[PATCH 0/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[PATCH 0/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[Oleg Nesterov]

Hello,

Steven, I am going to ask Ingo to pull this fix along with other
pending uprobes changes, but please let me know if you want to take
this patch.

Oleg.

 kernel/trace/trace_uprobe.c |   46 +++++++++++++++++++++++-------------------
 1 files changed, 25 insertions(+), 21 deletions(-)

[PATCH 1/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[Oleg Nesterov]

uprobe_perf_open()->uprobe_apply() can fail, but this error is wrongly
ignored. Change uprobe_perf_open() to do uprobe_perf_close() and return
the error code in this case.

Change uprobe_perf_close() to propogate the error from uprobe_apply()
as well, although it should not fail.

The patch looks more complicated because it moves uprobe_perf_close()
up to make it visible to uprobe_perf_open().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
---
 kernel/trace/trace_uprobe.c |   46 +++++++++++++++++++++++-------------------
 1 files changed, 25 insertions(+), 21 deletions(-)

diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
index 930e514..9aad3e2 100644
--- a/kernel/trace/trace_uprobe.c
+++ b/kernel/trace/trace_uprobe.c
@@ -1003,56 +1003,60 @@ uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
 	return __uprobe_perf_filter(&tu->filter, event->hw.tp_target->mm);
 }
 
-static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
+static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
 {
 	bool done;
 
 	write_lock(&tu->filter.rwlock);
 	if (event->hw.tp_target) {
-		/*
-		 * event->parent != NULL means copy_process(), we can avoid
-		 * uprobe_apply(). current->mm must be probed and we can rely
-		 * on dup_mmap() which preserves the already installed bp's.
-		 *
-		 * attr.enable_on_exec means that exec/mmap will install the
-		 * breakpoints we need.
-		 */
+		list_del(&event->hw.tp_list);
 		done = tu->filter.nr_systemwide ||
-			event->parent || event->attr.enable_on_exec ||
+			(event->hw.tp_target->flags & PF_EXITING) ||
 			uprobe_filter_event(tu, event);
-		list_add(&event->hw.tp_list, &tu->filter.perf_events);
 	} else {
+		tu->filter.nr_systemwide--;
 		done = tu->filter.nr_systemwide;
-		tu->filter.nr_systemwide++;
 	}
 	write_unlock(&tu->filter.rwlock);
 
 	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
+		return uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
 
 	return 0;
 }
 
-static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
+static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
 {
 	bool done;
+	int err;
 
 	write_lock(&tu->filter.rwlock);
 	if (event->hw.tp_target) {
-		list_del(&event->hw.tp_list);
+		/*
+		 * event->parent != NULL means copy_process(), we can avoid
+		 * uprobe_apply(). current->mm must be probed and we can rely
+		 * on dup_mmap() which preserves the already installed bp's.
+		 *
+		 * attr.enable_on_exec means that exec/mmap will install the
+		 * breakpoints we need.
+		 */
 		done = tu->filter.nr_systemwide ||
-			(event->hw.tp_target->flags & PF_EXITING) ||
+			event->parent || event->attr.enable_on_exec ||
 			uprobe_filter_event(tu, event);
+		list_add(&event->hw.tp_list, &tu->filter.perf_events);
 	} else {
-		tu->filter.nr_systemwide--;
 		done = tu->filter.nr_systemwide;
+		tu->filter.nr_systemwide++;
 	}
 	write_unlock(&tu->filter.rwlock);
 
-	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
-
-	return 0;
+	err = 0;
+	if (!done) {
+		err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
+		if (err)
+			uprobe_perf_close(tu, event);
+	}
+	return err;
 }
 
 static bool uprobe_perf_filter(struct uprobe_consumer *uc,
-- 
1.5.5.1

Re: [PATCH 1/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[Steven Rostedt]

On Wed, 23 Apr 2014 18:58:30 +0200
Oleg Nesterov <oleg@redhat.com> wrote:

