linux-kernel - Re: [PATCH v5 2/2] tracing: Add documentation for funcgraph-retval and graph_retval

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, 21 Mar 2023 11:24:44 +0800
From:   Donglin Peng <pengdonglin@...gfor.com.cn>
To:     Ding Hui <dinghui@...gfor.com.cn>, mhiramat@...nel.org,
        rostedt@...dmis.org, linux@...linux.org.uk, mark.rutland@....com,
        will@...nel.org, catalin.marinas@....com, palmer@...belt.com,
        paul.walmsley@...ive.com, tglx@...utronix.de,
        dave.hansen@...ux.intel.com, x86@...nel.org, mingo@...hat.com,
        xiehuan09@...il.com, huangcun@...gfor.com.cn,
        dolinux.peng@...il.com
Cc:     linux-trace-kernel@...r.kernel.org,
        linux-riscv@...ts.infradead.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 2/2] tracing: Add documentation for funcgraph-retval
 and graph_retval_hex

On 2023/3/21 10:31, Ding Hui wrote:
> On 2023/3/20 21:16, Donglin Peng wrote:
> 
>> +There are some limitations when using the funcgraph-retval currently:
>> +
>> +- Even if the function return type is void, a return value will still
>> +  be printed, and you can just ignore it.
>> +
>> +- Even if the return value is not an error code actually, it may be
>> +  displayed as an error code. You should read the code to check.
>> +  For example, both 0xfe and 0xfffe are be interpreted as -2.
> 
> For char and short types, displaying as signed decimal may be not 
> appropriate, because they are rarely used to store error code.
> 
> So in "smart" mode (graph_retval_hex=0), I suggest just smart convert 
> error value stored in int or pointer to signed decimal.

Yeah, I will remove the return value conversion for char and short
types in the "smart" mode.

> 
>> +- Only the value of the first return register will be recorded and
>> +  printed even if the return values may be stored in two registers
>> +  actually. For example, both the eax and edx are used to store a
>> +  64 bit return value in the x86 architecture, and the eax stores
>> +  the low 32 bit, the edx stores the high 32 bit, however only the
>> +  value stored in eax will be recorded and printed.
>> +
>>   You can put some comments on specific functions by using
>>   trace_printk() For example, if you want to put a comment inside
>>   the __might_sleep() function, you just have to include
>