linux-kernel - Re: [PATCH net v2] nfc: nci: Fix uninit-value in nci_rx

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] [day] [month] [year] [list]

Message-ID: <20240429164118.GB516117@kernel.org>
Date: Mon, 29 Apr 2024 17:41:18 +0100
From: Simon Horman <horms@...nel.org>
To: Ryosuke Yasuoka <ryasuoka@...hat.com>
Cc: krzk@...nel.org, davem@...emloft.net, edumazet@...gle.com,
	kuba@...nel.org, pabeni@...hat.com, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org, syoshida@...hat.com,
	syzbot+d7b4dc6cd50410152534@...kaller.appspotmail.com
Subject: Re: [PATCH net v2] nfc: nci: Fix uninit-value in nci_rx_work

On Mon, Apr 29, 2024 at 11:30:48PM +0900, Ryosuke Yasuoka wrote:
> On Sun, Apr 28, 2024 at 02:45:25PM +0100, Simon Horman wrote:
> > On Sat, Apr 27, 2024 at 07:35:54PM +0900, Ryosuke Yasuoka wrote:

..

> Thank you for your comment, Simon.
> 
> Yes, if it handles packets correctly in nci_{rsp,ntf,rx_data}_packet(),
> it should not reach invalid_pkt_free and it should continue to work in
> the for statement. Sorry, it is my mistake and need to fix it.
> 
> BTW, in the current implementation, if the payload is zero, it will free
> the skb and exit the for statement. I wonder it is intended. 
> 
> > > -		if (!nci_plen(skb->data)) {
> > > -			kfree_skb(skb);
> > > -			break;
> > > -		}
> 
> When the packet is invalid, it should be discarded but it should not
> exit the for statement by break. Instead, the skb should just free and
> it should handle the subsequent packet by continue. If yes, then it 
> may be like below,
> 
> 	for (; (skb = skb_dequeue(&ndev->rx_q)); kcov_remote_stop()) {
> 		kcov_remote_start_common(skb_get_kcov_handle(skb));
> 
> 		/* Send copy to sniffer */
> 		nfc_send_to_raw_sock(ndev->nfc_dev, skb,
> 				     RAW_PAYLOAD_NCI, NFC_DIRECTION_RX);
> 
> 		if (!skb->len)
> 			goto invalid_pkt_free;
> 
> 		/* Process frame */
> 		switch (nci_mt(skb->data)) {
> 		case NCI_MT_RSP_PKT:
> 			if (!nci_valid_size(skb, NCI_CTRL_HDR_SIZE))
> 				goto invalid_pkt_free;
> 			nci_rsp_packet(ndev, skb);
> 			continue;   <<<---
> 
> 		case NCI_MT_NTF_PKT:
> 			if (!nci_valid_size(skb, NCI_CTRL_HDR_SIZE))
> 				goto invalid_pkt_free;
> 			nci_ntf_packet(ndev, skb);
> 			continue;   <<<---
> 
> 		case NCI_MT_DATA_PKT:
> 			if (!nci_valid_size(skb, NCI_DATA_HDR_SIZE))
> 				goto invalid_pkt_free;
> 			nci_rx_data_packet(ndev, skb);
> 			continue;   <<<---
> 
> 		default:
> 			pr_err("unknown MT 0x%x\n", nci_mt(skb->data));
> 			goto invalid_pkt_free;
> 		}
> invalid_pkt_free:
> 		kfree_skb(skb);
> 	}
> 
> Could I hear your opinion?

Hi Yasuoka-san,

Thanks for pointing this out.

I agree that it is not good to 'break' after kfree_skb() for two reasons:

1. As you mention, the loop should keep going and process other skbs
2. kcov_remote_stop() needs to be called for each skb

I might have used a 'continue' above the invalid_pkt_free label.
But I think your suggestion - using 'continue' inside the switch statement
- is also correct, and seems fine to me.

Please post a v3 when you are ready.

-- 
pw-bot: changes-requested