* [PATCH] nfc: nci: Fix uninit-value in nci_rx_work()
@ 2024-08-03 12:18 zhanghao
2024-08-04 10:57 ` Simon Horman
0 siblings, 1 reply; 4+ messages in thread
From: zhanghao @ 2024-08-03 12:18 UTC (permalink / raw)
To: bongsu.jeon, krzk
Cc: syzbot+3da70a0abd7f5765b6ea, linux-kernel, netdev, zhanghao
Commit e624e6c3e777 ("nfc: Add a virtual nci device driver")
calls alloc_skb() with GFP_KERNEL as the argument flags.The
allocated heap memory was not initialized.This causes KMSAN
to detect an uninitialized value.
Reported-by: syzbot+3da70a0abd7f5765b6ea@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=3da70a0abd7f5765b6ea
Fixes: e624e6c3e777 ("nfc: Add a virtual nci device driver")
Link: https://lore.kernel.org/all/000000000000747dd6061a974686@google.com/T/
Signed-off-by: zhanghao <zhanghao1@kylinos.cn>
---
drivers/nfc/virtual_ncidev.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/nfc/virtual_ncidev.c b/drivers/nfc/virtual_ncidev.c
index 6b89d596ba9a..ae1592db131e 100644
--- a/drivers/nfc/virtual_ncidev.c
+++ b/drivers/nfc/virtual_ncidev.c
@@ -117,7 +117,7 @@ static ssize_t virtual_ncidev_write(struct file *file,
struct virtual_nci_dev *vdev = file->private_data;
struct sk_buff *skb;
- skb = alloc_skb(count, GFP_KERNEL);
+ skb = alloc_skb(count, GFP_KERNEL|__GFP_ZERO);
if (!skb)
return -ENOMEM;
base-commit: 17712b7ea0756799635ba159cc773082230ed028
--
2.39.2
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH] nfc: nci: Fix uninit-value in nci_rx_work()
2024-08-03 12:18 [PATCH] nfc: nci: Fix uninit-value in nci_rx_work() zhanghao
@ 2024-08-04 10:57 ` Simon Horman
2024-08-05 8:20 ` Krzysztof Kozlowski
2024-08-08 0:59 ` zhanghao
0 siblings, 2 replies; 4+ messages in thread
From: Simon Horman @ 2024-08-04 10:57 UTC (permalink / raw)
To: zhanghao
Cc: bongsu.jeon, krzk, syzbot+3da70a0abd7f5765b6ea, linux-kernel,
netdev
On Sat, Aug 03, 2024 at 08:18:17PM +0800, zhanghao wrote:
> Commit e624e6c3e777 ("nfc: Add a virtual nci device driver")
> calls alloc_skb() with GFP_KERNEL as the argument flags.The
> allocated heap memory was not initialized.This causes KMSAN
> to detect an uninitialized value.
>
> Reported-by: syzbot+3da70a0abd7f5765b6ea@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=3da70a0abd7f5765b6ea
Hi,
I wonder if the problem reported above is caused by accessing packet
data which is past the end of what is copied in virtual_ncidev_write().
I.e. count is unusually short and this is not being detected.
> Fixes: e624e6c3e777 ("nfc: Add a virtual nci device driver")
> Link: https://lore.kernel.org/all/000000000000747dd6061a974686@google.com/T/
> Signed-off-by: zhanghao <zhanghao1@kylinos.cn>
> ---
> drivers/nfc/virtual_ncidev.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/nfc/virtual_ncidev.c b/drivers/nfc/virtual_ncidev.c
> index 6b89d596ba9a..ae1592db131e 100644
> --- a/drivers/nfc/virtual_ncidev.c
> +++ b/drivers/nfc/virtual_ncidev.c
> @@ -117,7 +117,7 @@ static ssize_t virtual_ncidev_write(struct file *file,
> struct virtual_nci_dev *vdev = file->private_data;
> struct sk_buff *skb;
>
> - skb = alloc_skb(count, GFP_KERNEL);
> + skb = alloc_skb(count, GFP_KERNEL|__GFP_ZERO);
> if (!skb)
> return -ENOMEM;
I'm not sure this helps wrt initialising the memory as immediately below there
is;
if (copy_from_user(skb_put(skb, count), buf, count)) {
...
Which I assume will initialise count bytes of skb data.
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] nfc: nci: Fix uninit-value in nci_rx_work()
2024-08-04 10:57 ` Simon Horman
@ 2024-08-05 8:20 ` Krzysztof Kozlowski
2024-08-08 0:59 ` zhanghao
1 sibling, 0 replies; 4+ messages in thread
From: Krzysztof Kozlowski @ 2024-08-05 8:20 UTC (permalink / raw)
To: Simon Horman, zhanghao
Cc: bongsu.jeon, syzbot+3da70a0abd7f5765b6ea, linux-kernel, netdev
On 04/08/2024 12:57, Simon Horman wrote:
> On Sat, Aug 03, 2024 at 08:18:17PM +0800, zhanghao wrote:
>> Commit e624e6c3e777 ("nfc: Add a virtual nci device driver")
>> calls alloc_skb() with GFP_KERNEL as the argument flags.The
>> allocated heap memory was not initialized.This causes KMSAN
>> to detect an uninitialized value.
