From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jan Kiszka <jan.kiszka@siemens.com>
References: <CAOuPNLg60S+am654fq55vTMoFV1=PfMp3as5MZtt5BHUacFRBQ@mail.gmail.com>
 <2ee8fa04-7ad4-58d5-603b-1d41ad2533ec@siemens.com>
 <CAOuPNLiE-Fy3LJrozgsLE9ga0ibC+GOSHO6ciJppU_bnZtbx0g@mail.gmail.com>
 <73a5c9f6-5014-5113-1ddd-2fcf974273bb@siemens.com>
 <CAOuPNLjuB53ovnEBijNSqrgsKYcSGjLM-L=J4mfbRogwo_5kVA@mail.gmail.com>
 <CAOuPNLh2C8b1KC5D6+Lws_1UA0o0Sp_BaPb3mUTyebUPq_EduA@mail.gmail.com>
 <CAOuPNLiCfmUrDMWyX=N37hVJTBDEqwOtoOr1srf-55ks1qms+w@mail.gmail.com>
 <513744ad-f8fe-d54e-0618-fad492ed5bff@siemens.com>
 <5f46f00c-f09b-7439-93cf-4ff2a0343974@siemens.com>
Message-ID: <6da8a3e2-7a76-afb2-c46c-7c173a8e444c@siemens.com>
Date: Thu, 21 Jun 2018 16:57:32 +0200
MIME-Version: 1.0
In-Reply-To: <5f46f00c-f09b-7439-93cf-4ff2a0343974@siemens.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Subject: Re: [Xenomai] Kernel oops during rtnet loopback usage on x86_64
	(e1000e)
List-Id: Discussions about the Xenomai project <xenomai.xenomai.org>
List-Unsubscribe: <https://xenomai.org/mailman/options/xenomai>,
 <mailto:xenomai-request@xenomai.org?subject=unsubscribe>
List-Archive: <http://xenomai.org/pipermail/xenomai/>
List-Post: <mailto:xenomai@xenomai.org>
List-Help: <mailto:xenomai-request@xenomai.org?subject=help>
List-Subscribe: <https://xenomai.org/mailman/listinfo/xenomai>,
 <mailto:xenomai-request@xenomai.org?subject=subscribe>
To: Pintu Kumar <pintu.ping@gmail.com>, Greg Gallagher <greg@embeddedgreg.com>, Philippe Gerum <rpm@xenomai.org>
Cc: "Xenomai@xenomai.org" <xenomai@xenomai.org>

On 2018-06-21 15:41, Jan Kiszka wrote:
> On 2018-06-21 13:55, Jan Kiszka wrote:
>> On 2018-06-21 13:20, Pintu Kumar wrote:
>>> Dear Jan, Greg,
>>>
>>> Is there any pointer about this issue?
>>> This is blocking my next work..
>>
>> Does this solve the issue AND still generate valid UDP checksums (please
>> check with wireshark or against a normal networking stack?
>>
>> diff --git a/kernel/drivers/net/stack/ipv4/udp/udp.c b/kernel/drivers/net/stack/ipv4/udp/udp.c
>> index 8e80d3e0b..bb0b0fc12 100644
>> --- a/kernel/drivers/net/stack/ipv4/udp/udp.c
>> +++ b/kernel/drivers/net/stack/ipv4/udp/udp.c
>> @@ -556,18 +556,17 @@ static int rt_udp_getfrag(const void *p, unsigned char *to,
>>      if (offset)
>>  	    return rtnet_read_from_iov(ufh->fd, ufh->iov, ufh->iovlen, to, fraglen);
>>  
>> -    /* Checksum of the complete data part of the UDP message: */
>> -    for (i = 0; i < ufh->iovlen; i++) {
>> -            ufh->wcheck = csum_partial(ufh->iov[i].iov_base, ufh->iov[i].iov_len,
>> -                                       ufh->wcheck);
>> -    }
>> -
>>      ret = rtnet_read_from_iov(ufh->fd, ufh->iov, ufh->iovlen,
>>  			      to + sizeof(struct udphdr),
>>  			      fraglen - sizeof(struct udphdr));
>>      if (ret)
>>  	    return ret;
>>  
>> +    /* Checksum of the complete data part of the UDP message: */
>> +    ufh->wcheck = csum_partial(to + sizeof(struct udphdr),
>> +			       fraglen - sizeof(struct udphdr),
>> +			       ufh->wcheck);
>> +
>>      /* Checksum of the udp header: */
>>      ufh->wcheck = csum_partial((unsigned char *)ufh,
>>  			       sizeof(struct udphdr), ufh->wcheck);
>>
> 
> This was definitely wrong. Here is another try:
> 
> diff --git a/kernel/drivers/net/stack/ipv4/udp/udp.c b/kernel/drivers/net/stack/ipv4/udp/udp.c
> index 8e80d3e0b..6cf1d369e 100644
> --- a/kernel/drivers/net/stack/ipv4/udp/udp.c
> +++ b/kernel/drivers/net/stack/ipv4/udp/udp.c
> @@ -549,6 +549,7 @@ static int rt_udp_getfrag(const void *p, unsigned char *to,
>                            unsigned int offset, unsigned int fraglen)
>  {
>      struct udpfakehdr *ufh = (struct udpfakehdr *)p;
> +    unsigned int datalen = 0;
>      int i, ret;
>  
>  
> @@ -556,18 +557,18 @@ static int rt_udp_getfrag(const void *p, unsigned char *to,
>      if (offset)
>  	    return rtnet_read_from_iov(ufh->fd, ufh->iov, ufh->iovlen, to, fraglen);
>  
> -    /* Checksum of the complete data part of the UDP message: */
> -    for (i = 0; i < ufh->iovlen; i++) {
> -            ufh->wcheck = csum_partial(ufh->iov[i].iov_base, ufh->iov[i].iov_len,
> -                                       ufh->wcheck);
> -    }
> -
>      ret = rtnet_read_from_iov(ufh->fd, ufh->iov, ufh->iovlen,
>  			      to + sizeof(struct udphdr),
>  			      fraglen - sizeof(struct udphdr));
>      if (ret)
>  	    return ret;
>  
> +    /* Checksum of the complete data part of the UDP message: */
> +    for (i = 0; i < ufh->iovlen; i++)
> +	    datalen += ufh->iov[i].iov_len;
> +    ufh->wcheck = csum_partial(to + sizeof(struct udphdr), datalen,
> +			       ufh->wcheck);
> +
>      /* Checksum of the udp header: */
>      ufh->wcheck = csum_partial((unsigned char *)ufh,
>  			       sizeof(struct udphdr), ufh->wcheck);
> 

OK, this won't work either if fraglen < datalen. We have to rework this
more fundamentally.

As a workaround, run the kernel with nosmap.

Jan