Received-SPF: pass (google.com: best guess record for domain of linux-nfs-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] sunrpc: expiry_time should be seconds not timeval
To:     Roberto Bergantinos Corpas <rbergant@redhat.com>,
        bfields@fieldses.org
Cc:     chuck.lever@oracle.com, trond.myklebust@hammerspace.com,
        linux-nfs@vger.kernel.org
References: <20200124101154.22760-1-rbergant@redhat.com>
From:   Frank Sorenson <sorenson@redhat.com>
Message-ID: <438af54f-e1a8-467a-4ef1-821b67b7bb6c@redhat.com>
Date:   Fri, 24 Jan 2020 10:53:19 -0600
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.1.1
MIME-Version: 1.0
In-Reply-To: <20200124101154.22760-1-rbergant@redhat.com>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk

On 1/24/20 4:11 AM, Roberto Bergantinos Corpas wrote:
> When upcalling gssproxy, cache_head.expiry_time is set as a
> timeval, not seconds since boot. As such, RPC cache expiry
> logic will not clean expired objects created under
> auth.rpcsec.context cache.
> 
> This has proven to cause kernel memory leaks on field.
> 
> Signed-off-by: Roberto Bergantinos Corpas <rbergant@redhat.com>
> ---
>  net/sunrpc/auth_gss/svcauth_gss.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
> index 8be2f209982b..725cf5b5ae40 100644
> --- a/net/sunrpc/auth_gss/svcauth_gss.c
> +++ b/net/sunrpc/auth_gss/svcauth_gss.c
> @@ -1211,6 +1211,7 @@ static int gss_proxy_save_rsc(struct cache_detail *cd,
>  		dprintk("RPC:       No creds found!\n");
>  		goto out;
>  	} else {
> +		struct timespec boot;
>  
>  		/* steal creds */
>  		rsci.cred = ud->creds;
> @@ -1231,6 +1232,9 @@ static int gss_proxy_save_rsc(struct cache_detail *cd,
>  						&expiry, GFP_KERNEL);
>  		if (status)
>  			goto out;
> +
> +		getboottime(&boot);
> +		expiry -= boot.tv_sec;
>  	}
>  
>  	rsci.h.expiry_time = expiry;
> 

The accumulating  become apparent when the client uses kerberos tickets
with very short (10 seconds or fewer) lifetimes and renewable lifetimes:

mount server:/exports /mnt/tmp -overs=4,sec=krb5p
life="2s"
rlife="3s"
while true ; do
	while true ; do
		kinit -l $life -R >/dev/null 2>&1 && break
		echo 'PASSWORD' | kinit -l $life -r $rlife \
			>/dev/null 2>&1 && break
	done
	timeout -k 1 2 touch /mnt/tmp/foo
	echo -n .
done

Due to the entry expiration occurring 50 years in the future, the
customer had accumulated in excess of 400,000 entries in the cache over
about a month with just 6 nfs clients.  The entries, with all the
accompanying structs which had been allocated consumed over 2 GiB from
various slab caches.

A flush of the cache cleans everything out, however they will again
accumulate afterward.

This patch fixes the expiration issue.

Tested-By: Frank Sorenson <sorenson@redhat.com>


Frank
--
Frank Sorenson
sorenson@redhat.com
Principal Software Maintenance Engineer
Global Support Services - filesystems
Red Hat