netdev - Re: [PATCH net-next] net: dsa: b53: Slightly optimize b53_arl

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] [day] [month] [year] [list]

Message-ID: <4385d07a-e124-d2f2-fbd0-eda1d602167b@wanadoo.fr>
Date:   Fri, 21 Apr 2023 07:40:00 +0200
From:   Christophe JAILLET <christophe.jaillet@...adoo.fr>
To:     Florian Fainelli <f.fainelli@...il.com>,
        Andrew Lunn <andrew@...n.ch>,
        Vladimir Oltean <olteanv@...il.com>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>
Cc:     linux-kernel@...r.kernel.org, kernel-janitors@...r.kernel.org,
        netdev@...r.kernel.org
Subject: Re: [PATCH net-next] net: dsa: b53: Slightly optimize b53_arl_read()

Le 21/04/2023 à 02:40, Florian Fainelli a écrit :
> 
> 
> On 4/20/2023 1:44 PM, Christophe JAILLET wrote:
>> When the 'free_bins' bitmap is cleared, it is better to use its full
>> maximum size instead of only the needed size.
>> This lets the compiler optimize it because the size is now known at 
>> compile
>> time. B53_ARLTBL_MAX_BIN_ENTRIES is small (i.e. currently 4), so a 
>> call to
>> memset() is saved.
>>
>> Also, as 'free_bins' is local to the function, the non-atomic __set_bit()
>> can also safely be used here.
>>
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@...adoo.fr>
>> ---
>>   drivers/net/dsa/b53/b53_common.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/net/dsa/b53/b53_common.c 
>> b/drivers/net/dsa/b53/b53_common.c
>> index 3464ce5e7470..8c55fe0e0747 100644
>> --- a/drivers/net/dsa/b53/b53_common.c
>> +++ b/drivers/net/dsa/b53/b53_common.c
>> @@ -1627,7 +1627,7 @@ static int b53_arl_read(struct b53_device *dev, 
>> u64 mac,
>>       if (ret)
>>           return ret;
>> -    bitmap_zero(free_bins, dev->num_arl_bins);
>> +    bitmap_zero(free_bins, B53_ARLTBL_MAX_BIN_ENTRIES);
> 
> That one I am not a big fan, as the number of ARL bins is a function of 
> the switch model, and this illustrates it well.

Ok, up to you to take or not what looks the better solution.

 From my point of view, the "for (i = 0; i < dev->num_arl_bins" below 
illustrates it better.


Maybe, another approach to save the memset() call would be remove the 
bitmap_zero() call, and declare 'free_bins' as:

    DECLARE_BITMAP(free_bins, B53_ARLTBL_MAX_BIN_ENTRIES) = { };
(this syntax is already used in b53_configure_vlan())


The compiler should still be able to optimize the initialisation and 
this wouldn't, IMHO, introduce confusion about the intent.

Let me know if you prefer to leave this hunk as-is, or if this other 
alternative pleases you.


CJ

> 
>>       /* Read the bins */
>>       for (i = 0; i < dev->num_arl_bins; i++) {
>> @@ -1641,7 +1641,7 @@ static int b53_arl_read(struct b53_device *dev, 
>> u64 mac,
>>           b53_arl_to_entry(ent, mac_vid, fwd_entry);
>>           if (!(fwd_entry & ARLTBL_VALID)) {
>> -            set_bit(i, free_bins);
>> +            __set_bit(i, free_bins);
> 
> I would be keen on taking that hunk but keep the other as-is. Does that 
> work for you?
> -- 
> Florian
>