linux-kernel - Re: [PATCH] comedi: include poll wait state in busy assessment

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: <4c4400e0-dd8c-4694-8030-f35273e5522d@mev.co.uk>
Date: Tue, 22 Jul 2025 16:29:09 +0100
From: Ian Abbott <abbotti@....co.uk>
To: Jens Axboe <axboe@...nel.dk>, hsweeten@...ionengravers.com,
 Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc: LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] comedi: include poll wait state in busy assessment

On 22/07/2025 13:49, Jens Axboe wrote:
> syzbot reports a use-after-free in comedi in the below link, which is
> due to comedi gladly removing the allocated async area even though poll
> requests are still active on the wait_queue_head inside of it. This can
> cause a use-after-free when the poll entries are later triggered or
> removed, as the memory for the wait_queue_head has been freed. The
> notion of being busy in comedi seems mostly centered around sync
> syscalls, where the comedi subdevice stores the file in 'busy' as a
> marker. This is obviously broken for things like poll which can be
> persistent across syscalls.
> 
> Rename is_device_busy() to start_detach(), and move it under the
> dev->attach_lock. The latter serializes it with poll attempts. If
> start_detach() is successful, then it will have marked the device with
> ->detaching == 1 and this will prevent further poll attempts.
> 
> Similarly, have start_detach() check for active polls on the device, and
> return busy for that case.
> 
> Cc: stable@...r.kernel.org
> Fixes: 8ffdff6a8cfb ("staging: comedi: move out of staging directory")
> Link: https://lore.kernel.org/all/687bd5fe.a70a0220.693ce.0091.GAE@google.com/
> Reported-by: syzbot+01523a0ae5600aef5895@...kaller.appspotmail.com
> Tested-by: syzbot+01523a0ae5600aef5895@...kaller.appspotmail.com
> Signed-off-by: Jens Axboe <axboe@...nel.dk>
> 
> ---
> 
> diff --git a/drivers/comedi/comedi_fops.c b/drivers/comedi/comedi_fops.c
> index c83fd14dd7ad..58b034e45283 100644
> --- a/drivers/comedi/comedi_fops.c
> +++ b/drivers/comedi/comedi_fops.c
> @@ -782,24 +782,33 @@ void comedi_device_cancel_all(struct comedi_device *dev)
>   	}
>   }
>   
> -static int is_device_busy(struct comedi_device *dev)
> +static int start_detach(struct comedi_device *dev)
>   {
>   	struct comedi_subdevice *s;
> -	int i;
> +	int i, is_busy = 0;
>   
>   	lockdep_assert_held(&dev->mutex);
> +	lockdep_assert_held(&dev->attach_lock);
>   	if (!dev->attached)
>   		return 0;
>   
>   	for (i = 0; i < dev->n_subdevices; i++) {
>   		s = &dev->subdevices[i];
> -		if (s->busy)
> -			return 1;
> -		if (s->async && comedi_buf_is_mmapped(s))
> -			return 1;
> +		if (s->busy) {
> +			is_busy = 1;
> +			break;
> +		}
> +		if (!s->async)
> +			continue;
> +		if (comedi_buf_is_mmapped(s) ||
> +		    wq_has_sleeper(&s->async->wait_head)) {
> +			is_busy = 1;
> +			break;
> +		}
>   	}
> -
> -	return 0;
> +	if (!is_busy)
> +		dev->detaching = 1;
> +	return is_busy;
>   }
>   
>   /*
> @@ -825,8 +834,13 @@ static int do_devconfig_ioctl(struct comedi_device *dev,
>   		return -EPERM;
>   
>   	if (!arg) {
> -		if (is_device_busy(dev))
> +		/* prevent new polls */
> +		down_write(&dev->attach_lock);
> +		if (start_detach(dev)) {
> +			up_write(&dev->attach_lock);
>   			return -EBUSY;
> +		}
> +		up_write(&dev->attach_lock);
>   		if (dev->attached) {
>   			struct module *driver_module = dev->driver->module;
>   
> @@ -2479,7 +2493,7 @@ static __poll_t comedi_poll(struct file *file, poll_table *wait)
>   
>   	down_read(&dev->attach_lock);
>   
> -	if (!dev->attached) {
> +	if (!dev->attached || dev->detaching) {
>   		dev_dbg(dev->class_dev, "no driver attached\n");
>   		goto done;
>   	}
> diff --git a/include/linux/comedi/comedidev.h b/include/linux/comedi/comedidev.h
> index 4cb0400ad616..b2bec668785f 100644
> --- a/include/linux/comedi/comedidev.h
> +++ b/include/linux/comedi/comedidev.h
> @@ -545,6 +545,7 @@ struct comedi_device {
>   	const char *board_name;
>   	const void *board_ptr;
>   	unsigned int attached:1;
> +	unsigned int detaching:1;
>   	unsigned int ioenabled:1;
>   	spinlock_t spinlock;	/* generic spin-lock for low-level driver */
>   	struct mutex mutex;	/* generic mutex for COMEDI core */
> 

Thanks for the patch.  I'll post my version shortly.  One problem with 
this patch is that dev->detaching does not get cleared.

-- 
-=( Ian Abbott <abbotti@....co.uk> || MEV Ltd. is a company  )=-
-=( registered in England & Wales.  Regd. number: 02862268.  )=-
-=( Regd. addr.: S11 & 12 Building 67, Europa Business Park, )=-
-=( Bird Hall Lane, STOCKPORT, SK3 0XA, UK. || www.mev.co.uk )=-