linux-kernel - Re: [PATCH] Revert "net: pskb_trim_rcsum() and CHECKSUM

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: <dd6f13bc-c5f2-85f8-c08d-837bc024fc7c@gmail.com>
Date:   Mon, 18 Jun 2018 16:36:15 -0700
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Eric Dumazet <eric.dumazet@...il.com>,
        Mathieu Malaterre <malat@...ian.org>
Cc:     schwab@...ux-m68k.org, "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Christophe LEROY <christophe.leroy@....fr>,
        Meelis Roos <mroos@...ux.ee>, netdev@...r.kernel.org,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>
Subject: Re: [PATCH] Revert "net: pskb_trim_rcsum() and CHECKSUM_COMPLETE are
 friends"



On 06/18/2018 04:29 PM, Eric Dumazet wrote:
> 
> 
> On 06/18/2018 11:45 AM, Mathieu Malaterre wrote:
> 
>>
>> Here is what I get on my side
>>
>> [   53.628847] sungem: sungem wrong csum : 4e04/f97, len 64 bytes
>> [   53.667063] sungem: sungem wrong csum : eea8/6eec, len 149 bytes
>> [   58.648952] sungem: sungem wrong csum : 2095/3d06, len 64 bytes
>> [   58.669414] sungem: sungem wrong csum : 5245/b50, len 149 bytes
>> [   63.674451] sungem: sungem wrong csum : 2d8/5abd, len 149 bytes
>> [   68.678233] sungem: sungem wrong csum : b8fc/a498, len 149 bytes
>> [   73.685771] sungem: sungem wrong csum : 374/5a21, len 149 bytes
>> [   78.689089] sungem: sungem wrong csum : d81/5014, len 149 bytes
>> [   83.683261] sungem: sungem wrong csum : 4e04/f97, len 64 bytes
>> [   83.690193] sungem: sungem wrong csum : c2f7/9a9d, len 149 bytes
>> [   88.692511] sungem: sungem wrong csum : f4d8/68bc, len 149 bytes
>> [   93.699915] sungem: sungem wrong csum : 1370/4a25, len 149 bytes
>> [   98.703136] sungem: sungem wrong csum : e1b5/7bdf, len 149 bytes
>> [  103.704230] sungem: sungem wrong csum : 5321/a74, len 149 bytes
>> [  108.688912] sungem: sungem wrong csum : 2095/3d06, len 64 bytes
>> [  108.706559] sungem: sungem wrong csum : ddbc/7fd8, len 149 bytes
>> [  113.713189] sungem: sungem wrong csum : 5a65/330, len 149 bytes
>> [  113.891697] sungem: sungem wrong csum : 4e04/f97, len 64 bytes
>> [  118.717151] sungem: sungem wrong csum : f7c8/65cc, len 149 bytes
>> [  123.722680] sungem: sungem wrong csum : 3d7a/201b, len 149 bytes
>> [  128.726524] sungem: sungem wrong csum : c8fd/9497, len 149 bytes
>> [  133.732045] sungem: sungem wrong csum : de0d/7f87, len 149 bytes
>> [  135.529163] sungem: sungem wrong csum : 3089/b6dd, len 96 bytes
>> [  135.529208] eth0: hw csum failure
>> [  135.529220] CPU: 0 PID: 0 Comm: swapper Not tainted 4.17.0+ #7
>> [  135.529226] Call Trace:
>> [  135.529243] [dffedbe0] [c069ddac]
>> __skb_checksum_complete+0xf0/0x108 (unreliable)
> 
> Thanks, then I guess next step would be to dump the content of the frames
> having a wrong checksum, hoping we find an easy way to discard the CHECKSUM_COMPLETE
> in a selective way.
> 
> Otherwise, we will need to remove CHECKSUM_COMPLETE setting in this driver.
> 
> diff --git a/drivers/net/ethernet/sun/sungem.c b/drivers/net/ethernet/sun/sungem.c
> index 7a16d40a72d13cf1d522e8a3a396c826fe76f9b9..77a761f95be788bb86c8d917f613c9084818f826 100644
> --- a/drivers/net/ethernet/sun/sungem.c
> +++ b/drivers/net/ethernet/sun/sungem.c
> @@ -857,6 +857,14 @@ static int gem_rx(struct gem *gp, int work_to_do)
>  
>                 csum = (__force __sum16)htons((status & RXDCTRL_TCPCSUM) ^ 0xffff);
>                 skb->csum = csum_unfold(csum);
> +               {
> +               __wsum rsum = csum_partial(skb->data + ETH_HLEN, len - ETH_HLEN, 0);
> +               if (csum != csum_fold(rsum) && net_ratelimit())
> +                       pr_err("sungem wrong csum : %04x/%04x, len %u bytes\n",
> +                               csum, csum_fold(rsum), len);
> +                       print_hex_dump(KERN_ERR, "raw data: ", DUMP_PREFIX_OFFSET,


DUMP_PREFIX_ADDRESS might give us more information (say alignment problem, or crossing page boundaries)

> +                                           16, 1, skb->data, len, true);
> +               }
>                 skb->ip_summed = CHECKSUM_COMPLETE;
>                 skb->protocol = eth_type_trans(skb, gp->dev);
>  
>