linux-kernel - Re: [PATCH RFC] nfsd: fix error handling in nfsd

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-id: <169870163037.24305.14020614041859684912@noble.neil.brown.name>
Date:   Tue, 31 Oct 2023 08:33:50 +1100
From:   "NeilBrown" <neilb@...e.de>
To:     "Jeff Layton" <jlayton@...nel.org>
Cc:     "Chuck Lever" <chuck.lever@...cle.com>,
        "Olga Kornievskaia" <kolga@...app.com>,
        "Dai Ngo" <Dai.Ngo@...cle.com>, "Tom Talpey" <tom@...pey.com>,
        linux-nfs@...r.kernel.org, linux-kernel@...r.kernel.org,
        "Zhi Li" <yieli@...hat.com>, "Jeff Layton" <jlayton@...nel.org>
Subject: Re: [PATCH RFC] nfsd: fix error handling in nfsd_svc

On Tue, 31 Oct 2023, Jeff Layton wrote:
> Once we've set the nfsd_serv pointer in nfsd_svc, we still need to call
> nfsd_last_thread if the server fails to be started. Remove the special
> casing for nfsd_up_before case since shutting down the per-net stuff is
> also handled by nfsd_last_thread.
> 
> Finally, add a new special case at the start and skip doing anything if
> the service already exists, 0 threads were requested and
> serv->sv_nrthreads is 0.

This is very similar to my 
  Commit bf32075256e9 ("NFSD: simplify error paths in nfsd_svc()")

The main difference being that special case you mention.  I don't like
that bit.
If I run "rpc.nfsd 0" then I want the nfsd_svc to be destroyed, whether
there were threads running or not.

Is there a reason my patch isn't sufficient?

Thanks,
NeilBrown


> 
> Fixes: 9f28a971ee9f ("nfsd: separate nfsd_last_thread() from nfsd_put()")
> Reported-by: Zhi Li <yieli@...hat.com>
> Signed-off-by: Jeff Layton <jlayton@...nel.org>
> ---
> Here's what I was thinking for a targeted patch for stable. Testing it
> now, but I won't have results until tomorrow.
> ---
>  fs/nfsd/nfssvc.c | 15 +++++----------
>  1 file changed, 5 insertions(+), 10 deletions(-)
> 
> diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
> index 3deef000afa9..187b68769815 100644
> --- a/fs/nfsd/nfssvc.c
> +++ b/fs/nfsd/nfssvc.c
> @@ -787,7 +787,6 @@ int
>  nfsd_svc(int nrservs, struct net *net, const struct cred *cred)
>  {
>  	int	error;
> -	bool	nfsd_up_before;
>  	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
>  	struct svc_serv *serv;
>  
> @@ -797,8 +796,9 @@ nfsd_svc(int nrservs, struct net *net, const struct cred *cred)
>  	nrservs = max(nrservs, 0);
>  	nrservs = min(nrservs, NFSD_MAXSERVS);
>  	error = 0;
> +	serv = nn->nfsd_serv;
>  
> -	if (nrservs == 0 && nn->nfsd_serv == NULL)
> +	if (nrservs == 0 && (serv == NULL || serv->sv_nrthreads == 0))
>  		goto out;
>  
>  	strscpy(nn->nfsd_name, utsname()->nodename,
> @@ -808,22 +808,17 @@ nfsd_svc(int nrservs, struct net *net, const struct cred *cred)
>  	if (error)
>  		goto out;
>  
> -	nfsd_up_before = nn->nfsd_net_up;
>  	serv = nn->nfsd_serv;
>  
>  	error = nfsd_startup_net(net, cred);
>  	if (error)
>  		goto out_put;
>  	error = svc_set_num_threads(serv, NULL, nrservs);
> -	if (error)
> -		goto out_shutdown;
> -	error = serv->sv_nrthreads;
>  	if (error == 0)
> -		nfsd_last_thread(net);
> -out_shutdown:
> -	if (error < 0 && !nfsd_up_before)
> -		nfsd_shutdown_net(net);
> +		error = serv->sv_nrthreads;
>  out_put:
> +	if (serv->sv_nrthreads == 0)
> +		nfsd_last_thread(net);
>  	/* Threads now hold service active */
>  	if (xchg(&nn->keep_active, 0))
>  		svc_put(serv);
> 
> ---
> base-commit: 31b5a36c4b88b44c91cdd523997b1e86fb47339d
> change-id: 20231030-kdevops-5f7366897ef4
> 
> Best regards,
> -- 
> Jeff Layton <jlayton@...nel.org>
> 
>