[PATCH] tty: nozomi: Fix the error handling path of 'nozomi_card_init()'

[PATCH] tty: nozomi: Fix the error handling path of 'nozomi_card_init()'

[Christophe JAILLET]

The error handling path is broken and we may un-register things that have
never been registered.

Update the loops index accordingly.

Fixes: 9842c38e9176 ("kfifo: fix warn_unused_result")
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
PORT_MDM is 0, so it does not really matter, but I wonder why one of the
for loop starts at PORT_MDM, while the other one explicitly starts at 0.

We have the same in the remove function, where 0 is used in both cases.
---
 drivers/tty/nozomi.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/nozomi.c b/drivers/tty/nozomi.c
index b85d4beabc1f..0c80f25c8c3d 100644
--- a/drivers/tty/nozomi.c
+++ b/drivers/tty/nozomi.c
@@ -1378,7 +1378,7 @@ static int nozomi_card_init(struct pci_dev *pdev,
 			NOZOMI_NAME, dc);
 	if (unlikely(ret)) {
 		dev_err(&pdev->dev, "can't request irq %d\n", pdev->irq);
-		goto err_free_kfifo;
+		goto err_free_all_kfifo;
 	}
 
 	DBG1("base_addr: %p", dc->base_addr);
@@ -1416,13 +1416,15 @@ static int nozomi_card_init(struct pci_dev *pdev,
 	return 0;
 
 err_free_tty:
-	for (i = 0; i < MAX_PORT; ++i) {
+	for (i--; i >= 0; i--) {
 		tty_unregister_device(ntty_driver, dc->index_start + i);
 		tty_port_destroy(&dc->port[i].port);
 	}
 	free_irq(pdev->irq, dc);
+err_free_all_kfifo:
+	i = MAX_PORT;
 err_free_kfifo:
-	for (i = 0; i < MAX_PORT; i++)
+	for (i--; i >= PORT_MDM; i--)
 		kfifo_free(&dc->port[i].fifo_ul);
 err_free_sbuf:
 	kfree(dc->send_buf);
-- 
2.30.2