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;
Content-Type: text/plain;
        charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 12.4 \(3445.104.11\))
Subject: Re: [PATCH v3] SUNRPC/cache: Fix unsafe traverse caused double-free
 in cache_purge
From:   Chuck Lever <chuck.lever@oracle.com>
In-Reply-To: <e0dd0339-a15e-814d-ac5a-5f51bc15d73c@linux.alibaba.com>
Date:   Sun, 5 Apr 2020 14:07:05 -0400
Cc:     Bruce Fields <bfields@fieldses.org>, Neil Brown <neilb@suse.de>,
        Sasha Levin <sashal@kernel.org>,
        Linux NFS Mailing List <linux-nfs@vger.kernel.org>
Content-Transfer-Encoding: 7bit
Message-Id: <0308EB8A-A8B7-412C-8F93-A444DE47CB1D@oracle.com>
References: <4568a7cf87f110b8e59fda6f53fda34c550ab403.1586108200.git.wuyihao@linux.alibaba.com>
 <e0dd0339-a15e-814d-ac5a-5f51bc15d73c@linux.alibaba.com>
To:     Yihao Wu <wuyihao@linux.alibaba.com>
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk

Hi Yihao-

> On Apr 5, 2020, at 1:57 PM, Yihao Wu <wuyihao@linux.alibaba.com> wrote:
> 
> Deleting list entry within hlist_for_each_entry_safe is not safe unless
> next pointer (tmp) is protected too. It's not, because once hash_lock
> is released, cache_clean may delete the entry that tmp points to. Then
> cache_purge can walk to a deleted entry and tries to double free it.
> 
> Fix this bug by holding only the deleted entry's reference.
> 
> Signed-off-by: Yihao Wu <wuyihao@linux.alibaba.com>
> ---
> v1->v2: Use Neil's better solution
> v2->v3: Fix a checkscript warning
> 
> net/sunrpc/cache.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c
> index af0ddd28b081..b445874e8e2f 100644
> --- a/net/sunrpc/cache.c
> +++ b/net/sunrpc/cache.c
> @@ -541,7 +541,9 @@ void cache_purge(struct cache_detail *detail)
> 	dprintk("RPC: %d entries in %s cache\n", detail->entries, detail->name);
> 	for (i = 0; i < detail->hash_size; i++) {
> 		head = &detail->hash_table[i];
> -		hlist_for_each_entry_safe(ch, tmp, head, cache_list) {

If review/testing shows you need to respin this patch, I note that "tmp" is
now unused and should be removed. I've pulled v3 into my testing branch and
made that minor change. Thanks!


> +		while (!hlist_empty(head)) {
> +			ch = hlist_entry(head->first, struct cache_head,
> +					 cache_list);
> 			sunrpc_begin_cache_remove_entry(ch, detail);
> 			spin_unlock(&detail->hash_lock);
> 			sunrpc_end_cache_remove_entry(ch, detail);
> -- 
> 2.20.1.2432.ga663e714

--
Chuck Lever