Re: [PATCH] tpm_vtpm_proxy: fix race condition in /dev/vtpmx creation

["Jarkko Sakkinen" <jarkko@kernel.org>]

On Sat May 13, 2023 at 8:28 PM EEST, Jarkko Sakkinen wrote:
> /dev/vtpmx is made visible before the workqueue exist, which can lead
> memory corruption in the worst case, if workqueue is used before it has
> been fully initialized. Address this by changing the call order.
>
> Cc: Stefan Berger <stefanb@linux.vnet.ibm.com>
> Cc: stable@vger.kernel.org
> Fixes: 6f99612e2500 ("tpm: Proxy driver for supporting multiple emulated TPMs")
> Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@tuni.fi>
> ---
>  drivers/char/tpm/tpm_vtpm_proxy.c | 26 ++++++++++----------------
>  1 file changed, 10 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/char/tpm/tpm_vtpm_proxy.c b/drivers/char/tpm/tpm_vtpm_proxy.c
> index 5c865987ba5c..ef1367cf2f10 100644
> --- a/drivers/char/tpm/tpm_vtpm_proxy.c
> +++ b/drivers/char/tpm/tpm_vtpm_proxy.c
> @@ -50,7 +50,7 @@ struct proxy_dev {
>  /* all supported flags */
>  #define VTPM_PROXY_FLAGS_ALL  (VTPM_PROXY_FLAG_TPM2)
>  
> -static struct workqueue_struct *workqueue;
> +static struct workqueue_struct *vtpm_workqueue;
>  
>  static void vtpm_proxy_delete_device(struct proxy_dev *proxy_dev);
>  
> @@ -478,7 +478,7 @@ static void vtpm_proxy_work_stop(struct proxy_dev *proxy_dev)
>   */
>  static inline void vtpm_proxy_work_start(struct proxy_dev *proxy_dev)
>  {
> -	queue_work(workqueue, &proxy_dev->work);
> +	queue_work(vtpm_workqueue, &proxy_dev->work);
>  }
>  
>  /*
> @@ -697,30 +697,24 @@ static int __init vtpm_module_init(void)
>  {
>  	int rc;
>  
> -	rc = vtpmx_init();
> -	if (rc) {
> -		pr_err("couldn't create vtpmx device\n");
> +	vtpm_workqueue = create_workqueue("tpm-vtpm");
> +	if (!vtpm_workqueue) {
> +		rc = -ENOMEM;
>  		return rc;
>  	}
>  
> -	workqueue = create_workqueue("tpm-vtpm");
> -	if (!workqueue) {
> -		pr_err("couldn't create workqueue\n");
> -		rc = -ENOMEM;
> -		goto err_vtpmx_cleanup;
> +	rc = vtpmx_init();
> +	if (rc) {
> +		vtpmx_cleanup();

This should be destroy_workqueue(). I'll send a new version.

BR, Jarkko