netdev - Re: [PATCH] net: switchdev: restrict vid range abstraction

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: <1087251318.172776.1438390934326.JavaMail.zimbra@savoirfairelinux.com>
Date:	Fri, 31 Jul 2015 21:02:14 -0400 (EDT)
From:	Vivien Didelot <vivien.didelot@...oirfairelinux.com>
To:	Scott Feldman <sfeldma@...il.com>
Cc:	David <davem@...emloft.net>, netdev <netdev@...r.kernel.org>,
	Jiří Pírko <jiri@...nulli.us>,
	Florian Fainelli <f.fainelli@...il.com>,
	linux-kernel <linux-kernel@...r.kernel.org>,
	kernel <kernel@...oirfairelinux.com>
Subject: Re: [PATCH] net: switchdev: restrict vid range abstraction

Hi Scott,

On Jul 29, 2015, at 5:17 PM, Scott Feldman sfeldma@...il.com wrote:

> On Wed, Jul 29, 2015 at 12:14 PM, Vivien Didelot
> <vivien.didelot@...oirfairelinux.com> wrote:
>> Hi Scott, David,
>>
>> On Jul 29, 2015, at 2:28 PM, David davem@...emloft.net wrote:
>>
>>> From: Scott Feldman <sfeldma@...il.com>
>>> Date: Wed, 29 Jul 2015 00:31:44 -0700
>>>
>>>> Since the netlink request (for example vlan add) includes the range,
>>>> I'm not seeing how we can response with success for the satisfied
>>>> vlans in the range, and also respond with an error for the unsatisfied
>>>> vlans in the range.   In other words, from the netlink msgs
>>>> perspective, we need to treat a vlan range as all-or-nothing.  So in
>>>> your example, if hw can't add vlan 2, we fail the entire request to
>>>> add range 2-5.  This is where the prepare phase checks to make sure
>>>> the entire request can be satisfied before committing to hw.
>>
>> I made this change in order to start restricting the bridge abstraction
>> to switchdev, since IMHO its info flags do not add much value to the
>> switch chip drivers perspective.
>>
>> While a range might be convenient to a user, exposing it to drivers is
>> likely to end up writing the same vid_begin to vid_end for loop.
>>
>>> This was my concern with the change as well.
>>>
>>> The user asked for the range to be installed, so if any portion
>>> of it cannot be done we must not make any changes to the HW
>>> configuration and fail the entire request.
>>
>> I understand the concern with the netlink request.
>>
>> However, this can be confusing to someone. With the previous example:
>>
>>     bridge vlan add dev port0 vid 2-5 master
>>
>> must fail for the entire range (due to the single netlink request). But:
>>
>>     bridge vlan add dev port0 vid 2 master
>>
>> will silently fallback to software VLAN (assuming that the driver
>> correctly returned -EOPNOTSUPP in the prepare phase). In other words, no
>> changes has been committed to the hardware.
> 
> I see your concern now, I think.  net/bridge/br_netlink.c:br_afspec()
> does the range loop but doesn't rewind if something goes wrong with
> one of the vlans in the range.  The call into switchdev is
> one-at-a-time at that point.  If br_afspec() handled the rewind, would
> this address your concern?  We can keep the range support in the
> switchdev vlan obj, so 'self' can use it.

I am not sure is the rewind is needed. My concern was trying to handle
the fallback to software VLAN for a single VID within a range, so that
we can free a switch chip driver for this bridge-specific notion. But
because of the single netlink request, it seems not possible.

At which level does this fallback happen exactly?

Thanks,
-v
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html