On Sun, Sep 13, 2026 at 4:49 PM Eric Dumazet [off-list ref] wrote:
On Sun, Sep 13, 2026 at 4:32 PM [off-list ref] wrote:
quoted
Thank you for your contribution! Sashiko AI review found 1 potential
issue(s) to consider:
Critical: 0 · High: 1 · Medium: 0 · Low: 0 (1)
- [High] pppoatm_send() does not make the frame linear, although the
commit subject claims to "ensure a writable skb header and linear…
Pre-existing issues:
- [Low] net/atm/pppoatm.c never initialises pvcc->chan.hdrlen although
the channel requires LLC_LEN (4) bytes of headroom for e_llc…
--
pppoatm: ensure a writable skb header and linear data
pppoatm_send() now uses skb_cow_head() to get a writable header with room
for the LLC header, adds a pskb_may_pull(skb, 1) before looking at
skb->data[0], and defers the SC_COMP_PROT pull plus the LLC push until
after pppoatm_may_send() has succeeded, dropping the old "undo" heuristic.
A couple of questions about the linearity claim and about the channel
headroom hint.
quoted
diff --git a/net/atm/pppoatm.c b/net/atm/pppoatm.c
index 6da52d12df68e..5214786e61d11 100644
--- a/net/atm/pppoatm.c
+++ b/net/atm/pppoatm.c
@@ -292,10 +292,13 @@ static int pppoatm_send(struct ppp_channel *chan, struct sk_buff *skb)
struct atm_vcc *vcc;
int ret;
+ if (!pskb_may_pull(skb, 1)) {
+ kfree_skb(skb);
+ return DROP_PACKET;
+ }
+
[Severity: High]
The subject says "linear data" and the changelog says this pull is there
"to prevent out-of-bounds reads on zero-length or non-linear frames
(e.g. from bridging)", but does this pull actually make the frame linear?
pskb_may_pull(skb, 1) only guarantees the single byte read as
skb->data[0]; skb_cow_head() -> __skb_cow() -> pskb_expand_head()
unshares and grows the head but leaves the paged fragments in place.
Linear' as in : make sure the first byte is in skb->head, before
reading skb->data[0]
I guess I need to invent a new term : 1-byte-linear
Perhaps this will please/silence our AI friends.