linux-kernel - Re: [PATCH] tracing: Fix synth event printk format for str fields

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] [day] [month] [year] [list]

Message-ID: <9d234136-05b9-46b6-849f-901df83bf54c@arm.com>
Date: Wed, 19 Mar 2025 14:34:19 +0000
From: Douglas Raillard <douglas.raillard@....com>
To: "Masami Hiramatsu (Google)" <mhiramat@...nel.org>
Cc: rostedt@...dmis.org, Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
 linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org
Subject: Re: [PATCH] tracing: Fix synth event printk format for str fields

Hi,

On 19-03-2025 12:18, Masami Hiramatsu (Google) wrote:
> Hi,
> 
> On Tue, 18 Mar 2025 18:09:38 +0000
> Douglas RAILLARD <douglas.raillard@....com> wrote:
> 
>> From: Douglas Raillard <douglas.raillard@....com>
>>
>> The printk format for synth event uses "%.*s" to print string fields,
>> but then only passes the pointer part as var arg.
>>
>> Add the missing precision var arg.
> 
> I'm not sure what you want to. Would you mean showing the string length too?
> But I think actual output(*) shows only string, right?
> 
> (*) the output data which can be read from `trace` file.

The "%.*s" printf/printk format specifier has 2 components:
1. the "*" part that in printf documentation called the "precision"
2. the "s" part that is called the "specifier"

For a "s" specifier, the precision is interpreted as the number of chars to display from that string.
In any case, using "*" as precision means the value is passed dynamically, as a parameter to printf/printk,
before the actual char *. Both those args are consumed by printk to display "%.*s", the precision is never
displayed on its own.

In the current state, synthetic event use "%.*s" for string fields, but then only a single `__get_str(field)`
is passed to printk, so it's missing the precision arg. Both trace-cmd and our parser report an error because of that.

The alternative would be to just use "%s" instead of "%.*s", but I assume whoever wrote the code initially had a reason
to use the precision (maybe to print non-null-terminated strings ?), so I simply fixed the arguments.

> 
> Thank you,
> 
>>
>> Signed-off-by: Douglas Raillard <douglas.raillard@....com>
>> ---
>>   kernel/trace/trace_events_synth.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/kernel/trace/trace_events_synth.c b/kernel/trace/trace_events_synth.c
>> index e3f7d09e5512..274b9b1d9f7d 100644
>> --- a/kernel/trace/trace_events_synth.c
>> +++ b/kernel/trace/trace_events_synth.c
>> @@ -620,7 +620,8 @@ static int __set_synth_event_print_fmt(struct synth_event *event,
>>   		if (event->fields[i]->is_string &&
>>   		    event->fields[i]->is_dynamic)
>>   			pos += snprintf(buf + pos, LEN_OR_ZERO,
>> -				", __get_str(%s)", event->fields[i]->name);
>> +				", (int)__get_dynamic_array_len(%s), __get_str(%s)",
>> +				event->fields[i]->name, event->fields[i]->name);
>>   		else if (event->fields[i]->is_stack)
>>   			pos += snprintf(buf + pos, LEN_OR_ZERO,
>>   				", __get_stacktrace(%s)", event->fields[i]->name);
>> -- 
>> 2.43.0
>>