[ofa-general] Re: [IBSIM] add ReLink command
Al Chu
chu11 at llnl.gov
Tue Sep 2 13:12:27 PDT 2008
Hey Sasha,
> So if one asked for ReLinking whole node? I think it should be
> straightforward - restore links for all ports where previous_remote*
> exists. What do you think?
I didn't think of that before, but I think it's a good idea. So I
tweaked it to handle this case when a port isn't specified.
> Maybe "restore previously disconnected link(s)" help message? Actually it
> is almost same :)
Now that you mention it, I think "restore" is a better word to use than
"reconnect". So I've now tweaked it to "restore previously
unconnected".
The new patch is attached.
Thanks,
Al
On Sun, 2008-08-31 at 16:45 +0300, Sasha Khapyorsky wrote:
> Hi Al,
>
> On 16:01 Thu 28 Aug , Al Chu wrote:
> > Hey Sasha,
> >
> > This adds a "ReLink" command to ibsim. If a link was previously
> > unlinked, you can run "ReLink" to reconnect it to whatever it was
> > connected to before. It's easier than having to figure out what it was
> > connected to previously and input both the local and remote ends under
> > the "Link" command.
> >
> > The idea for this option came up when I was trying to simulate an entire
> > cluster going down then going back up. Scripting the cluster to go down
> > was easy ("Unlink" all CAs), but scripting it to come back up was a
> > little harder since I had to figure out all the other end ports to input
> > into "Link".
> >
> > Al
> >
> > --
> > Albert Chu
> > chu11 at llnl.gov
> > 925-422-5311
> > Computer Scientist
> > High Performance Systems Division
> > Lawrence Livermore National Laboratory
>
> > From ec9cf72ac3dc5950337aa577f49ada6b8887d579 Mon Sep 17 00:00:00 2001
> > From: Albert Chu <chu11 at llnl.gov>
> > Date: Thu, 28 Aug 2008 15:25:14 -0700
> > Subject: [PATCH] add relink command
> >
> >
> > Signed-off-by: Albert Chu <chu11 at llnl.gov>
> > ---
> > ibsim/sim.h | 2 +
> > ibsim/sim_cmd.c | 68 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
> > 2 files changed, 70 insertions(+), 0 deletions(-)
> >
> > diff --git a/ibsim/sim.h b/ibsim/sim.h
> > index f989252..32a4e20 100644
> > --- a/ibsim/sim.h
> > +++ b/ibsim/sim.h
> > @@ -206,6 +206,8 @@ struct Port {
> > char alias[ALIASLEN + 1];
> > Node *remotenode;
> > int remoteport;
> > + Node *previous_remotenode;
> > + int previous_remoteport;
> > int errrate;
> > uint16_t errattr;
> > Node *node;
> > diff --git a/ibsim/sim_cmd.c b/ibsim/sim_cmd.c
> > index d55fb4c..39eb316 100644
> > --- a/ibsim/sim_cmd.c
> > +++ b/ibsim/sim_cmd.c
> > @@ -149,14 +149,79 @@ static int do_link(FILE * f, char *line)
> > if (link_ports(lport, rport) < 0)
> > return -fprintf(f,
> > "# can't link: local/remote port are already connected\n");
> > +
> > + lport->previous_remotenode = NULL;
> > + rport->previous_remotenode = NULL;
> > +
> > + return 0;
> > +}
> > +
> > +static int do_relink(FILE * f, char *line)
> > +{
> > + Port *lport, *rport;
> > + Node *lnode;
> > + char *orig = 0;
> > + char *lnodeid = 0;
> > + char *s = line, name[NAMELEN], *sp;
> > + int lportnum = -1;
> > +
> > + // parse local
> > + if (strsep(&s, "\""))
> > + orig = strsep(&s, "\"");
> > +
> > + lnodeid = expand_name(orig, name, &sp);
> > + if (!sp && s && *s == '[')
> > + sp = s + 1;
> > +
> > + DEBUG("lnodeid %s port [%s", lnodeid, sp);
> > + if (!(lnode = find_node(lnodeid))) {
> > + fprintf(f, "# nodeid \"%s\" (%s) not found\n", orig, lnodeid);
> > + return -1;
> > + }
> > +
> > + if (sp) {
> > + lportnum = strtoul(sp, &sp, 0);
> > + if (lportnum < 1 || lportnum > lnode->numports) {
> > + fprintf(f, "# nodeid \"%s\": bad port %d\n",
> > + lnodeid, lportnum);
> > + return -1;
> > + }
> > + } else {
> > + fprintf(f, "# no local port\n");
> > + return -1;
>
> So if one asked for ReLinking whole node? I think it should be
> straightforward - restore links for all ports where previous_remote*
> exists. What do you think?
>
> > + }
> > +
> > + lport = node_get_port(lnode, lportnum);
> > +
> > + if (!lport->previous_remotenode) {
> > + fprintf(f, "# no previous link stored\n");
> > + return -1;
> > + }
> > +
> > + rport = node_get_port(lport->previous_remotenode, lport->previous_remoteport);
> > +
> > + if (link_ports(lport, rport) < 0)
> > + return -fprintf(f,
> > + "# can't link: local/remote port are already connected\n");
> > +
> > + lport->previous_remotenode = NULL;
> > + rport->previous_remotenode = NULL;
> > +
> > return 0;
> > }
> >
> > +
>
> No need extra lines between functions.
>
> > static void unlink_port(Node * lnode, Port * lport, Node * rnode, int rportnum)
> > {
> > Port *rport = node_get_port(rnode, rportnum);
> > Port *endport;
> >
> > + /* save current connection for potential relink later */
> > + lport->previous_remotenode = lport->remotenode;
> > + lport->previous_remoteport = lport->remoteport;
> > + rport->previous_remotenode = rport->remotenode;
> > + rport->previous_remoteport = rport->remoteport;
> > +
> > lport->remotenode = rport->remotenode = 0;
> > lport->remoteport = rport->remoteport = 0;
> > lport->remotenodeid[0] = rport->remotenodeid[0] = 0;
> > @@ -713,6 +778,7 @@ static int dump_help(FILE * f)
> > fprintf(f, "\tDump [nodeid] (def all network)\n");
> > fprintf(f, "\tRoute <from-lid> <to-lid>\n");
> > fprintf(f, "\tLink \"nodeid\"[port] \"remoteid\"[port]\n");
> > + fprintf(f, "\tReLink \"nodeid\"[port] : reconnect previously unconnected link\n");
>
> Maybe "restore previously disconnected link(s)" help message? Actually it
> is almost same :)
>
> Sasha
>
> > fprintf(f, "\tUnlink \"nodeid\" : remove all links of the node\n");
> > fprintf(f, "\tUnlink \"nodeid\"[port]\n");
> > fprintf(f,
> > @@ -814,6 +880,8 @@ int do_cmd(char *buf, FILE *f)
> > *
> > * please specify new command support below this comment.
> > */
> > + else if (!strncasecmp(line, "ReLink", cmd_len))
> > + r = do_relink(f, line);
> > else if (*line != '\n' && *line != '\0')
> > fprintf(f, "command \'%s\' unknown - skipped\n", line);
> >
> > --
> > 1.5.4.5
> >
>
--
Albert Chu
chu11 at llnl.gov
925-422-5311
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-add-relink-command.patch
Type: text/x-patch
Size: 4285 bytes
Desc: not available
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20080902/3d53a684/attachment.bin>
More information about the general
mailing list