[PATCH v2] nfc: pn533: fix OOB read in pn533_acr122_is_rx_frame_valid()

[PATCH v2] nfc: pn533: fix OOB read in pn533_acr122_is_rx_frame_valid()

[Deepanshu Kartikey]

frame->ccid.datalen is read directly from the USB response frame
and used, unchecked, as an index into frame->data[]. A malicious or
malfunctioning device can set this field to an arbitrary value,
causing the driver to read far outside the received buffer.

Bound ccid.datalen against the maximum possible ACR122 frame size
before using it. This replaces the existing datalen == 0 check,
since datalen < 2 already covers that case and additionally
rejects datalen == 1, which would still underflow the
"datalen - 2" offset used below.

Fixes: 9815c7cf22da ("NFC: pn533: Separate physical layer from the core implementation")
Reported-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=1853daab1a47603d4678
Tested-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
Assisted-by: Claude (Anthropic)
Link: https://lore.kernel.org/all/20260906054336.20187-1-kartikey406@gmail.com/T/ [v1]
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>

---
Changes in v2:
- Replace the existing "if (!frame->ccid.datalen)" check instead of
  adding the new bounds check alongside it, since the new check
  already subsumes it (per Johan Hovold).
---
 drivers/nfc/pn533/usb.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/nfc/pn533/usb.c b/drivers/nfc/pn533/usb.c
index efb07f944fce..9471c34b4d4b 100644
--- a/drivers/nfc/pn533/usb.c
+++ b/drivers/nfc/pn533/usb.c
@@ -319,7 +319,9 @@ static bool pn533_acr122_is_rx_frame_valid(void *_frame, struct pn533 *dev)
 	if (frame->ccid.type != 0x83)
 		return false;
 
-	if (!frame->ccid.datalen)
+	if (frame->ccid.datalen < 2 ||
+	    frame->ccid.datalen > PN533_ACR122_FRAME_MAX_PAYLOAD_LEN +
+	    PN533_ACR122_RX_FRAME_TAIL_LEN)
 		return false;
 
 	if (frame->data[frame->ccid.datalen - 2] == 0x63)
-- 
2.43.0

Re: [PATCH v2] nfc: pn533: fix OOB read in pn533_acr122_is_rx_frame_valid()

[David Heidelberg]

On 12/09/2026 06:03, Deepanshu Kartikey wrote:
> frame->ccid.datalen is read directly from the USB response frame
> and used, unchecked, as an index into frame->data[]. A malicious or
> malfunctioning device can set this field to an arbitrary value,
> causing the driver to read far outside the received buffer.
> 
> Bound ccid.datalen against the maximum possible ACR122 frame size
> before using it. This replaces the existing datalen == 0 check,
> since datalen < 2 already covers that case and additionally
> rejects datalen == 1, which would still underflow the
> "datalen - 2" offset used below.
> 
> Fixes: 9815c7cf22da ("NFC: pn533: Separate physical layer from the core implementation")
> Reported-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=1853daab1a47603d4678
> Tested-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
> Assisted-by: Claude (Anthropic)

Hello Deepanshu,

thank you for the patch, could you please correct the Assisted-by tag?

See: https://docs.kernel.org/process/coding-assistants.html

Thanks
David

> Link: https://lore.kernel.org/all/20260906054336.20187-1-kartikey406@gmail.com/T/ [v1]
> Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
> 
> ---
> Changes in v2:
> - Replace the existing "if (!frame->ccid.datalen)" check instead of
>    adding the new bounds check alongside it, since the new check
>    already subsumes it (per Johan Hovold).
> ---
>   drivers/nfc/pn533/usb.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)

Re: [PATCH v2] nfc: pn533: fix OOB read in pn533_acr122_is_rx_frame_valid()

[Deepanshu Kartikey]

On Mon, Sep 21, 2026 at 4:41 AM David Heidelberg <david@ixit.cz> wrote:
>
> On 12/09/2026 06:03, Deepanshu Kartikey wrote:
> > frame->ccid.datalen is read directly from the USB response frame
> > and used, unchecked, as an index into frame->data[]. A malicious or
> > malfunctioning device can set this field to an arbitrary value,
> > causing the driver to read far outside the received buffer.
> >
> > Bound ccid.datalen against the maximum possible ACR122 frame size
> > before using it. This replaces the existing datalen == 0 check,
> > since datalen < 2 already covers that case and additionally
> > rejects datalen == 1, which would still underflow the
> > "datalen - 2" offset used below.
> >
> > Fixes: 9815c7cf22da ("NFC: pn533: Separate physical layer from the core implementation")
> > Reported-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
> > Closes: https://syzkaller.appspot.com/bug?extid=1853daab1a47603d4678
> > Tested-by: syzbot+1853daab1a47603d4678@syzkaller.appspotmail.com
> > Assisted-by: Claude (Anthropic)
>
> Hello Deepanshu,
>
> thank you for the patch, could you please correct the Assisted-by tag?
>
> See: https://docs.kernel.org/process/coding-assistants.html
>
> Thanks
> David
>

Sure, I will send patch v3 shortly.

Thanks