linux-kernel - Re: [PATCH] tracing/histogram: Fix UAF in destroy_hist

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 16 Nov 2021 22:47:03 -0500
From:   Steven Rostedt <rostedt@...dmis.org>
To:     Kalesh Singh <kaleshsingh@...gle.com>
Cc:     kernel-team@...roid.com, mhiramat@...nel.org, zanussi@...nel.org,
        kernel test robot <oliver.sang@...el.com>,
        Ingo Molnar <mingo@...hat.com>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] tracing/histogram: Fix UAF in destroy_hist_field()

On Tue, 16 Nov 2021 18:12:22 -0800
Kalesh Singh <kaleshsingh@...gle.com> wrote:

> Calling destroy_hist_field() on an expression will recursively free
> any operands associated with the expression. If during expression
> parsing the operands of the expression are already set when an error
> is encountered, there is no need to explicity free the operands. Doing
> so will result in destroy_hist_field() being called twice for the
> operands and lead to a use-after-free (UAF) error.
> 
> Fix this by only calling destroy_hist_field() for the expression if the
> operands are already set.
> 
> Signed-off-by: Kalesh Singh <kaleshsingh@...gle.com>
> Fixes: 8b5d46fd7a38 ("tracing/histogram: Optimize division by constants")
> Reported-by: kernel test robot <oliver.sang@...el.com>
> ---
>  kernel/trace/trace_events_hist.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/trace/trace_events_hist.c b/kernel/trace/trace_events_hist.c
> index 5ea2c9ec54a6..e3856eaf2ac3 100644
> --- a/kernel/trace/trace_events_hist.c
> +++ b/kernel/trace/trace_events_hist.c
> @@ -2669,7 +2669,7 @@ static struct hist_field *parse_expr(struct hist_trigger_data *hist_data,
>  		if (!divisor) {
>  			hist_err(file->tr, HIST_ERR_DIVISION_BY_ZERO, errpos(str));
>  			ret = -EDOM;
> -			goto free;
> +			goto free_expr;
>  		}
>  
>  		/*
> @@ -2709,7 +2709,7 @@ static struct hist_field *parse_expr(struct hist_trigger_data *hist_data,
>  		expr->type = kstrdup_const(operand1->type, GFP_KERNEL);
>  		if (!expr->type) {
>  			ret = -ENOMEM;
> -			goto free;
> +			goto free_expr;
>  		}
>  
>  		expr->name = expr_str(expr, 0);
> @@ -2719,6 +2719,7 @@ static struct hist_field *parse_expr(struct hist_trigger_data *hist_data,
>  free:
>  	destroy_hist_field(operand1, 0);
>  	destroy_hist_field(operand2, 0);
> +free_expr:
>  	destroy_hist_field(expr, 0);
>  
>  	return ERR_PTR(ret);
> 
> base-commit: 8ab774587903771821b59471cc723bba6d893942

Wouldn't this be a simpler and more robust fix?

diff --git a/kernel/trace/trace_events_hist.c b/kernel/trace/trace_events_hist.c
index 5ea2c9ec54a6..aab69b4ffe11 100644
--- a/kernel/trace/trace_events_hist.c
+++ b/kernel/trace/trace_events_hist.c
@@ -2717,8 +2717,10 @@ static struct hist_field *parse_expr(struct hist_trigger_data *hist_data,
 
 	return expr;
 free:
-	destroy_hist_field(operand1, 0);
-	destroy_hist_field(operand2, 0);
+	if (!expr || expr->operand[0] != operand1)
+		destroy_hist_field(operand1, 0);
+	if (!expr || expr->operand[1] != operand2)
+		destroy_hist_field(operand2, 0);
 	destroy_hist_field(expr, 0);
 
 	return ERR_PTR(ret);


I'm worried about the complexity of having to know where to free what,
and not just figuring it out at the end.

-- Steve