linux-kernel - Re: [PATCH net V2 1/2] net: stmmac: Premature loop termination check was ignored on rx

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: <87sfe2gwd2.fsf@henneberg-systemdesign.com>
Date:   Sat, 18 Mar 2023 09:38:12 +0100
From:   Jochen Henneberg <jh@...neberg-systemdesign.com>
To:     Jakub Kicinski <kuba@...nel.org>
Cc:     netdev@...r.kernel.org,
        Giuseppe Cavallaro <peppe.cavallaro@...com>,
        Alexandre Torgue <alexandre.torgue@...s.st.com>,
        Jose Abreu <joabreu@...opsys.com>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Paolo Abeni <pabeni@...hat.com>,
        Maxime Coquelin <mcoquelin.stm32@...il.com>,
        Ong Boon Leong <boon.leong.ong@...el.com>,
        linux-stm32@...md-mailman.stormreply.com,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH net V2 1/2] net: stmmac: Premature loop termination
 check was ignored on rx


Jakub Kicinski <kuba@...nel.org> writes:

> On Thu, 16 Mar 2023 08:59:39 +0100 Jochen Henneberg wrote:
>> The premature loop termination check makes sense only in case of the
>> jump to read_again where the count may have been updated. But
>> read_again did not include the check.
>> 
>> Fixes: ec222003bd94 ("net: stmmac: Prepare to add Split Header support")
>> Signed-off-by: Jochen Henneberg <jh@...neberg-systemdesign.com>
>> ---
>>  drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>> index e4902a7bb61e..ea51c7c93101 100644
>> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>> @@ -5221,10 +5221,10 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit, u32 queue)
>>  			len = 0;
>>  		}
>>  
>> +read_again:
>>  		if (count >= limit)
>>  			break;
>
> Are you sure? Can you provide more detailed analysis?
> Do you observe a problem / error in real life or is this theoretical?

This is theoretical, I was hunting another bug and just stumbled over
the check which is, I think you agree, pointless right now. I did not
try to force execute that code path.

>
> As far as I can tell only path which jumps to read_again after doing
> count++ is via the drain_data jump, but I can't tell how it's
> discarding subsequent segments in that case..
>
>> -read_again:
>>  		buf1_len = 0;
>>  		buf2_len = 0;
>>  		entry = next_entry;

Correct. The read_again is triggered in case that the segment is not the
last segment of the frame:

		if (likely(status & rx_not_ls))
			goto read_again;

So in case there is no skb (queue error) it will keep increasing count
until the last segment has been found with released device DMA
ownership. So skb will not change while the goto loop is running, the
only thing that will change is that subsequent segments release device
DMA ownership. The dirty buffers are then cleaned up from
stmmac_rx_refill().

I think the driver code is really hard to read I have planned to cleanup
things later, however, this patch simply tries to prevent us from
returning a value greater than limit which could happen and would
definitely be wrong.