linux-kernel - Re: [PATCH] scsi: wd33c93: fix buffer overflow in SCSI message-in handling

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: <aPcb8MMIJ2ve64yD@infradead.org>
Date: Mon, 20 Oct 2025 22:36:48 -0700
From: Christoph Hellwig <hch@...radead.org>
To: Yuhao Jiang <danisjiang@...il.com>
Cc: James.Bottomley@...senpartnership.com, martin.petersen@...cle.com,
	linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org,
	stable@...r.kernel.org
Subject: Re: [PATCH] scsi: wd33c93: fix buffer overflow in SCSI message-in
 handling

This exploit really needs a catchy name.  Just think of how much
valuable data you could extract by selling malicious fake 8-bit
SCSI disks to retro computing enthusiasts and then exploiting their
SCSI HBA driver.

On Mon, Oct 20, 2025 at 09:08:04PM -0500, Yuhao Jiang wrote:
> A buffer overflow vulnerability exists in the wd33c93 SCSI driver's
> message handling where missing bounds checking allows a malicious
> SCSI device to overflow the incoming_msg[] buffer and corrupt kernel
> memory.
> 
> The issue occurs because:
> - incoming_msg[] is a fixed 8-byte buffer (line 235 in wd33c93.h)
> - wd33c93_intr() writes to incoming_msg[incoming_ptr] without
>   validating incoming_ptr is within bounds (line 935)
> - For EXTENDED_MESSAGE, incoming_ptr increments based on the device-
>   supplied length field (line 1085) with no maximum check
> - The validation at line 1001 only checks if the message is complete,
>   not if it exceeds buffer size
> 
> This allows an attacker controlling a SCSI device to craft an extended
> message with length field 0xFF, causing the driver to write 256 bytes
> into an 8-byte buffer. This can corrupt adjacent fields in the
> WD33C93_hostdata structure including function pointers, potentially
> leading to arbitrary code execution.
> 
> Add bounds checking in the MESSAGE_IN handler to ensure incoming_ptr
> does not exceed buffer capacity before writing. Reject oversized
> messages per SCSI protocol by sending MESSAGE_REJECT.
> 
> Reported-by: Yuhao Jiang <danisjiang@...il.com>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Cc: stable@...r.kernel.org
> Signed-off-by: Yuhao Jiang <danisjiang@...il.com>
> ---
>  drivers/scsi/wd33c93.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/scsi/wd33c93.c b/drivers/scsi/wd33c93.c
> index dd1fef9226f2..2d50a0a01726 100644
> --- a/drivers/scsi/wd33c93.c
> +++ b/drivers/scsi/wd33c93.c
> @@ -932,6 +932,19 @@ wd33c93_intr(struct Scsi_Host *instance)
>  		sr = read_wd33c93(regs, WD_SCSI_STATUS);	/* clear interrupt */
>  		udelay(7);
>  
> +		/* Prevent buffer overflow from malicious extended messages */
> +		if (hostdata->incoming_ptr >= sizeof(hostdata->incoming_msg)) {
> +			printk("wd33c93: Incoming message too long, rejecting\n");
> +			hostdata->incoming_ptr = 0;
> +			write_wd33c93_cmd(regs, WD_CMD_ASSERT_ATN);
> +			hostdata->outgoing_msg[0] = MESSAGE_REJECT;
> +			hostdata->outgoing_len = 1;
> +			write_wd33c93_cmd(regs, WD_CMD_NEGATE_ACK);
> +			hostdata->state = S_CONNECTED;
> +			spin_unlock_irqrestore(&hostdata->lock, flags);
> +			break;
> +		}
> +
>  		hostdata->incoming_msg[hostdata->incoming_ptr] = msg;
>  		if (hostdata->incoming_msg[0] == EXTENDED_MESSAGE)
>  			msg = EXTENDED_MESSAGE;
> -- 
> 2.34.1
> 
> 
---end quoted text---