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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <2d524100-7e34-7869-5e90-415614b767f3@ti.com>
Date: Thu, 12 Oct 2023 22:08:58 +0530
From: "Anwar, Md Danish" <a0501179@...com>
To: Andrew Lunn <andrew@...n.ch>, MD Danish Anwar <danishanwar@...com>
CC: Paolo Abeni <pabeni@...hat.com>, Jakub Kicinski <kuba@...nel.org>,
        Eric
 Dumazet <edumazet@...gle.com>,
        "David S. Miller" <davem@...emloft.net>,
        <linux-kernel@...r.kernel.org>, <netdev@...r.kernel.org>, <srk@...com>,
        Vignesh Raghavendra <vigneshr@...com>, <r-gunasekaran@...com>,
        Roger Quadros
	<rogerq@...nel.org>
Subject: Re: [PATCH] net: ti: icssg-prueth: Fix tx_total_bytes count

On 10/12/2023 8:58 PM, Andrew Lunn wrote:
> On Thu, Oct 12, 2023 at 10:51:12AM +0530, MD Danish Anwar wrote:
>> Hi Andrew,
>>
>> On 11/10/23 18:11, Andrew Lunn wrote:
>>>> @@ -29,7 +30,12 @@ void emac_update_hardware_stats(struct prueth_emac *emac)
>>>>  			     base + icssg_all_stats[i].offset,
>>>>  			     val);
>>>>  
>>>> +		if (!strncmp(icssg_ethtool_stats[i].name, "tx_good_frames", ETH_GSTRING_LEN))
>>>> +			tx_pkt_cnt = val;
>>>
>>> Doing a strncmp seems very expensive. Could you make use of
>>> icssg_stats.offset?
>>>
>>
>> Sure. I can define the offset of these two stats and then use them in if
>> condition as below.
>>
>> #define ICSSG_TX_PACKET_OFFSET 0xA0
>> #define ICSSG_TX_BYTE_OFFSET   0xEC
>>
>> if (icssg_ethtool_stats[i].offset == ICSSG_TX_PACKET_OFFSET)
>> 	tx_pkt_cnt = val;
>>
>> if (icssg_ethtool_stats[i].offset == ICSSG_TX_BYTE_OFFSET)
>> 	emac->stats[i] -= tx_pkt_cnt * 8;
> 
> That is much better. Also consider adding something like:
> 
> BUILD_BUG_ON(ICSSG_TX_PACKET_OFFSET < ICSSG_TX_BYTE_OFFSET)
> 
> I've no idea if this is correct. Just something to prove at build time
> that ICSSG_TX_PACKET_OFFSET is read before ICSSG_TX_BYTE_OFFSET.
> 

These registers are defined sequentially in the structure
miig_stats_regs. The offset for rx_packets is 0x0, rx_broadcast_frames
is 0x4 and so on. Basically the offset for i'th stat is i * sizeof(u32).

In the structure, tx_packet is defined first (index 40, offset 0xA0) and
then tx_bytes is defined (index 59, offset 0xEC).

In emac_update_hardware_stats() all these registers are read
sequentially. Meaning first tx_packet register is read and then tx_byte.

emac_update_hardware_stats() is called every 25s (by workqueue). Every
time first tx_packet is read and then tx_byte. So every time we are
decrementing tx_bytes by 8 bytes * num of packets, the num of packets
always exists and it is read before doing this calculation.

So I don't think any check is required to make sure
ICSSG_TX_PACKET_OFFSET is read before ICSSG_TX_BYTE_OFFSET.

The hardware design is such a way that these registers are read in a
sequence and the same sequence is followed in driver (struct
miig_stats_regs)

>      Andrew

-- 
Thanks and Regards,
Md Danish Anwar

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