From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Date: Wed, 11 Apr 2007 09:22:00 +0000 Subject: Re: [PATCH 4/5]: Rate-limit DCCP-Syncs Message-Id: <461CA8B8.8040501@trash.net> List-Id: References: <200704091058.57313@strip-the-willow> In-Reply-To: <200704091058.57313@strip-the-willow> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: dccp@vger.kernel.org Gerrit Renker wrote: > Quoting Ian McDonald: > | No, no, no. A userspace parameter in jiffies is just wrong I think. > | You change HZ and this doesn't automatically change. You could be > | doing this with different kernels on your machine even and setting > | sysctls in a file. Take a bit of time and put this in milliseconds. > The default value is set at compilation time when HZ is known. I used jiffies for the > main reason to make this sysctl consistent with the other, similar, runtime variables > which are also in units of jiffies (cf. Documentation/networking/ip-sysctl.txt): > > * inet_peer_minttl / inet_peer_maxttl / inet_peer_gc_mintime / inet_peer_gc_maxtime Actually these are in seconds, they use proc_dointvec_jiffies, which takes seconds and converts to jiffies. > * icmp_ratelimit - which is semantically the closest to this sysctl This unfortunately really is in jiffies. > Would you still rather have this in milliseconds? I agree with Ian that userspace interfaces should use fixed units that don't depend on compile-time options.