[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 16 Oct 2020 18:37:59 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Hemant Kumar <hemantk@...eaurora.org>
Cc: Loic Poulain <loic.poulain@...aro.org>, davem@...emloft.net,
netdev@...r.kernel.org, manivannan.sadhasivam@...aro.org,
eric.dumazet@...il.com
Subject: Re: [PATCH v6 2/3] net: Add mhi-net driver
Please trim your replies.
On Fri, 16 Oct 2020 17:52:28 -0700 Hemant Kumar wrote:
> > +#include <linux/if_arp.h>
> > +#include <linux/mhi.h>
> > +#include <linux/mod_devicetable.h>
> > +#include <linux/module.h>
> > +#include <linux/netdevice.h>
> > +#include <linux/skbuff.h>
> > +#include <linux/u64_stats_sync.h>
> > +
> > +#define MIN_MTU ETH_MIN_MTU
> > +#define MAX_MTU 0xffff
> uci driver patch series would take care of this :)
References? What's the UCI patch series?
> > + if (mhi_queue_is_full(mdev, DMA_TO_DEVICE))
> > + netif_stop_queue(ndev);
> is it possible to enable flow control on tx queue if mhi_queue_skb
> returns non zero value and return NETDEV_TX_BUSY instead of dropping
> packet at MHI layer ? If this is possible you dont need to export new
> API as well.
That's not the preferred implementation. Drivers should stop the queues
in advance to avoid requeuing to the qdiscs.
> > + ndev->tx_queue_len = 1000;
> some of the above parameters can go in to driver data specific to sw and
> hw channels.
Really curious to hear what you're talking about here.
Is this already in tree? Was netdev CCed on it?
Powered by blists - more mailing lists