Re: [newnfs/client] SIGINFO aborts transfer and produces `permission denied'

From: Rick Macklem <rmacklem_at_uoguelph.ca>
Date: Sat, 18 Jul 2009 09:35:33 -0400 (EDT)
On Sat, 18 Jul 2009, Anonymous wrote:

>
> Yep, I can reproduce it as easily on 8.0-BETA2 snapshot under qemu
>
> # uname -vm
> FreeBSD 8.0-BETA2 #0: Wed Jul 15 23:25:30 UTC 2009
> root_at_almeida.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC  i386
>
I'll try to get around to testing it this weekend, but if you'd
like to test the following patch, I think it might fix the problem.

The new krpc only checks NFSMNT_INT at connect and not every rpc.

rick
ps: The line #s assume the other patch you tested is already applied.
--- untested exp. nfs client patch ---
--- fs/nfsclient/nfs_clvfsops.c.sav2	2009-07-18 09:19:07.000000000 -0400
+++ fs/nfsclient/nfs_clvfsops.c	2009-07-18 09:29:14.000000000 -0400
_at__at_ -1037,7 +1037,7 _at__at_
  {
  	struct nfsmount *nmp;
  	struct nfsnode *np;
-	int error, trycnt, ret, clearintr;
+	int error, trycnt, ret;
  	struct nfsvattr nfsva;
  	static u_int64_t clval = 0;

_at__at_ -1152,20 +1152,8 _at__at_
  		nmp->nm_sockreq.nr_vers = NFS_VER2;


-	/*
-	 * For Connection based sockets (TCP,...) do the connect here,
-	 * but make it interruptible, even for non-interuptible mounts.
-	 */
-	if ((nmp->nm_flag & NFSMNT_INT) == 0) {
-		nmp->nm_flag |= NFSMNT_INT;
-		clearintr = 1;
-	} else {
-		clearintr = 0;
-	}
  	if ((error = newnfs_connect(nmp, &nmp->nm_sockreq, cred, td, 0)))
  		goto bad;
-	if (clearintr)
-		nmp->nm_flag &= ~NFSMNT_INT;

  	/*
  	 * A reference count is needed on the nfsnode representing the
Received on Sat Jul 18 2009 - 11:32:33 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:52 UTC