All of lore.kernel.org
 help / color / mirror / Atom feed
From: Steven Rostedt <rostedt@goodmis.org>
To: Breno Leitao <leitao@debian.org>
Cc: mark.rutland@arm.com, mhiramat@kernel.org,
	linux-trace-kernel@vger.kernel.org
Subject: Re: ftrace: UAF in ftrace_regex_open() on a removed tracing instance
Date: Fri, 28 Aug 2026 12:34:30 -0400	[thread overview]
Message-ID: <20260828123430.22d43b44@gandalf.local.home> (raw)
In-Reply-To: <apGwPEj65dcbZFxS@gmail.com>

On Fri, 28 Aug 2026 08:59:39 -0700
Breno Leitao <leitao@debian.org> wrote:

> 
> My LLM came up with something like the following, and it doesn't
> reproduce the problem anymore. Very similar to your code above:

I know the solution but there's subtle things I want to make sure that
works.

> 
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index f9d80c7bd9f16..37845399fe5cf 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -4788,22 +4788,37 @@ ftrace_regex_open(struct ftrace_ops *ops, int flag,
>  static int
>  ftrace_filter_open(struct inode *inode, struct file *file)
>  {
> -	struct ftrace_ops *ops = inode->i_private;
> +	struct trace_array *tr = inode->i_private;
> +	int ret;
>  
> -	/* Checks for tracefs lockdown */
> -	return ftrace_regex_open(ops,
> -			FTRACE_ITER_FILTER | FTRACE_ITER_DO_PROBES,
> -			inode, file);
> +	/* Checks for tracefs lockdown, and that the instance is still alive */
> +	ret = tracing_check_open_get_tr(tr);
> +	if (ret)
> +		return ret;
> +
> +	ret = ftrace_regex_open(tr->ops,
> +				FTRACE_ITER_FILTER | FTRACE_ITER_DO_PROBES,
> +				inode, file);
> +	/* ftrace_regex_open() holds its own reference on success */
> +	trace_array_put(tr);
> +	return ret;
>  }
>  
>  static int
>  ftrace_notrace_open(struct inode *inode, struct file *file)
>  {
> -	struct ftrace_ops *ops = inode->i_private;
> +	struct trace_array *tr = inode->i_private;
> +	int ret;
> +
> +	/* Checks for tracefs lockdown, and that the instance is still alive */
> +	ret = tracing_check_open_get_tr(tr);
> +	if (ret)
> +		return ret;
>  
> -	/* Checks for tracefs lockdown */
> -	return ftrace_regex_open(ops, FTRACE_ITER_NOTRACE,
> -				 inode, file);
> +	ret = ftrace_regex_open(tr->ops, FTRACE_ITER_NOTRACE, inode, file);
> +	/* ftrace_regex_open() holds its own reference on success */
> +	trace_array_put(tr);
> +	return ret;
>  }
>  
>  /* Type for quick search ftrace basic regexes (globs) from filter_parse_regex */
> @@ -7496,11 +7511,15 @@ void ftrace_create_filter_files(struct ftrace_ops *ops,
>  				struct dentry *parent)
>  {
>  
> +	/*
> +	 * Pass the trace array and not @ops, as @ops is freed when the
> +	 * instance is removed, but the trace array can still be validated.
> +	 */

I hate these comments. It's really pointless to state this. It's a "no shit" comment ;-)


>  	trace_create_file("set_ftrace_filter", TRACE_MODE_WRITE, parent,
> -			  ops, &ftrace_filter_fops);
> +			  ops->private, &ftrace_filter_fops);

But this also assumes that ops->private is always a trace_array. Which may
someday change and we will get a hidden bug if it does.

-- Steve


>  
>  	trace_create_file("set_ftrace_notrace", TRACE_MODE_WRITE, parent,
> -			  ops, &ftrace_notrace_fops);
> +			  ops->private, &ftrace_notrace_fops);
>  }
>  
>  /*


WARNING: multiple messages have this Message-ID (diff)
From: Steven Rostedt <rostedt@goodmis.org>
To: Breno Leitao <leitao@debian.org>
Cc: mark.rutland@arm.com, mhiramat@kernel.org,
	linux-trace-kernel@vger.kernel.org
Subject: Re: ftrace: UAF in ftrace_regex_open() on a removed tracing instance
Date: Fri, 28 Aug 2026 12:34:52 -0400	[thread overview]
Message-ID: <20260828123430.22d43b44@gandalf.local.home> (raw)
Message-ID: <20260828163452.FSU7ejbJe9onK4amKMjmoCOJqemfXwQZjCrHsdy33eY@z> (raw)
In-Reply-To: <apGwPEj65dcbZFxS@gmail.com>

On Fri, 28 Aug 2026 08:59:39 -0700
Breno Leitao <leitao@debian.org> wrote:

> 
> My LLM came up with something like the following, and it doesn't
> reproduce the problem anymore. Very similar to your code above:

I know the solution but there's subtle things I want to make sure that
works.

