2020-08-30 08:24:05

by Jia-Ju Bai

[permalink] [raw]
Subject: [PATCH] media: pci: ttpci: av7110: avoid compiler optimization of reading data[0] in debiirq()

In debiirq(), data_0 stores the value of data[0], but it can be dropped
by compiler optimization. Thus, data[0] is read through READ_ONCE().

Fixes: 6499a0db9b0f ("media: pci: ttpci: av7110: fix possible buffer overflow caused by bad DMA value in debiirq()")
Reported-by: Pavel Machek <[email protected]>
Signed-off-by: Jia-Ju Bai <[email protected]>
---
drivers/media/pci/ttpci/av7110.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/pci/ttpci/av7110.c b/drivers/media/pci/ttpci/av7110.c
index bf36b1e22b63..f7d098d5b198 100644
--- a/drivers/media/pci/ttpci/av7110.c
+++ b/drivers/media/pci/ttpci/av7110.c
@@ -406,7 +406,7 @@ static void debiirq(unsigned long cookie)
case DATA_CI_GET:
{
u8 *data = av7110->debi_virt;
- u8 data_0 = data[0];
+ u8 data_0 = READ_ONCE(data[0]);

if (data_0 < 2 && data[2] == 0xff) {
int flags = 0;
--
2.17.1


2020-08-30 08:35:18

by Sean Young

[permalink] [raw]
Subject: Re: [PATCH] media: pci: ttpci: av7110: avoid compiler optimization of reading data[0] in debiirq()

On Sun, Aug 30, 2020 at 04:20:42PM +0800, Jia-Ju Bai wrote:
> In debiirq(), data_0 stores the value of data[0], but it can be dropped
> by compiler optimization. Thus, data[0] is read through READ_ONCE().
>
> Fixes: 6499a0db9b0f ("media: pci: ttpci: av7110: fix possible buffer overflow caused by bad DMA value in debiirq()")
> Reported-by: Pavel Machek <[email protected]>

Pavel reported that your patch was garbage, if you are trying to defend
against a malicious pci device. READ_ONCE() will not help here.

Sean

> Signed-off-by: Jia-Ju Bai <[email protected]>
> ---
> drivers/media/pci/ttpci/av7110.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/media/pci/ttpci/av7110.c b/drivers/media/pci/ttpci/av7110.c
> index bf36b1e22b63..f7d098d5b198 100644
> --- a/drivers/media/pci/ttpci/av7110.c
> +++ b/drivers/media/pci/ttpci/av7110.c
> @@ -406,7 +406,7 @@ static void debiirq(unsigned long cookie)
> case DATA_CI_GET:
> {
> u8 *data = av7110->debi_virt;
> - u8 data_0 = data[0];
> + u8 data_0 = READ_ONCE(data[0]);
>
> if (data_0 < 2 && data[2] == 0xff) {
> int flags = 0;
> --
> 2.17.1

2020-09-09 08:56:29

by Pavel Machek

[permalink] [raw]
Subject: Re: [PATCH] media: pci: ttpci: av7110: avoid compiler optimization of reading data[0] in debiirq()

On Sun 2020-08-30 09:30:36, Sean Young wrote:
> On Sun, Aug 30, 2020 at 04:20:42PM +0800, Jia-Ju Bai wrote:
> > In debiirq(), data_0 stores the value of data[0], but it can be dropped
> > by compiler optimization. Thus, data[0] is read through READ_ONCE().
> >
> > Fixes: 6499a0db9b0f ("media: pci: ttpci: av7110: fix possible buffer overflow caused by bad DMA value in debiirq()")
> > Reported-by: Pavel Machek <[email protected]>
>
> Pavel reported that your patch was garbage, if you are trying to defend
> against a malicious pci device. READ_ONCE() will not help here.

I would not use exactly those words, but agreed; we should have some
explanation that it is feasible to protect against malicious av7110
device, first.

Best regards,
Pavel
--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html


Attachments:
(No filename) (908.00 B)
signature.asc (188.00 B)
Digital signature
Download all attachments