Re: [PATCH] Re: sendfile02 (was:Re: [LTP] Modularized gzip'd tarballs created....)

[Robert W. <ro...@us...>]

Thanks for the patch Sundar, I patched the CVS version and it will be
included in the Feb release.

- Robbie

Robert V. Williamson <ro...@us...>
Linux Test Project
IBM Linux Technology Center
Phone: (512) 838-9295   T/L: 678-9295
Fax: (512) 838-4603
http://ltp.sourceforge.net
====================
"Only two things are infinite, the universe and human stupidity, and I'm
not sure about the former." -Albert Einstein


                                                                                                                                   
                      "V.R.Sundar"                                                                                                 
                      <vr_sundar@attbi.        To:       Robert Williamson/Austin/IBM@IBMUS                                        
                      com>                     cc:       ltp...@li...                                            
                                               Subject:  [PATCH] Re: sendfile02 (was:Re: [LTP] Modularized gzip'd tarballs         
                      01/19/2003 10:18          created....)                                                                       
                      PM                                                                                                           
                                                                                                                                   




Robbie,

The patch for sendfile02 is attached.
(1) Call create_server from within do_sendfile so that a child is forked
for each test iteration.
(2) Don't wait for child in case of a sendfile error. kill child in case
of error.
(3) Fixed small errors in waitpid and recvfrom

let me know if you have any problems.

sundar
On Wed, 2003-01-15 at 10:36, Robert Williamson wrote:
> Sundar,
>
>   I think the second way is better overall.  Although delivery of the
> packet has nothing to do with the correctness of sendfile, some users of
> this test also use it to test their networking.
>
> - Robbie
>
> Robert V. Williamson <ro...@us...>
> Linux Test Project
> IBM Linux Technology Center
> Phone: (512) 838-9295   T/L: 678-9295
> Fax: (512) 838-4603
> http://ltp.sourceforge.net
> ====================
> "Only two things are infinite, the universe and human stupidity, and I'm
> not sure about the former." -Albert Einstein
>
>
>
>                       "V.R.Sundar"
>                       <vr_...@at...>            To:
ltp...@li..., Dan Kegel <da...@ke...>
>                       Sent by:                         cc:
>                       ltp...@li...ur        Subject:
sendfile02 (was:Re: [LTP] Modularized gzip'd tarballs created....)
>                       ceforge.net
>
>
>                       01/15/2003 01:44 AM
>
>
>
>
>
> On Mon, 2003-01-13 at 23:15, Dan Kegel wrote:
> > Incidentally, the [or a] LTP sendfile
> > test creates a child process for no reason...
> > and continues to work on the 2nd, 3rd, and 4th
> > calls to its test function even though
> > the test function kills the child after
> > the 1st one.   No need to have a recipient for
> > UDP packets, is there?  (This is all hearsay...
> > Vasan Sundar noticed this, and can say more.)
>
> The parent creates only one child process (which waits for a udp packet
> and exits when it has received it) but loops 4 times, each time sending
> a packet and waiting for the child to exit . After the first time, the
> one child has already exited and further waitpids return with -1 and
> ECHILD (there is a small bug that makes it return with EFAULT right
> now). This is not the right  thing to do but doesn't affect the test
> result. The test returns success if sendfile succeeds - the child's
> return value is not checked.
>
> The main problem though is the error handling if sendfile fails the
> first time. The parent goes on to wait for the child. But the child is
> waiting for data which doesn't come as sendfile failed. So both go on
> waiting forever. This is what happens on the ppc405 - sendfile fails
> with EINVAL for some reason.
>
> I would like to fix this by changing the testcase to not fork at all and
> then remove the waitpids. The socket used in sendfile is UDP so we don't
> need a recipient. ( thats what happens now 75% of the time anyway). The
> child does not check the data received and the parent doesn't test the
> child's return code. What we lose is the indication that a packet was
> delivered which anyway is not guaranteed and has nothing to do with
> whether sendfile fails/succeeds
>
> The other alternative would be to fork the correct number of children,
> modify the error checking to kill the child on error and exit directly.
> Additionally maybe add some checking on the child and/or parent side for
> the data or  return code.
>
> sundar
> --
> V.R.Sundar <vr_...@at...>
>
>
>
> -------------------------------------------------------
> This SF.NET email is sponsored by: Take your first step towards giving
> your online business a competitive advantage. Test-drive a Thawte SSL
> certificate - our easy online guide will show you how. Click here to get
> started: http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0027en
> _______________________________________________
> Ltp-list mailing list
> Ltp...@li...
> https://lists.sourceforge.net/lists/listinfo/ltp-list
>
>
>
>
>
>
> -------------------------------------------------------
> This SF.NET email is sponsored by: A Thawte Code Signing Certificate
> is essential in establishing user confidence by providing assurance of
> authenticity and code integrity. Download our Free Code Signing guide:
> http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0028en
> _______________________________________________
> Ltp-list mailing list
> Ltp...@li...
> https://lists.sourceforge.net/lists/listinfo/ltp-list
--
V.R.Sundar <vr_...@at...>



#### sendfile.patch has been removed from this note on January 20, 2003 by
Robert Williamson