> uprobe_perf_open()->uprobe_apply() can fail, but this error is wrongly
> ignored. Change uprobe_perf_open() to do uprobe_perf_close() and return
> the error code in this case.
> 
> Change uprobe_perf_close() to propogate the error from uprobe_apply()
> as well, although it should not fail.
> 
> The patch looks more complicated because it moves uprobe_perf_close()
> up to make it visible to uprobe_perf_open().
> 
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
>  kernel/trace/trace_uprobe.c |   46 +++++++++++++++++++++++-------------------
>  1 files changed, 25 insertions(+), 21 deletions(-)
> 
> diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
> index 930e514..9aad3e2 100644
> --- a/kernel/trace/trace_uprobe.c
> +++ b/kernel/trace/trace_uprobe.c
> @@ -1003,56 +1003,60 @@ uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
>  	return __uprobe_perf_filter(&tu->filter, event->hw.tp_target->mm);
>  }
>  
> -static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
> +static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)

Egad, this confused the heck out of me. I didn't notice the swap in
functions and was wondering what you were doing. I didn't realize this
is what you meant by moving the uprobe_perf_close() up. I was thinking
you moved the call up or something, not the function itself physically
in the file.

/me tries to continue dazed and confused.

>  {
>  	bool done;
>  
>  	write_lock(&tu->filter.rwlock);
>  	if (event->hw.tp_target) {
> -		/*
> -		 * event->parent != NULL means copy_process(), we can avoid
> -		 * uprobe_apply(). current->mm must be probed and we can rely
> -		 * on dup_mmap() which preserves the already installed bp's.
> -		 *
> -		 * attr.enable_on_exec means that exec/mmap will install the
> -		 * breakpoints we need.
> -		 */
> +		list_del(&event->hw.tp_list);
>  		done = tu->filter.nr_systemwide ||
> -			event->parent || event->attr.enable_on_exec ||
> +			(event->hw.tp_target->flags & PF_EXITING) ||
>  			uprobe_filter_event(tu, event);
> -		list_add(&event->hw.tp_list, &tu->filter.perf_events);
>  	} else {
> +		tu->filter.nr_systemwide--;
>  		done = tu->filter.nr_systemwide;
> -		tu->filter.nr_systemwide++;
>  	}
>  	write_unlock(&tu->filter.rwlock);
>  
>  	if (!done)
> -		uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
> +		return uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
>  
>  	return 0;
>  }
>  
> -static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
> +static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
>  {
>  	bool done;
> +	int err;
>  
>  	write_lock(&tu->filter.rwlock);
>  	if (event->hw.tp_target) {
> -		list_del(&event->hw.tp_list);
> +		/*
> +		 * event->parent != NULL means copy_process(), we can avoid
> +		 * uprobe_apply(). current->mm must be probed and we can rely
> +		 * on dup_mmap() which preserves the already installed bp's.
> +		 *
> +		 * attr.enable_on_exec means that exec/mmap will install the
> +		 * breakpoints we need.
> +		 */
>  		done = tu->filter.nr_systemwide ||
> -			(event->hw.tp_target->flags & PF_EXITING) ||
> +			event->parent || event->attr.enable_on_exec ||
>  			uprobe_filter_event(tu, event);
> +		list_add(&event->hw.tp_list, &tu->filter.perf_events);
>  	} else {
> -		tu->filter.nr_systemwide--;
>  		done = tu->filter.nr_systemwide;
> +		tu->filter.nr_systemwide++;
>  	}
>  	write_unlock(&tu->filter.rwlock);
>  
> -	if (!done)
> -		uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
> -
> -	return 0;
> +	err = 0;
> +	if (!done) {
> +		err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
> +		if (err)
> +			uprobe_perf_close(tu, event);
> +	}
> +	return err;

You can add by Acked-by, but next time, please make this into two
patches. One to do the move, the other to do the change.

Thanks!

