Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 12 Oct 2022 16:31:48 +0300
From:   Vladimir Oltean <olteanv@gmail.com>
To:     Christian Marangi <ansuelsmth@gmail.com>
Cc:     Andrew Lunn <andrew@lunn.ch>,
        Vivien Didelot <vivien.didelot@gmail.com>,
        Florian Fainelli <f.fainelli@gmail.com>,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        "Russell King (Oracle)" <rmk+kernel@armlinux.org.uk>,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        Pawel Dembicki <paweldembicki@gmail.com>,
        Lech Perczak <lech.perczak@gmail.com>
Subject: Re: [net PATCH 1/2] net: dsa: qca8k: fix inband mgmt for big-endian
 systems
Message-ID: <20221012133148.6apqbip3kvnjuafu@skbuf>
References: <20221010111459.18958-1-ansuelsmth@gmail.com>
 <Y0a2KD9pVeoYkHkK@lunn.ch>
 <6346b921.a70a0220.64f4e.a0bc@mx.google.com>
 <6346ba28.050a0220.f0e18.949b@mx.google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <6346ba28.050a0220.f0e18.949b@mx.google.com>
Precedence: bulk

On Wed, Oct 12, 2022 at 02:59:17PM +0200, Christian Marangi wrote:
> > > Humm...
> > > 
> > > This might have the same alignment issue as the second patch. In fact,
> > > because the Ethernet header is 14 bytes in size, it is often
> > > deliberately out of alignment by 2 bytes, so that the IP header is
> > > aligned. You should probably be using get_unaligned_le32() when
> > > accessing members of mgmt_ethhdr.
> > > 
> > > 	  Andrew
> > 
> > Should I replace everything to get_unaligned_le32? Or this is only
> > needed for the mgmt_ethhdr as it's 12 bytes?
> > 
> > The skb data is all 32 bit contiguous stuff so it should be safe? Or
> > should we treat that also as unalligned just to make sure?
> > 
> > Same question for patch 2. the rest of the mib in skb data are all 32 or
> > 64 values contiguous so wonder if we just take extra care of the
> > mgmt_ethhdr. 
> > 
> 
> Also also... Should I use put_unalligned to fill the mgmt_ethhdr?

Documentation/core-api/unaligned-memory-access.rst section "Alignment vs. Networking"
says that the IP header is aligned to a 4 byte boundary.

Relative to the IP header, skb_mac_header(skb) is a pointer that's 14
bytes behind, right?

14 bytes behind something aligned to a 4 byte boundary is something
that's not aligned to a 4 byte boundary. That's why Andrew is suggesting
to use the unaligned helper for accesses (both put and get).

On-stack data structures don't need this, the compiler should take care
of aligning them and their fields appropriately. The trouble is with
pointers generated using manual arithmetics.