Re: strtst fails with LiS-2.18 and MontaVista CGE 3.1 (band_test and fifo_test)

Longhi,

I fixed a bunch of these kinds of configuration problems in LiS 2.18.1.
You might want to try the new release and see if your problems disappear.
Otherwise, I will help you work forward on 2.18.1.

--brian

On Wed, 06 Apr 2005, Longhi Stefano wrote:

> 
>    Hello,
> 
>    I'm running LiS-2.18 compiled against MontaVista CGE 3.1 and have some
>    problems.
> 
> 
> 
>    1) To make LiS compile with MVCGE, I had to include <linux/namei.h> in
>    linux-mdep.c     (undefined    struct    nameidata)    and    redefine
>    EXPORT_SYMBOL_NOVERS  in  exports.c  because  otherwise I  had a parse
>    error before
>    "this_object_must_be_defined_as_export_objs_in_the_Makefile".
>    I  suppose something  is  not  correctly  defined in my configuration,
>    but I  cannot  fint  what.  It  seems  that  this  error  appears when
>    EXPORT_SYMTAB  is not defined, even though EXPORT_SYMTAB is defined at
>    the  beginning  of  exports.c.  Anyway,  with  my corrections, streams
>    module  loads  and exported symbols are shown with ksyms, but I wonder
>    if my changes are correct.
> 
>    2)  strtst  sometimes  fails  during the Queue band test (about 50% of
>    strtst runs).
>    To test different configurations, since MontaVista CGE 3.1 is based on
>    2.4.20  kernel,  I  compiled  LiS  and  run  strtst with 2.4.20 kernel
>    downloaded   from  kernel.org  and  everything  went  fine,  but  with
>    2.4.20-20.9  kernel  provided with RedHat 9.B I have the same problems
>    and  Queue  band test fails (note that I used the same .config file to
>    compile  both  kernels),  therefore  it seems it is not only a problem
>    related to MVCGE environment.
>    Searching in the mail archive I found similar problem descriptions and
>    I tried to verify whether the failure is related to a delay problem.
>    So,  I  commented  out the set_debug_mask(0x00000000L) instructions in
>    the  band_test  function  (the  receive_band(2): seq=31024 Bad message
>    always  happens  soon  after  that  instruction)  and  I  had  no more
>    failures.
>    My  question  is: is the band_test failure only due to delays that are
>    introduced  to  log debug  information? is it related only with strtst
>    execution or could I have problems whith my modules that use streams?
>    In  spite  of my change  to  band_test I  had the problem described at
>    point 3). (I run a script that executes a loop of strtst).
> 
>    3) running strtst I have sporadic Oops during FIFO/pipe test:
>    after  strtst  starts,  fifo_test: open("/dev/fifo.0",O_RDWR) #1: fd 4
>    mode 020666 dev 0xe85f
>    I have:
>      lis_qopen([1]q@0xc491b584,...)    calling    module    "fifo"   open
>    @0xd0861360 (dev=e80005f,flag=2,sflag=1)
>      Unable to handle kernel paging request at virtual address 0000ee20
> 
>    After that, strtst hangs and I have to reboot to use streams again.
> 
>    In attachment there are traces related to problems 2) and 3).
>    Anyone who can help me is welcome!
> 
>    Many thanks,
> 
>        Stefano
> 
> References
> 
>    1. mailto:q@0xc491b584


-- 
Brian F. G. Bidulock    ¦ The reasonable man adapts himself to the ¦
[email protected]    ¦ world; the unreasonable one persists in  ¦
http://www.openss7.org/ ¦ trying  to adapt the  world  to himself. ¦
                        ¦ Therefore  all  progress  depends on the ¦
                        ¦ unreasonable man. -- George Bernard Shaw ¦