>>
>> Reported-by: syzbot+3da70a0abd7f5765b6ea@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=3da70a0abd7f5765b6ea
>
> Hi,
>
> I wonder if the problem reported above is caused by accessing packet
> data which is past the end of what is copied in virtual_ncidev_write().
> I.e. count is unusually short and this is not being detected.
>
>> Fixes: e624e6c3e777 ("nfc: Add a virtual nci device driver")
>> Link: https://lore.kernel.org/all/000000000000747dd6061a974686@google.com/T/
>> Signed-off-by: zhanghao <zhanghao1@kylinos.cn>
>> ---
>> drivers/nfc/virtual_ncidev.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/nfc/virtual_ncidev.c b/drivers/nfc/virtual_ncidev.c
>> index 6b89d596ba9a..ae1592db131e 100644
>> --- a/drivers/nfc/virtual_ncidev.c
>> +++ b/drivers/nfc/virtual_ncidev.c
>> @@ -117,7 +117,7 @@ static ssize_t virtual_ncidev_write(struct file *file,
>> struct virtual_nci_dev *vdev = file->private_data;
>> struct sk_buff *skb;
>>
>> - skb = alloc_skb(count, GFP_KERNEL);
>> + skb = alloc_skb(count, GFP_KERNEL|__GFP_ZERO);
>> if (!skb)
>> return -ENOMEM;
>
> I'm not sure this helps wrt initialising the memory as immediately below there
> is;
>
> if (copy_from_user(skb_put(skb, count), buf, count)) {
> ...
>
> Which I assume will initialise count bytes of skb data.
Yeah, this looks like hiding the real issue.
Best regards,
Krzysztof
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: Re: [PATCH] nfc: nci: Fix uninit-value in nci_rx_work()
2024-08-04 10:57 ` Simon Horman
2024-08-05 8:20 ` Krzysztof Kozlowski
@ 2024-08-08 0:59 ` zhanghao
1 sibling, 0 replies; 4+ messages in thread
From: zhanghao @ 2024-08-08 0:59 UTC (permalink / raw)
To: horms
Cc: bongsu.jeon, krzk, linux-kernel, netdev,
syzbot+3da70a0abd7f5765b6ea, zhanghao1
On 04/08/2024 12:57, Simon Horman wrote:
> > On Sat, Aug 03, 2024 at 08:18:17PM +0800, zhanghao wrote:
> >> Commit e624e6c3e777 ("nfc: Add a virtual nci device driver")
> >> calls alloc_skb() with GFP_KERNEL as the argument flags.The
> >> allocated heap memory was not initialized.This causes KMSAN
> >> to detect an uninitialized value.
> >>
> >> Reported-by: syzbot+3da70a0abd7f5765b6ea@syzkaller.appspotmail.com
> >> Closes: https://syzkaller.appspot.com/bug?extid=3da70a0abd7f5765b6ea
> >
> > Hi,
> >
> > I wonder if the problem reported above is caused by accessing packet
> > data which is past the end of what is copied in virtual_ncidev_write().
> > I.e. count is unusually short and this is not being detected.
> >
Yes, you're right.I debug the kernel and find that skb->data is " ".The
length is less than 2.Using nci_plen to access skb->data in nci_rx_work()
triggers kmsan to detect access to uninitialized memory.
I wonder if I can use skb->len to determine length in the
nci_rx_work().I tested it and there was no problem.
> >> Fixes: e624e6c3e777 ("nfc: Add a virtual nci device driver")
> >> Link: https://lore.kernel.org/all/000000000000747dd6061a974686@google.com/T/
> >> Signed-off-by: zhanghao <zhanghao1@kylinos.cn>
> >> ---
> >> drivers/nfc/virtual_ncidev.c | 2 +-
> >> 1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/nfc/virtual_ncidev.c b/drivers/nfc/virtual_ncidev.c
> >> index 6b89d596ba9a..ae1592db131e 100644
> >> --- a/drivers/nfc/virtual_ncidev.c
> >> +++ b/drivers/nfc/virtual_ncidev.c
> >> @@ -117,7 +117,7 @@ static ssize_t virtual_ncidev_write(struct file *file,
> >> struct virtual_nci_dev *vdev = file->private_data;
> >> struct sk_buff *skb;
> >>
> >> - skb = alloc_skb(count, GFP_KERNEL);
> >> + skb = alloc_skb(count, GFP_KERNEL|__GFP_ZERO);
> >> if (!skb)
> >> return -ENOMEM;
> >
> > I'm not sure this helps wrt initialising the memory as immediately below there
> > is;
> >
> > if (copy_from_user(skb_put(skb, count), buf, count)) {
> > ...
> >
> > Which I assume will initialise count bytes of skb data.
>
> Yeah, this looks like hiding the real issue.
>
> Best regards,
> Krzysztof
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2024-08-08 1:00 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-08-03 12:18 [PATCH] nfc: nci: Fix uninit-value in nci_rx_work() zhanghao
2024-08-04 10:57 ` Simon Horman
2024-08-05 8:20 ` Krzysztof Kozlowski
2024-08-08 0:59 ` zhanghao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).