Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Sun, 20 Nov 2022 14:38:52 -0500
From:   Steven Rostedt <rostedt@goodmis.org>
To:     Yang Jihong <yangjihong1@huawei.com>
Cc:     <mhiramat@kernel.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2] tracing: Fix infinite loop in tracing_read_pipe on
 overflowed print_trace_line
Message-ID: <20221120143852.7e84b50d@rorschach.local.home>
In-Reply-To: <20221118102521.62362-1-yangjihong1@huawei.com>
References: <20221118102521.62362-1-yangjihong1@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Fri, 18 Nov 2022 18:25:21 +0800
Yang Jihong <yangjihong1@huawei.com> wrote:

> print_trace_line may overflow seq_file buffer. If the event is not
> consumed, the while loop keeps peeking this event, causing a infinite loop.
> 
> Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
> ---
> 
> Changes since v1:
>   - Print partial line to show the broken trace event when overflowed print_trace_line
> 
>  kernel/trace/trace.c | 27 ++++++++++++++++++++++++++-
>  1 file changed, 26 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index 47a44b055a1d..81c36dc80212 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -6786,7 +6786,32 @@ tracing_read_pipe(struct file *filp, char __user *ubuf,
>  
>  		ret = print_trace_line(iter);
>  		if (ret == TRACE_TYPE_PARTIAL_LINE) {
> -			/* don't print partial lines */
> +			/*
> +			 * If one trace_line of the tracer overflows seq_file
> +			 * buffer, trace_seq_to_user returns -EBUSY.
> +			 * In this case, we need to consume, otherwise,
> +			 * while loop will peek this event next time,
> +			 * resulting in an infinite loop.
> +			 */
> +			if (trace_seq_has_overflowed(&iter->seq)) {

We need to only do this if save_len is zero. Because the reason that it
returns TRACE_TYPE_PARTIAL_LINE is usually because it overflowed.

This loops until the trace_seq is full, so it's OK to have it overflow.
The case I believe you are fixing, is the case were one
print_trace_line() actually fills the entire trace_seq in one shot. In
which case, it will never print, and in that case, save_len will be zero.

-- Steve


> +				/*
> +				 * Print the partial line,
> +				 * that way we can see the broken trace event.
> +				 */
> +				char dots[] = "...\n";
> +
> +				iter->seq.full = 0;
> +				if (seq_buf_buffer_left(&iter->seq.seq) < strlen(dots)) {
> +					iter->seq.seq.len =
> +						seq_buf_used(&iter->seq.seq) - strlen(dots);
> +				}
> +				trace_seq_puts(&iter->seq, dots);
> +
> +				trace_consume(iter);
> +				break;
> +			}
> +
> +			/* In other cases, don't print partial lines */
>  			iter->seq.seq.len = save_len;
>  			break;
>  		}