Re: Bug on CCID 3 of DCCP

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


Hello Gerrit:

We are developing the product on a Linux which runs in a quite small
file system, of about 1 GB, and we chose to follow the instructions
from a Linux from Scratch. This is why we chose the 2.6.35.4 version
of the Linux kernel. If it is necessary we will migrate to a newer
kernel version, although it may pose problems to the rest of the
software we use (as a matter of fact, we have already found problems
with other components of our product, but that's a completely
different story).

What worries me most of that bug is that I found it extremely hard to
reproduce at will. We noticed the bug because from time to time the
system crashed, and we investigated and found that those crashes were
linked to CCID3 of DCCP, which is a rather important protocol for our
product. But we cannot predict when the system will crash, how we can
make it to crash, or even why it crashes.

According to the messages we got in the kernel panic, we saw that line
of ccid3.c, but we still don't understand why that condition is a bug
-- so severe that it "deserves" a kernel panic. Reading the relevant
RFCs we think that the condition inside the BUG_ON statement could
trigger a divide-by-zero exception, but we are not sure enough about
this.

Since we don't fully understand that, we can't understand what
conditions may trigger the condition to become true and crash the
system. We would really appreciate any information that could help us
to understand these two topics. We have also observed that in the last
kernel version the BUG_ON statement still appears, so we are worried
that, even if we upgrade the kernel, the bug could reappear.

We haven't previous experience using DCCP, and we don't have
experience on Linux kernel development, which makes things a bit
harder for us.

I've tried to download the test tree, but I'm not sure whether I've
done it right -- we haven't ever used git before. I compared the
net/dccp/ directory of the tree and the same directory of version
3.2.2 and there are only few differences (the majority are about using
the "bool" type instead of "int" for certain boolean fields).

Thank your for your answer.

Jordi

On 4 February 2012 17:03, Gerrit Renker <gerrit@xxxxxxxxxxxxxx> wrote:
> Hi Jordi,
>
> 2.6.35.4 was announced 26 August 2010, almost 1.5 years ago. Since then
> some more CCID-3 specific changes have been committed, but there are
> still some significant changes in the 'test tree' which you can get
> from
>        git://eden-feed.erg.abdn.ac.uk/dccp_exp
>
> using subtree 'dccp' or 'ccid5' (which includes CCID-3, CCID-4, an
> experimental CCID-0 without congestion control and an experimental
> CCID-5 variant of CCID-2 using TCP cubic). The experimental variants
> aside, the test tree has seen more development and prolonged testing.
>
> Since it has happened several times in the past that bugs which showed up
> in the mainline could not be reproduced with the test tree, can you please
> first see if the bug re-appears with the test tree?
>
> There are two ways of doing this
>
>  1. git, details are on
>    http://www.linuxfoundation.org/collaborate/workgroups/networking/dccptesting#Experimental_DCCP_source_tree
>
>  2. a patched tarball or patches
>    These are provided on the same webserver as the above git:// url,
>    tarball and patches are generated at least once weekly from the
>    latest net tree by David Miller.
>
> If the problem does persist with the test tree, I will take care of working on the issue and ask for more
> input to reproduce the problem.
>
> If the problem does not persist with the test tree, I will make an effort to get the test tree patches
> submitted soon.
>
> Gerrit
>
> Quoting Jordi Salvador:
> | Hello:
> |
> | First of alsl, thank you for your quick answer. The kernel we are using
> | is 2.6.35.4 (and 2.6.35.14, since a few days ago). We haven't tried
> | Gerrit's development code, yet.
> |
> | I tried to read the code and understand it, and I also read the
> | relevant RFCs (for example, 4342 and 5348, and also the draft
> | specified in the source code), and I haven't yet been able to
> | understand why that situation arises. We have not much experience with
> | the Linux kernel source code.
> |
> | Thanks again for your answer.
> |
> | On 2 February 2012 17:38, Ian McDonald <ian.mcdonald@xxxxxxxxxxx> wrote:
> | > Hi Jordi
> | >
> | > I am not actively working on this code at present so sending to
> | > dccp@xxxxxxxxxxxxxxx.
> | >
> | > You will get asked a few questions so will ask them to you now:
> | > - what version of kernel are you using?
> | > - have you tried Gerrit's development code?
> | >
> | > The other thing with an experimental area of the Linux kernel is that
> | > if you are lucky someone will respond, otherwise you might need to
> | > learn the code and try and fix yourself. If you do not wish to do this
> | > you might wish to engage a professional firm. I went and fixed much of
> | > the code because I had problems I needed to resolve, for example.
> | >
> | > If I do get time then I may look at it, but highly unlikely I am afraid.
> | >
> | > Regards
> | >
> | > Ian
> | >
> | > On 2 February 2012 15:48, Jordi Salvador <salvador@xxxxxxxxxxxxxxxxx> wrote:
> | >>
> | >> Hello Mr. McDonald:
> | >>
> | >> We are developing an application on Linux which sends data using the
> | >> DCCP protocol with CCID 3. Sometimes, while sending data, we get a bug
> | >> message, originated in net/dccp/ccids/ccid3.c, line 240, and then the
> | >> system freezes. Looking at the kernel source code we find the
> | >> following:
> | >>
> | >> BUG_ON(hc->tx_p && !hc->tx_x_calc);
> | >>
> | >> We haven't been able to reproduce that bug (it just happens randomly),
> | >> and we don't exactly understand the meaning of the condition that
> | >> triggers the bug, nor how that condition may hold true. Could you
> | >> point out us some ideas which could help us to understand why we are
> | >> experiencing that bug and to understand its meaning?
> | >>
> | >> Thank you in advance.
> | >> --
> | >> Jordi Salvador
> | >> +34934980980
> | >> +34660842146
> | >> salvador@xxxxxxxxxxxxxxxxx
> | >>
> | >> Warning: This e-mail is privileged, confidential and contains private
> | >> information. Any reading, retention, distribution or copying of this
> | >> communication by any person other than its intended recipient is
> | >> prohibited.
> |
> |
> |
> | --
> | Jordi Salvador
> | +34934980980
> | +34660842146
> | salvador@xxxxxxxxxxxxxxxxx
> |
> | Warning: This e-mail is privileged, confidential and contains private
> | information. Any reading, retention, distribution or copying of this
> | communication by any person other than its intended recipient is
> | prohibited.
> | --
> | To unsubscribe from this list: send the line "unsubscribe dccp" in
> | the body of a message to majordomo@xxxxxxxxxxxxxxx
> | More majordomo info at  http://vger.kernel.org/majordomo-info.html
> |
>
> --



-- 
Jordi Salvador
+34934980980
+34660842146
salvador@xxxxxxxxxxxxxxxxx

Warning: This e-mail is privileged, confidential and contains private
information. Any reading, retention, distribution or copying of this
communication by any person other than its intended recipient is
prohibited.
--
To unsubscribe from this list: send the line "unsubscribe dccp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Linux Kernel]     [IETF DCCP]     [Linux Networking]     [Git]     [Security]     [Linux Assembly]     [Bugtraq]     [Photo]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Linux Resources]

Powered by Linux