Re: [PATCH V4 5/6] net: maclorawan: Implement maclorawan class module

[Jian-Hong Pan <starnight@xxxxxxxxxxxx>]

I made a fake skb and passed it to lrw_parse_frame() function for
testing.  I use print_hex_dump() function to show the skb's content.
Here is the original content in the skb->data and the length is 20 bytes.

[   33.732033] 00000000: 40 04 03 02 01 00 00 00 00 27 76 d3 2d 1b 79
a0  @........'v.-.y.
[   33.732065] 00000010: 18 38 fb a6                                      .8..

Byte 0: MHDR field, value is 0x40.
Byte 1 ~ 4: DevAddr field, value is 0x04 0x03 0x02 0x01.
Byte 5: FCtrl field, value is 0x00.
Byte 6 ~ 7: FCnt field, value is 0x00 0x00.
Byte 8: FPort field, value is 0x00.
Byte 9 ~ 15: Encrypted payload
Byte 16 ~ 19: MIC field value is 0x18 0x38 0xfb 0xa6.

> > +void
> > +lrw_parse_frame(struct lrw_session *ss, struct sk_buff *skb)
> > +{
> > +     struct lrw_fhdr *fhdr = &ss->rx_fhdr;
> > +     __le16 *p_fcnt;
> > +
> > +     pr_debug("%s: %s\n", LORAWAN_MODULE_NAME, __func__);
> > +
> > +     /* Get message type */
> > +     fhdr->mtype = skb->data[0];
> > +     skb_pull(skb, LRW_MHDR_LEN);

print_hex_dump skb here:
[   33.732202] 00000000: 04 03 02 01 00 00 00 00 27 76 d3 2d 1b 79 a0
18  ........'v.-.y..
[   33.732204] 00000010: 38 fb a6

> This does not seem robust. There is no point at which you actually check
> the message size is valid etc

Thanks!  It is a potential bug.
It should check skb->len >= length of MHDR + DevAddr + FCtrl + FCnt + MIC.
These are required fields for (Un)confirmed Data Up/Down messages.


print_hex_dump skb here:
[   33.732211] 00000000: 00 27 76 d3 2d 1b 79 a0 18 38 fb a6
   .'v.-.y..8..

> > +     fhdr->fopts_len = fhdr->fctrl & 0xF;
> > +     if (fhdr->fopts_len > 0) {
> > +             memcpy(fhdr->fopts, skb->data, fhdr->fopts_len);
> > +             skb_pull(skb, fhdr->fopts_len);
> > +     }

print_hex_dump skb here:
[   33.732213] 00000000: 00 27 76 d3 2d 1b 79 a0 18 38 fb a6
   .'v.-.y..8..

> In fact you appear to copy random kernel memory into a buffer

It copied fhdr->fopts_len bytes from skb->data to fhdr->fopts if
fhdr->fopts_len > 0.
https://www.kernel.org/doc/html/latest/core-api/kernel-api.html?highlight=memcpy#c.memcpy

> > +
> > +     /* TODO: Parse frame options */
> > +
> > +     /* Remove message integrity code */
> > +     skb_trim(skb, skb->len - LRW_MIC_LEN);

print_hex_dump skb here:
[   33.732216] 00000000: 00 27 76 d3 2d 1b 79 a0
   .'v.-.y.

> and then try and trim the buffer to a negative size ?

It removed 4 tail bytes (MIC).  (skb->len - LRW_MIC_LEN) is the final
new length as skb_trim()'s 2nd argument len.
https://www.kernel.org/doc/html/latest/networking/kapi.html?highlight=skb_trim#c.skb_trim

I found another bug which did not initialize rx_skb_list.  So,
lrw_parse_frame() may be passed a mystery skb.

Please keep reviewing.  That is appreciated.

Thank you,
Jian-Hong Pan