-- Steve

>  }
>  
>  static bool uprobe_perf_filter(struct uprobe_consumer *uc,

Re: [PATCH 1/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[Srikar Dronamraju]

Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
(with 2 nits that you can ignore)

>  		done = tu->filter.nr_systemwide ||
> -			event->parent || event->attr.enable_on_exec ||
> +			(event->hw.tp_target->flags & PF_EXITING) ||
>  			uprobe_filter_event(tu, event);
> -		list_add(&event->hw.tp_list, &tu->filter.perf_events);
>  	} else {
> +		tu->filter.nr_systemwide--;
>  		done = tu->filter.nr_systemwide;
> -		tu->filter.nr_systemwide++;
>  	}

Nit: I think 2 lines can be made into
 done = --tu->filter.nr_systemwide;


<snipped>
>  		done = tu->filter.nr_systemwide ||
> -			(event->hw.tp_target->flags & PF_EXITING) ||
> +			event->parent || event->attr.enable_on_exec ||
>  			uprobe_filter_event(tu, event);
> +		list_add(&event->hw.tp_list, &tu->filter.perf_events);
>  	} else {
> -		tu->filter.nr_systemwide--;
>  		done = tu->filter.nr_systemwide;
> +		tu->filter.nr_systemwide++;
>  	}

Nit: Similarly lines can be made into
 done = tu->filter.nr_systemwide++;


-- 
Thanks and Regards
Srikar Dronamraju

[PATCH v2 0/2] uprobes/tracing: uprobe_perf_open()->uprobe_apply() fix

[Oleg Nesterov]

On 04/23, Steven Rostedt wrote:
>
> Egad, this confused the heck out of me. I didn't notice the swap in
> functions and was wondering what you were doing. I didn't realize this
> is what you meant by moving the uprobe_perf_close() up. I was thinking
> you moved the call up or something, not the function itself physically
> in the file.
...
>
> You can add by Acked-by, but next time, please make this into two
> patches. One to do the move, the other to do the change.

OK...

I tried to lessen the number of patches I have ;) But I agree, this simple
fix looks too complicated without preparation which only moves the code.

Let me split it then. Result is the same do I preserved the acks, this is
what I am going to add to my tree.

Thanks!

Oleg.

[PATCH v2 1/2] uprobes/tracing: Make uprobe_perf_close() visible to uprobe_perf_open()

[Oleg Nesterov]

Preparation. Move uprobe_perf_close() up before uprobe_perf_open() to
avoid the forward declaration in the next patch and make it readable.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
---
 kernel/trace/trace_uprobe.c |   36 ++++++++++++++++++------------------
 1 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
index 930e514..1ed0030 100644
--- a/kernel/trace/trace_uprobe.c
+++ b/kernel/trace/trace_uprobe.c
@@ -1003,54 +1003,54 @@ uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
 	return __uprobe_perf_filter(&tu->filter, event->hw.tp_target->mm);
 }
 
-static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
+static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
 {
 	bool done;
 
 	write_lock(&tu->filter.rwlock);
 	if (event->hw.tp_target) {
-		/*
-		 * event->parent != NULL means copy_process(), we can avoid
-		 * uprobe_apply(). current->mm must be probed and we can rely
-		 * on dup_mmap() which preserves the already installed bp's.
-		 *
-		 * attr.enable_on_exec means that exec/mmap will install the
-		 * breakpoints we need.
-		 */
+		list_del(&event->hw.tp_list);
 		done = tu->filter.nr_systemwide ||
-			event->parent || event->attr.enable_on_exec ||
+			(event->hw.tp_target->flags & PF_EXITING) ||
 			uprobe_filter_event(tu, event);
-		list_add(&event->hw.tp_list, &tu->filter.perf_events);
 	} else {
+		tu->filter.nr_systemwide--;
 		done = tu->filter.nr_systemwide;
-		tu->filter.nr_systemwide++;
 	}
 	write_unlock(&tu->filter.rwlock);
 
 	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
+		uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
 
 	return 0;
 }
 
-static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
+static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
 {
 	bool done;
 
 	write_lock(&tu->filter.rwlock);
 	if (event->hw.tp_target) {
-		list_del(&event->hw.tp_list);
+		/*
+		 * event->parent != NULL means copy_process(), we can avoid
+		 * uprobe_apply(). current->mm must be probed and we can rely
+		 * on dup_mmap() which preserves the already installed bp's.
+		 *
+		 * attr.enable_on_exec means that exec/mmap will install the
+		 * breakpoints we need.
+		 */
 		done = tu->filter.nr_systemwide ||
-			(event->hw.tp_target->flags & PF_EXITING) ||
+			event->parent || event->attr.enable_on_exec ||
 			uprobe_filter_event(tu, event);
+		list_add(&event->hw.tp_list, &tu->filter.perf_events);
 	} else {
-		tu->filter.nr_systemwide--;
 		done = tu->filter.nr_systemwide;
+		tu->filter.nr_systemwide++;
 	}
 	write_unlock(&tu->filter.rwlock);
 
