linux-kernel - Re: [PATCH v4 net-next 2/3] octeontx2-af: Knobs for NPC default rule counters

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20241104102822.GA1011185@hyd1403.caveonetworks.com>
Date: Mon, 4 Nov 2024 15:58:22 +0530
From: Linu Cherian <lcherian@...vell.com>
To: Jakub Kicinski <kuba@...nel.org>
CC: <davem@...emloft.net>, <sgoutham@...vell.com>, <netdev@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <gakula@...vell.com>,
        <hkelam@...vell.com>, <sbhatta@...vell.com>, <jerinj@...vell.com>,
        <edumazet@...gle.com>, <pabeni@...hat.comi>, <jiri@...nulli.us>,
        <corbet@....net>, <linux-doc@...r.kernel.org>
Subject: Re: [PATCH v4 net-next 2/3] octeontx2-af: Knobs for NPC default rule
 counters

Hi Jakub,

On 2024-11-04 at 01:23:10, Jakub Kicinski (kuba@...nel.org) wrote:
> On Tue, 29 Oct 2024 09:27:38 +0530 Linu Cherian wrote:
> > +	struct npc_install_flow_rsp rsp = { 0 };
> 
> @rsp is reused in the loop, either it doesn't have to be inited at all,
> or it has to be inited before every use

Ack. It doesnt have to be inited. Will remove.

> 
> > +	struct npc_mcam *mcam = &rvu->hw->mcam;
> > +	struct rvu_npc_mcam_rule *rule;
> > +	int blkaddr;
> > +
> > +	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_NPC, 0);
> > +	if (blkaddr < 0)
> > +		return -EINVAL;
> > +
> > +	mutex_lock(&mcam->lock);
> > +	list_for_each_entry(rule, &mcam->mcam_rules, list) {
> > +		if (!is_mcam_entry_enabled(rvu, mcam, blkaddr, rule->entry))
> > +			continue;
> > +		if (!rule->default_rule)
> > +			continue;
> > +		if (enable && !rule->has_cntr) { /* Alloc and map new counter */
> > +			__rvu_mcam_add_counter_to_rule(rvu, rule->owner,
> > +						       rule, &rsp);
> > +			if (rsp.counter < 0) {
> > +				dev_err(rvu->dev, "%s: Err to allocate cntr for default rule (err=%d)\n",
> > +					__func__, rsp.counter);
> > +				break;
> 
> shouldn't you "unwind" in this case? We'll leave the counter enabled
> for some rules and disabled for others


Wanted to keep a best effort approach here, will make it clear in the documentation.

> 
> > +			}
> > +			npc_map_mcam_entry_and_cntr(rvu, mcam, blkaddr,
> > +						    rule->entry, rsp.counter);
> > +		}
> > +
> > +		if (enable && rule->has_cntr) /* Reset counter before use */ {
> > +			rvu_write64(rvu, blkaddr,
> > +				    NPC_AF_MATCH_STATX(rule->cntr), 0x0);
> > +			continue;
> 
> so setting to enabled while already enabled resets the value?
> If so that's neither documented, nor.. usual.


Will move this code under,  if (enable && !rule->has_cntr) case.
Infact, all enablement happens only from devlink control, hence the
above if check is really not required and can be confusing. 

> 
> > +		}
> > +
> > +		if (!enable && rule->has_cntr) /* Free and unmap counter */ {
> > +			__rvu_mcam_remove_counter_from_rule(rvu, rule->owner,
> > +							    rule);
> > +		}
> 
> unnecesary parenthesis

Ack.


Thanks,
Linu Cherian.