[PATCH] tty: nozomi: Fix a resource leak in an error handling function

[PATCH] tty: nozomi: Fix a resource leak in an error handling function

[Christophe JAILLET]

A 'request_irq()' call is not balanced by a corresponding 'free_irq()' in
the error handling path, as already done in the remove function.

Add it.

Fixes: 9842c38e9176 ("kfifo: fix warn_unused_result")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
I also wonder if the loop above is correct. The 'i < MAX_PORT' looks really
spurious to me.
'tty_port_destroy' can be called twice for the same entry (once before
branching in the error handling path, and once in here) and
'tty_unregister_device'/'tty_port_destroy' will be called on entries
that have not been 'tty_port_init'ed or 'tty_port_register_device'd.
I don't know if it may be an issue.
---
 drivers/tty/nozomi.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/tty/nozomi.c b/drivers/tty/nozomi.c
index 9a2d78ace49b..b270e137ef9b 100644
--- a/drivers/tty/nozomi.c
+++ b/drivers/tty/nozomi.c
@@ -1420,6 +1420,7 @@ static int nozomi_card_init(struct pci_dev *pdev,
 		tty_unregister_device(ntty_driver, dc->index_start + i);
 		tty_port_destroy(&dc->port[i].port);
 	}
+	free_irq(pdev->irq, dc);
 err_free_kfifo:
 	for (i = 0; i < MAX_PORT; i++)
 		kfifo_free(&dc->port[i].fifo_ul);
-- 
2.30.2

Re: [PATCH] tty: nozomi: Fix a resource leak in an error handling function

[Dan Carpenter]

On Sun, May 09, 2021 at 07:22:33PM +0200, Christophe JAILLET wrote:
> A 'request_irq()' call is not balanced by a corresponding 'free_irq()' in
> the error handling path, as already done in the remove function.
> 
> Add it.
> 
> Fixes: 9842c38e9176 ("kfifo: fix warn_unused_result")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> I also wonder if the loop above is correct. The 'i < MAX_PORT' looks really
> spurious to me.
> 'tty_port_destroy' can be called twice for the same entry (once before
> branching in the error handling path, and once in here) and
> 'tty_unregister_device'/'tty_port_destroy' will be called on entries
> that have not been 'tty_port_init'ed or 'tty_port_register_device'd.
> I don't know if it may be an issue.


Calling tty_port_destroy() twice is fine, but I think calling
tty_unregister_device() for unregistered devices will lead to a NULL
dereference in cdev_del().

regards,
dan carpenter

Re: [PATCH] tty: nozomi: Fix a resource leak in an error handling function

[Jiri Slaby]

On 09. 05. 21, 19:22, Christophe JAILLET wrote:
> A 'request_irq()' call is not balanced by a corresponding 'free_irq()' in
> the error handling path, as already done in the remove function.
> 
> Add it.
> 
> Fixes: 9842c38e9176 ("kfifo: fix warn_unused_result")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>

Reviewed-by: Jiri Slaby <jirislaby@kernel.org>

> ---
> I also wonder if the loop above is correct. The 'i < MAX_PORT' looks really
> spurious to me.
> 'tty_port_destroy' can be called twice for the same entry (once before
> branching in the error handling path, and once in here) and
> 'tty_unregister_device'/'tty_port_destroy' will be called on entries
> that have not been 'tty_port_init'ed or 'tty_port_register_device'd.
> I don't know if it may be an issue.

Yes. The fail path handling is very broken there. Both the code of 
err_free_tty label, and of the err_free_kfifo label. The loops should 
have been _something_ (I didn't invest much thinking into it, so it's 
likely wrong) like:
for (i--; i--; ) {
    ...
}


> ---
>   drivers/tty/nozomi.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/tty/nozomi.c b/drivers/tty/nozomi.c
> index 9a2d78ace49b..b270e137ef9b 100644
> --- a/drivers/tty/nozomi.c
> +++ b/drivers/tty/nozomi.c
> @@ -1420,6 +1420,7 @@ static int nozomi_card_init(struct pci_dev *pdev,
>   		tty_unregister_device(ntty_driver, dc->index_start + i);
>   		tty_port_destroy(&dc->port[i].port);
>   	}
> +	free_irq(pdev->irq, dc);
>   err_free_kfifo:
>   	for (i = 0; i < MAX_PORT; i++)
>   		kfifo_free(&dc->port[i].fifo_ul);
> 

thanks,
-- 
js
suse labs