> 
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index f9d80c7bd9f16..37845399fe5cf 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -4788,22 +4788,37 @@ ftrace_regex_open(struct ftrace_ops *ops, int flag,
>  static int
>  ftrace_filter_open(struct inode *inode, struct file *file)
>  {
> -	struct ftrace_ops *ops = inode->i_private;
> +	struct trace_array *tr = inode->i_private;
> +	int ret;
>  
> -	/* Checks for tracefs lockdown */
> -	return ftrace_regex_open(ops,
> -			FTRACE_ITER_FILTER | FTRACE_ITER_DO_PROBES,
> -			inode, file);
> +	/* Checks for tracefs lockdown, and that the instance is still alive */
> +	ret = tracing_check_open_get_tr(tr);
> +	if (ret)
> +		return ret;
> +
> +	ret = ftrace_regex_open(tr->ops,
> +				FTRACE_ITER_FILTER | FTRACE_ITER_DO_PROBES,
> +				inode, file);
> +	/* ftrace_regex_open() holds its own reference on success */
> +	trace_array_put(tr);
> +	return ret;
>  }
>  
>  static int
>  ftrace_notrace_open(struct inode *inode, struct file *file)
>  {
> -	struct ftrace_ops *ops = inode->i_private;
> +	struct trace_array *tr = inode->i_private;
> +	int ret;
> +
> +	/* Checks for tracefs lockdown, and that the instance is still alive */
> +	ret = tracing_check_open_get_tr(tr);
> +	if (ret)
> +		return ret;
>  
> -	/* Checks for tracefs lockdown */
> -	return ftrace_regex_open(ops, FTRACE_ITER_NOTRACE,
> -				 inode, file);
> +	ret = ftrace_regex_open(tr->ops, FTRACE_ITER_NOTRACE, inode, file);
> +	/* ftrace_regex_open() holds its own reference on success */
> +	trace_array_put(tr);
> +	return ret;
>  }
>  
>  /* Type for quick search ftrace basic regexes (globs) from filter_parse_regex */
> @@ -7496,11 +7511,15 @@ void ftrace_create_filter_files(struct ftrace_ops *ops,
>  				struct dentry *parent)
>  {
>  
> +	/*
> +	 * Pass the trace array and not @ops, as @ops is freed when the
> +	 * instance is removed, but the trace array can still be validated.
> +	 */

I hate these comments. It's really pointless to state this. It's a "no shit" comment ;-)


>  	trace_create_file("set_ftrace_filter", TRACE_MODE_WRITE, parent,
> -			  ops, &ftrace_filter_fops);
> +			  ops->private, &ftrace_filter_fops);

But this also assumes that ops->private is always a trace_array. Which may
someday change and we will get a hidden bug if it does.

This is the exact reason I want to look deeper at it.

-- Steve


>  
>  	trace_create_file("set_ftrace_notrace", TRACE_MODE_WRITE, parent,
> -			  ops, &ftrace_notrace_fops);
> +			  ops->private, &ftrace_notrace_fops);
>  }
>  
>  /*


  reply	other threads:[~2026-08-28 16:33 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-08-28 13:38 ftrace: UAF in ftrace_regex_open() on a removed tracing instance Breno Leitao
2026-08-28 14:42 ` Steven Rostedt
2026-08-28 15:59   ` Breno Leitao
2026-08-28 16:34     ` Steven Rostedt [this message]
2026-08-28 16:34       ` Steven Rostedt
2026-08-28 17:49       ` Steven Rostedt
2026-08-28 18:04     ` Steven Rostedt
2026-08-28 19:55       ` Steven Rostedt

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260828123430.22d43b44@gandalf.local.home \
    --to=rostedt@goodmis.org \
    --cc=leitao@debian.org \
    --cc=linux-trace-kernel@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=mhiramat@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.