linux-kernel - Re: [PATCH] drivers: pnp: proc.c: Handle errors while attaching devices

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: <CSBw5RbPfjHIC-pzsVOv-S6X9ir8mhLYp3cP3P6UizGPb-wc-OUBfXFXvAPr7fISCC2Bo_r1hHsoltEEEK7VXi8UyTNHc51lH_att8eNZqk=@protonmail.com>
Date:   Sun, 25 Apr 2021 01:06:06 +0000
From:   Barnabás Pőcze <pobrn@...tonmail.com>
To:     Anupama K Patil <anupamakpatil123@...il.com>
Cc:     Jaroslav Kysela <perex@...ex.cz>,
        "Rafael J. Wysocki" <rafael.j.wysocki@...el.com>,
        "linux-acpi@...r.kernel.org" <linux-acpi@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "skhan@...uxfoundation.org" <skhan@...uxfoundation.org>,
        "bkkarthik@...u.pes.edu" <bkkarthik@...u.pes.edu>,
        "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
        "kernelnewbies@...nelnewbies.org" <kernelnewbies@...nelnewbies.org>
Subject: Re: [PATCH] drivers: pnp: proc.c: Handle errors while attaching devices

Hi


2021. április 24., szombat 21:43 keltezéssel, Anupama K Patil írta:

> isapnp_proc_init() does not look at the return value from
> isapnp_proc_attach_device(). Check for this return value in
> isapnp_proc_detach_device().
>
> Cleanup in isapnp_proc_detach_device and
> isapnp_proc_detach_bus() for cleanup.
>
> Changed sprintf() to the kernel-space function scnprintf() as it returns
> the actual number of bytes written.
>
> Removed unnecessary variables de, e of type 'struct proc_dir_entry' to
> save memory.
>
> Suggested-by: Shuah Khan <skhan@...uxfoundation.org>
> Co-developed-by: B K Karthik <bkkarthik@...u.pes.edu>
> Signed-off-by: B K Karthik <bkkarthik@...u.pes.edu>
> Signed-off-by: Anupama K Patil <anupamakpatil123@...il.com>
> ---
>  drivers/pnp/isapnp/proc.c | 40 +++++++++++++++++++++++++++++----------
>  1 file changed, 30 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/pnp/isapnp/proc.c b/drivers/pnp/isapnp/proc.c
> index 785a796430fa..46ebc24175b7 100644
> --- a/drivers/pnp/isapnp/proc.c
> +++ b/drivers/pnp/isapnp/proc.c
> @@ -54,34 +54,54 @@ static const struct proc_ops isapnp_proc_bus_proc_ops = {
>  	.proc_read	= isapnp_proc_bus_read,
>  };
>
> +static int isapnp_proc_detach_device(struct pnp_dev *dev)
> +{
> +	proc_remove(dev->procent);
> +	dev->procent = NULL;
> +	return 0;
> +}
> +
> +static int isapnp_proc_detach_bus(struct pnp_card *bus)
> +{
> +	proc_remove(bus->procdir);

Is there any reason for not setting `bus->procdir` to `NULL`
similarly to the previous function?


> +	return 0;
> +}
> +

Is there any reason why the previous two functions return something? It doesn't
seem to be necessary.


>  static int isapnp_proc_attach_device(struct pnp_dev *dev)
>  {
>  	struct pnp_card *bus = dev->card;
> -	struct proc_dir_entry *de, *e;
>  	char name[16];
>
> -	if (!(de = bus->procdir)) {
> -		sprintf(name, "%02x", bus->number);
> -		de = bus->procdir = proc_mkdir(name, isapnp_proc_bus_dir);
> -		if (!de)
> +	if (!bus->procdir) {
> +		scnprintf(name, 16, "%02x", bus->number);

I think `sizeof(name)` would be preferable to hard-coding 16.


> +		bus->procdir = proc_mkdir(name, isapnp_proc_bus_dir);
> +		if (!bus->procdir)
>  			return -ENOMEM;
>  	}
> -	sprintf(name, "%02x", dev->number);
> -	e = dev->procent = proc_create_data(name, S_IFREG | S_IRUGO, de,
> +	scnprintf(name, 16, "%02x", dev->number);

Here as well.


> +	dev->procent = proc_create_data(name, S_IFREG | S_IRUGO, bus->procdir,
>  					    &isapnp_proc_bus_proc_ops, dev);

Please align the continuation properly.


> -	if (!e)
> +	if (!dev->procent) {
> +		isapnp_proc_detach_bus(bus);

I'm not sure if this should be here. If I'm not mistaken, the code
creates a procfs directory for a bus when it first sees a `pnp_dev` from that bus.
This call removes the whole directory for the bus, and with that, the files of
those `pnp_dev`s which were successfully created earlier.


>  		return -ENOMEM;
> -	proc_set_size(e, 256);
> +	}
> +	proc_set_size(dev->procent, 256);
>  	return 0;
>  }
>
>  int __init isapnp_proc_init(void)
>  {
>  	struct pnp_dev *dev;
> +	int dev_attach;
>
>  	isapnp_proc_bus_dir = proc_mkdir("bus/isapnp", NULL);

You could add a check to see if this `proc_mkdir()` call succeeds, and
possibly return early if it does not.


>  	protocol_for_each_dev(&isapnp_protocol, dev) {
> -		isapnp_proc_attach_device(dev);
> +		dev_attach = isapnp_proc_attach_device(dev);
> +		if (!dev_attach) {

`isapnp_proc_attach_device()` returns 0 on success, so the condition should be inverted.
And maybe `err` or something like that would be a better name than `dev_attach`.


> +			pr_info("procfs: pnp: Unable to attach the device, not enough memory");

If I'm not mistaken, allocation failures are logged, so this is probably not needed.


> +			isapnp_proc_detach_device(dev);

I'm also not sure if this is needed here. If `isapnp_proc_attach_device()` returns
an error, then `dev->procdir` could not have been "created". In other words,
if the execution reaches this point, `proc_create_data()` could not have succeeded
because either it had not yet been called or it had failed.


> +			return -ENOMEM;

It is usually preferable to return the error code you receive. E.g.:

  err = isapnp_proc_attach_device(...);
  if (err) {
    ...
    return err;
  }


> +		}
>  	}
>  	return 0;
>  }
> --
> 2.25.1
>


Regards,
Barnabás Pőcze