netdev - Re: [PATCH V3 net-next 06/10] net: hns3: add ethtool priv-flag for DIM

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]

Date:   Mon, 16 Nov 2020 16:41:45 +0800
From:   tanhuazhong <tanhuazhong@...wei.com>
To:     Jakub Kicinski <kuba@...nel.org>,
        Michal Kubecek <mkubecek@...e.cz>,
        Andrew Lunn <andrew@...n.ch>, Jiri Pirko <jiri@...nulli.us>
CC:     <davem@...emloft.net>, <netdev@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <salil.mehta@...wei.com>,
        <yisen.zhuang@...wei.com>, <linuxarm@...wei.com>
Subject: Re: [PATCH V3 net-next 06/10] net: hns3: add ethtool priv-flag for
 DIM



On 2020/11/15 2:54, Jakub Kicinski wrote:
> On Thu, 12 Nov 2020 11:33:14 +0800 Huazhong Tan wrote:
>> Add a control private flag in ethtool for enable/disable
>> DIM feature.
>>
>> Signed-off-by: Huazhong Tan <tanhuazhong@...wei.com>
> 
> Please work on a common ethtool API for the configuration instead of
> using private flags.
> 
> Private flags were overused because the old IOCTL-based ethtool was
> hard to extend, but we have a netlink API now.
> 
> For example here you're making a choice between device and DIM
> implementation of IRQ coalescing. You can add a new netlink attribute
> to the ETHTOOL_MSG_COALESCE_GET/ETHTOOL_MSG_COALESCE_SET commands which
> controls the type of adaptive coalescing (if enabled).
> 

The device's implementation of IRQ coalescing will be removed, if DIM 
works ok for a long time. So could this private flag for DIM be 
uptreamed as a transition scheme? And adding a new netlink attrtibute to 
controls the type of adaptive coalescing seems useless for other drivers.

> 
> One question I don't think we have a strong answer for is how to handle
> this extension from ethtool_ops point of view. Should we add a new
> "extended" op which drivers may start implementing? Or separate the
> structure passed in to the ops from the one used as uAPI?
> 
> Thoughts anyone?
> 
> 
> Huazhong Tan, since the DIM and EQ/CQ patches may require more
> infrastructure work feel free to repost the first 4 patches separately,
> I can apply those as is.
> 

ok, thanks.

> .
>