Re: [9front] [PATCH] Permissions for child boards in /srv

[ori@eigenstate.org]

Quoth Jacob Moody <moody@mail.posixcafe.org>:
> 
> Here's the patch in total now against the now reverted version
> in tree, happy to provide more context. Since this is more or less
> a full rewrite, I've included the full devsrv.c as an attachment if
> that is easier to dig through.
> 

Applied this on shithub, let's see if the front falls off.

I'll also be putting together some code to make more use
of it. Maybe we also want auth/box to grow a way of
running with a private /srv? Maybe even unconditionally?

Something like:

diff d457233c70ef6aa28861dd2978e92968ffba0920 uncommitted
--- a/sys/src/cmd/auth/box.c
+++ b/sys/src/cmd/auth/box.c
@@ -145,11 +145,11 @@
 {
 	char *b;
 	Dir *d;
-	char devs[1024];
-	int dfd;
+	char devs[1024], srvname[24], srvpath[32];
+	int dfd, sfd;
 	char *parts[256];
 	int mflags[256];
-	int nparts;
+	int n, nparts;
 
 	nparts = 0;
 	memset(devs, 0, sizeof devs);
@@ -192,6 +192,14 @@
 
 	rfork(RFNAMEG|RFFDG);
 	skelfs();
+	sfd = open("/srv/clone", OREAD);
+	if(sfd < 0)
+		sysfatal("could not open /srv/clone: %r");
+	if((n = read(sfd, srvname, sizeof(srvname))) == -1)
+		sysfatal("could not read srv id: %r");
+	snprint(srvpath, sizeof(srvpath), "/srv/%.*s", n, srvname);
+	if(bind(srvpath, "/srv", MREPL) < 0)
+		sysfatal("could not bind %s: %r", srvpath);
 	dfd = open("/dev/drivers", OWRITE|OCEXEC);
 	if(dfd < 0)
 		sysfatal("could not /dev/drivers: %r");