[PATCH] uio: Fix an Oops on load

[PATCH] uio: Fix an Oops on load

[Dan Carpenter]

I was trying to solve a double free but I introduced a more serious
NULL dereference bug.  The problem is that if there is an IRQ which
triggers immediately, then we need "info->uio_dev" but it's not set yet.

This patch puts the original initialization back to how it was and just
sets info->uio_dev to NULL on the error path so it should solve both
the Oops and the double free.

Fixes: f019f07ecf6a ("uio: potential double frees if __uio_register_device() fails")
Reported-by: Mathias Thore <Mathias.Thore@infinera.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
 drivers/uio/uio.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/uio/uio.c b/drivers/uio/uio.c
index 85644669fbe7..0a357db4b31b 100644
--- a/drivers/uio/uio.c
+++ b/drivers/uio/uio.c
@@ -961,6 +961,8 @@ int __uio_register_device(struct module *owner,
 	if (ret)
 		goto err_uio_dev_add_attributes;
 
+	info->uio_dev = idev;
+
 	if (info->irq && (info->irq != UIO_IRQ_CUSTOM)) {
 		/*
 		 * Note that we deliberately don't use devm_request_irq
@@ -972,11 +974,12 @@ int __uio_register_device(struct module *owner,
 		 */
 		ret = request_irq(info->irq, uio_interrupt,
 				  info->irq_flags, info->name, idev);
-		if (ret)
+		if (ret) {
+			info->uio_dev = NULL;
 			goto err_request_irq;
+		}
 	}
 
-	info->uio_dev = idev;
 	return 0;
 
 err_request_irq:
-- 
2.11.0

RE: [PATCH] uio: Fix an Oops on load

[Mathias Thore]

Thanks for the quick fix. I've tested it and it seems OK.

Tested-by: Mathias Thore <Mathias.Thore@infinera.com>


-----Original Message-----
From: Dan Carpenter [mailto:dan.carpenter@oracle.com] 
Sent: Friday, 26 October 2018 09:20
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>; Mathias Thore <Mathias.Thore@infinera.com>
Cc: linux-kernel@vger.kernel.org
Subject: [PATCH] uio: Fix an Oops on load

CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.


I was trying to solve a double free but I introduced a more serious NULL dereference bug.  The problem is that if there is an IRQ which triggers immediately, then we need "info->uio_dev" but it's not set yet.

This patch puts the original initialization back to how it was and just sets info->uio_dev to NULL on the error path so it should solve both the Oops and the double free.

Fixes: f019f07ecf6a ("uio: potential double frees if __uio_register_device() fails")
Reported-by: Mathias Thore <Mathias.Thore@infinera.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
 drivers/uio/uio.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/uio/uio.c b/drivers/uio/uio.c index 85644669fbe7..0a357db4b31b 100644
--- a/drivers/uio/uio.c
+++ b/drivers/uio/uio.c
@@ -961,6 +961,8 @@ int __uio_register_device(struct module *owner,
        if (ret)
                goto err_uio_dev_add_attributes;

+       info->uio_dev = idev;
+
        if (info->irq && (info->irq != UIO_IRQ_CUSTOM)) {
                /*
                 * Note that we deliberately don't use devm_request_irq @@ -972,11 +974,12 @@ int __uio_register_device(struct module *owner,
                 */
                ret = request_irq(info->irq, uio_interrupt,
                                  info->irq_flags, info->name, idev);
-               if (ret)
+               if (ret) {
+                       info->uio_dev = NULL;
                        goto err_request_irq;
+               }
        }

-       info->uio_dev = idev;
        return 0;

 err_request_irq:
--
2.11.0