 
-	daemonize("jffs2_gcd_mtd%d", c->mtd->index);
 	allow_signal(SIGKILL);
 	allow_signal(SIGSTOP);
 	allow_signal(SIGCONT);
@@ -107,6 +108,11 @@ static int jffs2_garbage_collect_thread(void *_c)
 		 * the GC thread get there first. */
 		schedule_timeout_interruptible(msecs_to_jiffies(50));
 
+		if (kthread_should_stop()) {
+			D1(printk(KERN_DEBUG "jffs2_garbage_collect_thread():  kthread_stop() called.\n"));
+			goto die;
+		}
+
 		/* Put_super will send a SIGKILL and then wait on the sem.
 		 */
 		while (signal_pending(current) || freezing(current)) {
Signature · 1 line
1.5.6.5
#4 lore raw

PATCH 1/1 jffs2_start_garbage_collect_thread() return value cleanup

There is no user of this return value in the kernel. Change it to return void instead.

---
 fs/jffs2/background.c |    7 +------
 fs/jffs2/os-linux.h   |    2 +-
 2 files changed, 2 insertions(+), 7 deletions(-)
fs/jffs2/background.c +1 -6
index 3ff50da..3ae7e5c 100644
@@ -30,10 +30,9 @@ void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c)
 }
 
 /* This must only ever be called when no GC thread is currently running */
-int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
+void jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 {
 	struct task_struct *tsk;
-	int ret = 0;
 
 	BUG_ON(c->gc_task);
 
@@ -44,15 +43,11 @@ int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c)
 	if (IS_ERR(tsk)) {
 		printk(KERN_WARNING "fork failed for JFFS2 garbage collect thread: %ld\n", -PTR_ERR(tsk));
 		complete(&c->gc_thread_exit);
-		ret = PTR_ERR(tsk);
 	} else {
 		/* Wait for it... */
 		D1(printk(KERN_DEBUG "JFFS2: Garbage collect thread is pid %d\n", tsk->pid));
 		wait_for_completion(&c->gc_thread_start);
-		ret = tsk->pid;
 	}
-
-	return ret;
 }
 
 void jffs2_stop_garbage_collect_thread(struct jffs2_sb_info *c)
fs/jffs2/os-linux.h +1 -1
index 2228380..f3184ac 100644
@@ -151,7 +151,7 @@ static inline void jffs2_erase_pending_trigger(struct jffs2_sb_info *c)
 }
 
 /* background.c */
-int jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c);
+void jffs2_start_garbage_collect_thread(struct jffs2_sb_info *c);
 void jffs2_stop_garbage_collect_thread(struct jffs2_sb_info *c);
 void jffs2_garbage_collect_trigger(struct jffs2_sb_info *c);
 
Signature · 1 line
1.5.6.5
#5 lore raw

Re: [PATCH] [JFFS2] jffs2_start_garbage_collect_thread() return value cleanup

Quoted 2 lines · On Tue, 2 Jun 2009 15:11:26 +0300 Gerard Lledo <[email protected]> wrote:

There is no user of this return value in the kernel. Change it to return void instead.

Both these patches were missing your signed-off-by:. As your first version of the first patch wasn't missing it, I added it to both these patches, OK?

#6 lore raw

Re: [PATCH] [JFFS2] jffs2_start_garbage_collect_thread() return value cleanup

Quoted 10 lines · Andrew Morton
Quoted 2 lines · On Tue, 2 Jun 2009 15:11:26 +0300 Gerard Lledo <[email protected]> wrote:

There is no user of this return value in the kernel. Change it to return void instead.

Both these patches were missing your signed-off-by:. As your first version of the first patch wasn't missing it, I added it to both these patches, OK?

Thats fine. I just forgot to add it. Thanks, and sorry for the trouble/noise.

maintainer
#7 lore raw

Re: [PATCH] [JFFS2] jffs2_start_garbage_collect_thread() return value cleanup

In reply to Gerard Lledo
Quoted 2 lines · Gerard Lledo

There is no user of this return value in the kernel. Change it to return void instead.

NAK. I hate this type of patch.

A function _should_ return an error value indicating success or failure, if there's _any_ chance that it (or a future rewrite of it) may fail.

It's up to the _callers_ to act on that result, or not, as they see fit.

If you don't do that, you end up with whole stacks of functions which neither handle nor propagate errors -- and it's a _complete_ pain when you later find that you _do_ have to make one of the innermost functions return an error.

Gerard, if you persist in sending this kind of patch, please could I ask you to send me a cloth doll of yourself? I would like to stick pins in it next time I have to retrofit error handling to a stack of functions that lack it.

It's not even as if this patch is going to give any real performance improvement. It's a single register load in a once-per-mount code path.

There _may_ be some places where this kind of patch is reasonable -- but the commit comment should have a clear explanation of why the return value will _never_ be needed, rather than merely observing that it is not _currently_ used. Or show that there is a _significant_ improvement obtained by doing so, perhaps.

Andrew, may I suggest that you look for such justification in future patches of this type?

Signature · 2 lines
David Woodhouse                            Open Source Technology Centre
[email protected]                              Intel Corporation
#8 lore raw

Re: [PATCH] [JFFS2] jffs2_start_garbage_collect_thread() return value cleanup

Quoted 10 lines · On Wed, 24 Jun 2009 11:09:01 +0100 David Woodhouse <[email protected]> wrote:
Quoted 2 lines · Gerard Lledo

There is no user of this return value in the kernel. Change it to return void instead.

NAK. I hate this type of patch.

A function _should_ return an error value indicating success or failure, if there's _any_ chance that it (or a future rewrite of it) may fail.

It's up to the _callers_ to act on that result, or not, as they see fit.

True.

Quoted 2 lines

Andrew, may I suggest that you look for such justification in future patches of this type?

eh, I sometimes don't even look at them. I just save them up in case of maintainer fumblage.

I might retain this one as a "jffs2 fails to check jffs2_start_garbage_collect_thread() return value" bug report :)