> Date: Tue, 17 Mar 2020 14:12:14 +0100
> From: Klemens Nanni <[email protected]>
> Content-Type: text/plain; charset=us-ascii
> Content-Disposition: inline
>
> On Sat, Mar 07, 2020 at 05:44:12PM +0100, Klemens Nanni wrote:
> > Next to the IO device path to be used in ldom.conf we can also print the
> > name which actually tells in human readable form what device we're
> > looking at.
> >
> > Those names are printed in similar format by Solaris `ldm list-io' and
> > directly match the the structure seen in the iLOM shell.
> >
> > $ doas ldomctl list-io
> > PATH NAME
> > /@400/@2/@0/@8 /SYS/MB/PCIE0
> > /@500/@2/@0/@a /SYS/MB/PCIE1
> > /@400/@2/@0/@4 /SYS/MB/PCIE2
> > /@500/@2/@0/@6 /SYS/MB/PCIE3
> > /@400/@2/@0/@0 /SYS/MB/PCIE4
> > /@500/@2/@0/@0 /SYS/MB/PCIE5
> > /@400/@1/@0/@8 /SYS/MB/PCIE6
> > /@500/@1/@0/@6 /SYS/MB/PCIE7
> > /@400/@1/@0/@c /SYS/MB/PCIE8
> > /@500/@1/@0/@0 /SYS/MB/PCIE9
> > /@400/@2/@0/@e /SYS/MB/SASHBA
> > /@400/@1/@0/@4 /SYS/MB/NET0
> > /@500/@1/@0/@5 /SYS/MB/NET2
> >
> > OK?
> Anyone? I'll probably just commit in a few days unless someone objects,
> printing device names is a nice aid in partitioning machines, especially
> since picking the wrong iodevice needs power cycling and might even
> prevent the primary domain from booting since another guest domain
> accidentially got its root disk device.
>
> PS: with the new mdprint port on ports@ one can easily inspect the PRI
> and verify:
>
> $ mdprint -pqng component.assignable-path,component.nac ./pri |
> > grep assignable
> component|assignable-path="/@400/@2/@0/@8"|nac="/SYS/MB/PCIE0"
> component|assignable-path="/@500/@2/@0/@a"|nac="/SYS/MB/PCIE1"
> component|assignable-path="/@400/@2/@0/@4"|nac="/SYS/MB/PCIE2"
> component|assignable-path="/@500/@2/@0/@6"|nac="/SYS/MB/PCIE3"
> component|assignable-path="/@400/@2/@0/@0"|nac="/SYS/MB/PCIE4"
> component|assignable-path="/@500/@2/@0/@0"|nac="/SYS/MB/PCIE5"
> component|assignable-path="/@400/@1/@0/@8"|nac="/SYS/MB/PCIE6"
> component|assignable-path="/@500/@1/@0/@6"|nac="/SYS/MB/PCIE7"
> component|assignable-path="/@400/@1/@0/@c"|nac="/SYS/MB/PCIE8"
> component|assignable-path="/@500/@1/@0/@0"|nac="/SYS/MB/PCIE9"
> component|assignable-path="/@400/@2/@0/@e"|nac="/SYS/MB/SASHBA"
> component|assignable-path="/@400/@1/@0/@4"|nac="/SYS/MB/NET0"
> component|assignable-path="/@500/@1/@0/@5"|nac="/SYS/MB/NET2"
>
> That's how ldomctl works around iodevices: They're "component" nodes in
> the tree that have an "assignable-path" attribute; ldomctl picks those,
> assigns whatever is specified in ldom.conf to guest domains, and assigns
> all remaining components to the primary domain.
Looks like I missed this. I like this. However:
> Index: config.c
> ===================================================================
> RCS file: /cvs/src/usr.sbin/ldomctl/config.c,v
> retrieving revision 1.34
> diff -u -p -r1.34 config.c
> --- config.c 21 Feb 2020 19:39:28 -0000 1.34
> +++ config.c 7 Mar 2020 16:39:25 -0000
> @@ -48,6 +48,7 @@ TAILQ_HEAD(, core) cores;
>
> struct component {
> const char *path;
> + const char *nac;
> int assigned;
>
> struct md_node *hv_node;
> @@ -216,6 +217,7 @@ pri_init_components(struct md *md)
> struct component *component;
> struct md_node *node;
> const char *path;
> + const char *nac;
> const char *type;
>
> TAILQ_INIT(&components);
> @@ -228,6 +230,10 @@ pri_init_components(struct md *md)
> if (md_get_prop_str(md, node, "assignable-path", &path)) {
> component = xzalloc(sizeof(*component));
> component->path = path;
> + if (md_get_prop_str(md, node, "nac", &nac))
> + component->nac = nac;
> + else
> + component->nac = "-";
> TAILQ_INSERT_TAIL(&components, component, link);
> }
>
> @@ -2889,7 +2896,8 @@ list_components(void)
>
> pri_init(pri);
>
> + printf("PATH\t\tNAME\n");
Using tabs to make things line up isn't going to work very well. So
better use field widths in the printf calls.
> TAILQ_FOREACH(component, &components, link) {
> - printf("%s\n", component->path);
> + printf("%s\t%s\n", component->path, component->nac);
> }
> }
>
