Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:4 as permitted sender) client-ip=2620:137:e000::3:4;
Date:   Tue, 5 Dec 2023 19:40:50 -0800
From:   Jakub Kicinski <kuba@kernel.org>
To:     Justin Lai <justinlai0215@realtek.com>
Cc:     "davem@davemloft.net" <davem@davemloft.net>,
        "edumazet@google.com" <edumazet@google.com>,
        "pabeni@redhat.com" <pabeni@redhat.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "netdev@vger.kernel.org" <netdev@vger.kernel.org>,
        "andrew@lunn.ch" <andrew@lunn.ch>,
        Ping-Ke Shih <pkshih@realtek.com>,
        Larry Chiu <larry.chiu@realtek.com>
Subject: Re: [PATCH net-next v13 01/13] rtase: Add pci table supported in
 this module
Message-ID: <20231205194050.7033cc2b@kernel.org>
In-Reply-To: <ae4807e31b53452ebf176098d95cf1fb@realtek.com>
References: <20231130114327.1530225-1-justinlai0215@realtek.com>
        <20231130114327.1530225-2-justinlai0215@realtek.com>
        <20231201203602.7e380716@kernel.org>
        <ae4807e31b53452ebf176098d95cf1fb@realtek.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Wed, 6 Dec 2023 03:28:32 +0000 Justin Lai wrote:
> > > +static void rtase_remove_one(struct pci_dev *pdev) {
> > > +     struct net_device *dev = pci_get_drvdata(pdev);
> > > +     struct rtase_private *tp = netdev_priv(dev);
> > > +     struct rtase_int_vector *ivec;
> > > +     u32 i;
> > > +
> > > +     for (i = 0; i < tp->int_nums; i++) {
> > > +             ivec = &tp->int_vector[i];
> > > +             netif_napi_del(&ivec->napi);  
> > 
> > NAPI instances should be added on ndo_open()  
> 
> Do you want me to call netif_napi_add() in the .ndo_open function,
> and netif_napi_del() in the .ndo_stop function? However, I saw that
> many drivers do it in probe and remove. What is the purpose of doing
> this in .ndo_open and .ndo_stop?

They will sit in a fixed-size hash table used for NAPI lookup in the
core. Not a big deal, but not the best way either.

I think the main thing that prompted me to ask was that I couldn't find
napi_disable() in the first few patches. You should probably call it on
close, otherwise making sure NAPI is not running when you start freeing
rings is hard. synchronize_irq() will not help you at all if you're
using NAPI.