 	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
+		uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
 
 	return 0;
 }
-- 
1.5.5.1

[PATCH v2 2/2] uprobes/tracing: Fix uprobe_perf_open() on uprobe_apply() failure

[Oleg Nesterov]

uprobe_perf_open()->uprobe_apply() can fail, but this error is wrongly
ignored. Change uprobe_perf_open() to do uprobe_perf_close() and return
the error code in this case.

Change uprobe_perf_close() to propogate the error from uprobe_apply()
as well, although it should not fail.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
---
 kernel/trace/trace_uprobe.c |   14 +++++++++-----
 1 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c
index 1ed0030..9aad3e2 100644
--- a/kernel/trace/trace_uprobe.c
+++ b/kernel/trace/trace_uprobe.c
@@ -1020,7 +1020,7 @@ static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
 	write_unlock(&tu->filter.rwlock);
 
 	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
+		return uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
 
 	return 0;
 }
@@ -1028,6 +1028,7 @@ static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
 static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
 {
 	bool done;
+	int err;
 
 	write_lock(&tu->filter.rwlock);
 	if (event->hw.tp_target) {
@@ -1049,10 +1050,13 @@ static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
 	}
 	write_unlock(&tu->filter.rwlock);
 
-	if (!done)
-		uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
-
-	return 0;
+	err = 0;
+	if (!done) {
+		err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
+		if (err)
+			uprobe_perf_close(tu, event);
+	}
+	return err;
 }
 
 static bool uprobe_perf_filter(struct uprobe_consumer *uc,
-- 
1.5.5.1

Re: [PATCH 1/1] uprobes/tracing: uprobe_perf_open() forgets to handle the error from uprobe_apply()

[Oleg Nesterov]

On 04/24, Srikar Dronamraju wrote:
>
> Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

Thanks, included!

> > +		tu->filter.nr_systemwide--;
> >  		done = tu->filter.nr_systemwide;
> > -		tu->filter.nr_systemwide++;
> >  	}
>
> Nit: I think 2 lines can be made into
>  done = --tu->filter.nr_systemwide;
...
> > -		tu->filter.nr_systemwide--;
> >  		done = tu->filter.nr_systemwide;
> > +		tu->filter.nr_systemwide++;
> >  	}
>
> Nit: Similarly lines can be made into
>  done = tu->filter.nr_systemwide++;

Yes, perhaps, but this has nothing to do with this patch, it does not
change this code.

And can't resist... you know, initially I wrote this code this way, but
then I decided to make it more straightforward to avoid the potential nits
from reviewers ;)

Oleg.

Re: [PATCH v2 0/2] uprobes/tracing: uprobe_perf_open()->uprobe_apply() fix

[Steven Rostedt]

On Thu, 24 Apr 2014 13:54:58 +0200
Oleg Nesterov <oleg@redhat.com> wrote:

> On 04/23, Steven Rostedt wrote:
> >
> > Egad, this confused the heck out of me. I didn't notice the swap in
> > functions and was wondering what you were doing. I didn't realize this
> > is what you meant by moving the uprobe_perf_close() up. I was thinking
> > you moved the call up or something, not the function itself physically
> > in the file.
> ...
> >
> > You can add by Acked-by, but next time, please make this into two
> > patches. One to do the move, the other to do the change.
> 
> OK...
> 
> I tried to lessen the number of patches I have ;) But I agree, this simple
> fix looks too complicated without preparation which only moves the code.

Yeah, sometimes a simple fix just seems wrong to break into two. But if
it helps in code review, it's definitely worth it.


> 
> Let me split it then. Result is the same do I preserved the acks, this is
> what I am going to add to my tree.

They look good (and much easier to review).

-- Steve