Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [syzbot] net-next boot error: WARNING: refcount bug in
 fib_create_info
To:     Jakub Kicinski <kuba@kernel.org>,
        Pavel Skripkin <paskripkin@gmail.com>
Cc:     syzbot <syzbot+c5ac86461673ef58847c@syzkaller.appspotmail.com>,
        davem@davemloft.net, dsahern@kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        syzkaller-bugs@googlegroups.com, yoshfuji@linux-ipv6.org
References: <0000000000005e090405c8a9e1c3@google.com>
 <02372175-c3a1-3f8e-28fe-66d812f4c612@gmail.com>
 <e6eab0c9-7b2e-179b-b9c0-459dd9a75ed1@gmail.com>
 <20210803140435.19e560fe@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
From:   David Ahern <dsahern@gmail.com>
Message-ID: <80ea9025-05a7-ab21-ed53-5527356c4464@gmail.com>
Date:   Tue, 3 Aug 2021 16:32:02 -0600
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:78.0)
 Gecko/20100101 Thunderbird/78.12.0
MIME-Version: 1.0
In-Reply-To: <20210803140435.19e560fe@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 8/3/21 3:04 PM, Jakub Kicinski wrote:
>>>
>>> diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
>>> index f29feb7772da..bb9949f6bb70 100644
>>> --- a/net/ipv4/fib_semantics.c
>>> +++ b/net/ipv4/fib_semantics.c
>>> @@ -1428,6 +1428,7 @@ struct fib_info *fib_create_info(struct fib_config
>>> *cfg,
>>>    	}
>>>
>>>    	fib_info_cnt++;
>>> +	refcount_set(&fi->fib_treeref, 1);
>>>    	fi->fib_net = net;
>>>    	fi->fib_protocol = cfg->fc_protocol;
>>>    	fi->fib_scope = cfg->fc_scope;
>>
>> Oops, it's already fixed in -next, so
>>
>> #syz fix: ipv4: Fix refcount warning for new fib_info
>>
>>
>> BTW: there is one more bug with refcounts:
>>
>> link_it:
>> 	ofi = fib_find_info(fi);
>> 	if (ofi) {
>> 		fi->fib_dead = 1;
>> 		free_fib_info(fi);
>> 		refcount_inc(&ofi->fib_treeref);
>>
>> 		^^^^^^^^^^^^^^^^^^^^^^^
>> 		/ *fib_treeref is 0 here */
> 
> Why 0? ofi is an existing object it's already initialized.

yes, it is an existing object with a non-0 refcount.

> 
>> 		return ofi;
>> 	}
>>
>> 	refcount_set(&fi->fib_treeref, 1);
>>
>>
>> diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
>> index f29feb7772da..38d1fc4d0be1 100644
>> --- a/net/ipv4/fib_semantics.c
>> +++ b/net/ipv4/fib_semantics.c
>> @@ -1543,6 +1543,8 @@ struct fib_info *fib_create_info(struct fib_config 
>> *cfg,
>>   	}
>>
>>   link_it:
>> +	refcount_set(&fi->fib_treeref, 1);

moving the refcount_set here causes all kinds of problems with the
release and error paths in this function.