Re: [LTP] Patch to record kernel dump message to each testcases log

[Subrata Modak <subrata@linux.vnet.ibm.com>]

On Wed, 2010-10-13 at 13:30 +0800, hake huang wrote:
> 2010/10/13 Garrett Cooper <yanegomi@gmail.com>
> >
> > On Tue, Oct 12, 2010 at 9:38 PM, hake huang <hakehuang@gmail.com> wrote:
> > > Hi
> > >
> > > I see this simple solution is no quite fit. But the kernel dump message for
> > > each test case is necessary for analysing. I can work on this further, with
> > > below help:
> > > 1. Where can I find the reason for the IIRC break?
> >
> > IIRC -> If I remember correctly.
> so I need ask Subrata about this problem, right? what is the existing
> functionality?

I believe the functionality your original patch was proposing is already
in LTP . I guess they will solve your problem. See the following
commits:

git show e3bc3758d3e4d17a0e04009ef873df53819b9ec6
git show bd9d440be1da8020c93ac92e497e3a16e66e9fa3

Just use:

./runltp -K <DMESG_DIRECTORY_NAME>

option while running ltp.

Regards--
Subrata

> >
> > > 2. Can it  be an option to avoid the size limitation on some embedded
> > > system?
> >
> > No. This is outside the realm of LTP, but we must remain conscious of
> > the fact that there are folks running LTP on cellphones, etc, where
> > space is very much a luxury.
> I see. but why can't they run in quit_mode?
> >
> > > 3.  A kind of seperating tag is needed for kernel message, right? e.g
> > > <kernel_dump_start>dump message<kernel_dump_end>
> >
> > I think that adding an optional capture feature might be worthwhile,
> > but you'll no doubt quickly find with some processes that the dmesg
> > buffer will fill up quickly.
> I have seen such events, and such cases need to be checked manually.
> >
> > > 4.  I see in ltp-pan.c the file descriptors, terminals are changed in child
> > > process. if the system(3) has problem, the worst cases is there are no
> > > kernel messge output in parent process, but the pan can still function well.
> >
> > Can you say that with 100% certainty? One of the potential problems
> > with system is the fact that it allocates a shell and depending on how
> > processes are executed and how file descriptors are cleaned up,
> >
> > Actually, pan makes zero attempts at closing open file descriptors and
> > the like, so the problem is still very much with pan.
> >
> So your concerns is about file descriptors which open in ltp process,
> I think this is an existing problem. I mean maybe pan need to attempt
> to close open file descriptors. system(3) only increases the change of
> occurrence caused by this problem, right?
> > > if not use system(3), can you give your suggestion on this, exec(3) seems to
> > > have the same problem as system(3)?
> >
> > Yes and no. In this case the problems that exist with open file
> > descriptor inheritance, etc are unfortunately no worse from that
> > perspective than your proposed code would do, but there are still
> > other problems that need to be resolved with your proposed change
> > anyhow that are much more pressing.
> >
> > -Garrett


------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today.
http://p.sf.net/sfu/beautyoftheweb
_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list