From mboxrd@z Thu Jan  1 00:00:00 1970
From: Cyril Hrubis <chrubis@suse.cz>
Date: Tue, 31 Jan 2017 16:23:34 +0100
Subject: [LTP] [PATCH 1/2] test_net.sh: add tst_set_sysctl()
In-Reply-To: <22b74bd5-fb9d-83b5-d8c1-2934a23388b0@oracle.com>
References: <1485354100-24769-1-git-send-email-alexey.kodanev@oracle.com>
 <20170131100744.GD29227@rei.lan>
 <4017cf9c-3957-0a91-4fa3-8bafdad2f4c4@oracle.com>
 <20170131124128.GF29227@rei.lan>
 <22b74bd5-fb9d-83b5-d8c1-2934a23388b0@oracle.com>
Message-ID: <20170131152333.GA27085@rei.lan>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it

Hi!
> >>>> +# tst_set_sysctl NAME VALUE [safe]
> >>>> +# It can handle netns case when sysctl not namespaceified.
> >>>> +tst_set_sysctl()
> >>>> +{
> >>>> +	local name="$1"
> >>>> +	local value="$2"
> >>>> +	local safe=
> >>>> +	[ "$3" = "safe" ] && safe="-s" || safe=""
> >>> Uh, this syntax is confusing, besides isn't safe set to empty string
> >>> anyway?
> >> What if we set TST_NO_CLEANUP just before calling cleanup()
> >> or unset TST_CLEANUP... then we can use ROD, etc. in cleanup?
> > You lost me here. How is this connected to the code I've commented?
> 
> In the 2nd patch, 'safe' option is used like this:
> 
> do_cleanup()
> {
>     tst_set_sysctl net.ipv4.tcp_cong...
> }
> 
> do_test()
> {
>     tst_set_sysctl net.ipv4.tcp_cong... safe
> }
> 
> So, the main reason for splitting the code with 'safe' option is to be
> able to call this function inside cleanup. But if we handle cleanup
> recursion inside the shell library, we can get rid of 'safe' option
> and simplify tst_set_sysctls().

Ah, I wasn't clear enough. I only meant that setting safe to empty
string, i.e. || safe="" does not seem to have any effect at all, since
it's empty anyway, or isn't it?

Also we may be able to handle recursion in the shell library, but the
main reason I avoided doing that is to keep the C library and shell
library consistent. And the C library cannot be fixed until we get rid
of last testcase using the old library (tst_res.c). Or at least I did
not figured out how to do that. But maybe converting all oldlib
testcases that use pthreads and getting rid of locking in tst_res.c
would suffice...

-- 
Cyril Hrubis
chrubis@suse.cz