[LTP] [PATCH 1/6] ftp/ftp01: Use tst_net.sh

[Petr Vorel]

Hi Akihiko,

> Hi Petr,

> On 2022/10/21 15:22, Petr Vorel wrote:
> > Hi Akihiko,

> > I'm sorry I overlooked your message at v1
> > https://lore.kernel.org/ltp/9894db50-6319-a818-c995-3ba9ab102c4b@daynix.com/

> > I guess script expects $PASSWD to be set.
> > Also trying to follow the instruction, none of them (allowing root in
> > /etc/ftpusers or filling /root/.netrc with machine localhost) work for me:
> > https://github.com/linux-test-project/ltp/tree/master/testcases/network#ftp-and-telnet-setup

> > I suppose any ftpd implementation must be run on the remote server (localhost,
> > if tst_net.sh uses netns - the default or on remote host, if ssh is used).
> > Otherwise ftp command "user $RUSER $PASSWD" cannot work, right?

> > Although it should behave the same regardless netns or ssh is used,
> > how do you test v3, with ssh or with netns (the default)? Or both?

> Thanks for sharing your fork. I tried it with SSH, and I had to make the
> following modifications to run it correctly:
> - I had to restore commented out "cd $TST_NET_DATAROOT".
Ack, this is correct.
> - I had to change the arguments of test_get and test_put back to "binary"
> and "ascii".
>   It is used to set the transfer mode (search for "echo $1").
Ah, I'm sorry I overlooked this. This could be solved with passing it as a first
parameter and shift, but I'll stop to be picky on an implementation detail.

> > It'd be nice if there was connection detection in the setup and tst_brk TCONF if
> > login does not work. The best would be to have configuration in the setup() +
> > cleanup in the cleanup()).

> > There are tests, which do vsftpd configuration: ftp-download-stress.sh,
> > ftp-upload-stress.sh.

> > BTW in the past we seriously considered to delete these highlevel smoke tests.
> > LTP is concentrated to test kernel API and internals, thus we didn't see much
> > value with smoke tests like this (that's why they haven't been rewritten to use
> > new API), specially if they require complex setup and get false positives when
> > SUT not configured properly. That's why it'd be nice to at least TCONF (if not
> > doing whole setup).

> I just modified this test because it is annoying to set up rsh just to fix
> this test so I would rather not put more effort for further improvement.
Understand, ack. Thanks for your work!

> Personally I don't object to remove this test either.
The fastest solution is to merge your fixed version.
But there should be even more modifications:
RHOST is obsolete, instead $(tst_ipaddr rhost) should be used.

Therefore we should decide if this smoke test (and other FTP tests in LTP) is
worth effort. If yes, I should force myself to create ftp_lib.sh and migrate
everything to work similar way as ftp-upload-stress.sh.

Kind regards,
Petr

> Regards,
> Akihiko Odaki


> > Kind regards,
> > Petr