Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Received-SPF: pass (mx2e12: connection is authenticated)
Subject: Re: [PATCH] KCOV: Introduced tracing unique covered PCs
To:     Dmitry Vyukov <dvyukov@google.com>
Cc:     Andrey Konovalov <andreyknvl@google.com>,
        Jonathan Corbet <corbet@lwn.net>,
        Andrew Morton <akpm@linux-foundation.org>,
        Wei Yongjun <weiyongjun1@huawei.com>,
        Maciej Grochowski <maciej.grochowski@pm.me>,
        kasan-dev <kasan-dev@googlegroups.com>,
        "open list:DOCUMENTATION" <linux-doc@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        syzkaller <syzkaller@googlegroups.com>
References: <20210211080716.80982-1-info@alexander-lochmann.de>
 <CACT4Y+YwRE=YNQYmQ=7RWde33830YOYr5pEAoYbrofY2JG43MA@mail.gmail.com>
From:   Alexander Lochmann <info@alexander-lochmann.de>
Message-ID: <01a9177f-bfd5-251a-758f-d3c68bafd0cf@alexander-lochmann.de>
Date:   Sun, 14 Mar 2021 22:29:45 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.7.1
MIME-Version: 1.0
In-Reply-To: <CACT4Y+YwRE=YNQYmQ=7RWde33830YOYr5pEAoYbrofY2JG43MA@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: de-DE
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 12.02.21 13:54, Dmitry Vyukov wrote:
> 
> I think we could make KCOV_IN_CTXSW sign bit and then express the check as:
> 
> void foo2(unsigned mode) {
>   if (((int)(mode & 0x8000000a)) > 0)
>     foo();
> }
> 
> 0000000000000020 <foo2>:
>   20: 81 e7 0a 00 00 80    and    $0x8000000a,%edi
>   26: 7f 08                jg     30 <foo2+0x10>
>   28: c3                    retq
> 
So ((int)(mode & (KCOV_IN_CTXSW | needed_mode))) > 0?
> 
> 
> 
>>  }
>>
>>  static notrace unsigned long canonicalize_ip(unsigned long ip)
>> @@ -191,18 +192,26 @@ void notrace __sanitizer_cov_trace_pc(void)
>>         struct task_struct *t;
>>         unsigned long *area;
>>         unsigned long ip = canonicalize_ip(_RET_IP_);
>> -       unsigned long pos;
>> +       unsigned long pos, idx;
>>
>>         t = current;
>> -       if (!check_kcov_mode(KCOV_MODE_TRACE_PC, t))
>> +       if (!check_kcov_mode(KCOV_MODE_TRACE_PC | KCOV_MODE_UNIQUE_PC, t))
>>                 return;
>>
>>         area = t->kcov_area;
>> -       /* The first 64-bit word is the number of subsequent PCs. */
>> -       pos = READ_ONCE(area[0]) + 1;
>> -       if (likely(pos < t->kcov_size)) {
>> -               area[pos] = ip;
>> -               WRITE_ONCE(area[0], pos);
>> +       if (likely(t->kcov_mode == KCOV_MODE_TRACE_PC)) {
> 
> Does this introduce an additional real of t->kcov_mode?
> If yes, please reuse the value read in check_kcov_mode.
Okay. How do I get that value from check_kcov_mode() to the caller?
Shall I add an additional parameter to check_kcov_mode()?
> 
> 
>> +               /* The first 64-bit word is the number of subsequent PCs. */
>> +               pos = READ_ONCE(area[0]) + 1;
>> +               if (likely(pos < t->kcov_size)) {
>> +                       area[pos] = ip;
>> +                       WRITE_ONCE(area[0], pos);
>> +               }
>> +       } else {
>> +               idx = (ip - canonicalize_ip((unsigned long)&_stext)) / 4;
>> +               pos = idx % BITS_PER_LONG;
>> +               idx /= BITS_PER_LONG;
>> +               if (likely(idx < t->kcov_size))
>> +                       WRITE_ONCE(area[idx], READ_ONCE(area[idx]) | 1L << pos);
>>         }
>>  }
>>  EXPORT_SYMBOL(__sanitizer_cov_trace_pc);
>> @@ -474,6 +483,7 @@ static int kcov_mmap(struct file *filep, struct vm_area_struct *vma)
>>                 goto exit;
>>         }
>>         if (!kcov->area) {
>> +               kcov_debug("mmap(): Allocating 0x%lx bytes\n", size);
>>                 kcov->area = area;
>>                 vma->vm_flags |= VM_DONTEXPAND;
>>                 spin_unlock_irqrestore(&kcov->lock, flags);
>> @@ -515,6 +525,8 @@ static int kcov_get_mode(unsigned long arg)
>>  {
>>         if (arg == KCOV_TRACE_PC)
>>                 return KCOV_MODE_TRACE_PC;
>> +       else if (arg == KCOV_UNIQUE_PC)
>> +               return KCOV_MODE_UNIQUE_PC;
> 
> As far as I understand, users can first do KCOV_INIT_UNIQUE and then
> enable KCOV_TRACE_PC, or vice versa.
> It looks somewhat strange. Is it intentional? 
I'll fix that.
It's not possible to
> specify buffer size for KCOV_INIT_UNIQUE, so most likely the buffer
> will be either too large or too small for a trace.
No, the buffer will be calculated by KCOV_INIT_UNIQUE based on the size
of the text segment.

- Alex

-- 
Alexander Lochmann                PGP key: 0xBC3EF6FD
Heiliger Weg 72                   phone:  +49.231.28053964
D-44141 Dortmund                  mobile: +49.151.15738323