On Wed, May 30, 2018 at 7:13 PM, Duncan Hare dh@synoia.com wrote:
Subject: Re: [U-Boot] [PATCH 2/2] net: ping,arp: Fix cache alignment issues Message-ID: 099414db-a264-a3ac-8aae-fb8aa0d969ee@denx.de Content-Type: text/plain; charset=utf-8; format=flowed
On 28.05.2018 08:33, Baruch Siach wrote:
From: Jon Nettleton jon@solid-run.com
Both ping_receive and arp_receive would transmit a received packet back out using its original point. This causes problems with certain network cards that add a custom header to the packet. Specifically the mvneta driver for the Armada series boards has a 2 byte Marvell header that is bypassed and passed along to the system, but that 2 byte offset now causes a misalignment if it is attempted to be sent back out.
Rather than changing the driver to memcpy all the received packets to cache aligned buffers we instead change the two offending network commands to copy the packet into a cache aligned net_tx_packet before sending it back out.
This fixes occasional messages like:
CACHE: Misaligned operation at range [3fc01082, 3fc010c2]
This is not a good idea. I believe it should be implemented in the driver. Putting it in in the app code leads to confusion as to its purpose.
This is merely aligning these two commands to work the same way that all the other net commands work. All of bootp.c, cdp.c, rarp.c, and lastly net.c use the global net_tx_packet, which is cache aligned, to pass a packet into the network driver stack. The patch is standardizing all of the net commands to behave the same way, along the way it fixes a problem that could exist with many drivers. Why debug this possible problem in every driver when it can be fixed in a common codeplace. This is most likely the reason that all the other libraries are written in this manner. I think it is more confusing for every network driver developer to possibly have to debug this issue because the behaviour is inconsistent between different network utilities.
-Jon