Date:   Wed, 24 Nov 2021 19:15:02 -0800
From:   Jakub Kicinski <kuba@kernel.org>
To:     Volodymyr Mytnyk <volodymyr.mytnyk@plvision.eu>
Cc:     netdev@vger.kernel.org, Taras Chornyi <taras.chornyi@plvision.eu>,
        Mickey Rachamim <mickeyr@marvell.com>,
        Serhiy Pshyk <serhiy.pshyk@plvision.eu>,
        Volodymyr Mytnyk <vmytnyk@marvell.com>,
        Taras Chornyi <tchornyi@marvell.com>,
        "David S. Miller" <davem@davemloft.net>,
        linux-kernel@vger.kernel.org,
        Yevhen Orlov <yevhen.orlov@plvision.eu>
Subject: Re: [PATCH net-next 1/3] net: prestera: acl: migrate to new vTCAM
 api
Message-ID: <20211124191502.5b497e84@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
In-Reply-To: <1637686684-2492-2-git-send-email-volodymyr.mytnyk@plvision.eu>
References: <1637686684-2492-1-git-send-email-volodymyr.mytnyk@plvision.eu>
        <1637686684-2492-2-git-send-email-volodymyr.mytnyk@plvision.eu>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Tue, 23 Nov 2021 18:58:00 +0200 Volodymyr Mytnyk wrote:
> From: Volodymyr Mytnyk <vmytnyk@marvell.com>
> 
> - Add new vTCAM HW API to configure HW ACLs.
> - Migrate acl to use new vTCAM HW API.
> - No counter support in this patch-set.
> 
> Co-developed-by: Yevhen Orlov <yevhen.orlov@plvision.eu>
> Signed-off-by: Yevhen Orlov <yevhen.orlov@plvision.eu>
> Signed-off-by: Volodymyr Mytnyk <vmytnyk@marvell.com>

>  struct prestera_acl_ruleset {
> +	struct rhash_head ht_node; /* Member of acl HT */
> +	struct prestera_acl_ruleset_ht_key ht_key;
>  	struct rhashtable rule_ht;
> -	struct prestera_switch *sw;
> -	u16 id;
> +	struct prestera_acl *acl;
> +	unsigned long rule_count;
> +	refcount_t refcount;
> +	void *keymask;
> +	bool offload;
> +	u32 vtcam_id;
> +	u16 pcl_id;

put the pcl_id earlier for better packing?

>  };

> +struct prestera_acl_vtcam {
> +	struct list_head list;
> +	__be32 keymask[__PRESTERA_ACL_RULE_MATCH_TYPE_MAX];
> +	bool is_keymask_set;
> +	refcount_t refcount;
> +	u8 lookup;

same here, 1B types together

>  	u32 id;
>  };

> +int prestera_acl_ruleset_keymask_set(struct prestera_acl_ruleset *ruleset,
> +				     void *keymask)
>  {
> -	prestera_hw_acl_ruleset_del(ruleset->sw, ruleset->id);
> -	rhashtable_destroy(&ruleset->rule_ht);
> -	kfree(ruleset);
> +	void *__keymask;
> +
> +	if (!keymask || !ruleset)

Can this legitimately happen? No defensive programming, please.

> +		return -EINVAL;
> +
> +	__keymask = kmalloc(ACL_KEYMASK_SIZE, GFP_KERNEL);
> +	if (!__keymask)
> +		return -ENOMEM;
> +
> +	memcpy(__keymask, keymask, ACL_KEYMASK_SIZE);

kmemdup()

> +	ruleset->keymask = __keymask;
> +
> +	return 0;
>  }