CVE-2026-98158

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ppp_async: drop the errored frame instead of resetting its headroom<br /> <br /> ppp_receive_nonmp_frame() prepends a two-byte direction tag before running<br /> the pass/active BPF filters:<br /> <br /> *(__be16 *)skb_push(skb, 2) = htons(PPP_FILTER_INBOUND_TAG);<br /> <br /> Nothing on the receive path guarantees those two bytes of headroom. The<br /> frame-error path in ppp_async&amp;#39;s process_input_packet() resets a reused skb&amp;#39;s<br /> headroom to zero while claiming to restore it to a freshly allocated state -<br /> but a fresh skb from dev_alloc_skb() carries NET_SKB_PAD:<br /> <br /> err:<br /> if (skb) {<br /> /* make skb appear as freshly allocated */<br /> skb_trim(skb, 0);<br /> skb_reserve(skb, - skb_headroom(skb));<br /> }<br /> <br /> ap-&gt;rpkt still points at that skb, so the next frame is reassembled into it<br /> with no headroom at all. A peer that sends a bad-FCS frame followed by one<br /> beginning ff 03 then leaves a single byte of headroom by the time the filter<br /> tag is pushed, which lands one byte below skb-&gt;head:<br /> <br /> skbuff: skb_under_panic: len:49 put:2 head:ffff888003c10000<br /> data:ffff888003c0ffff tail:0x30 end:0x640 dev:<br /> kernel BUG at net/core/skbuff.c:214!<br /> RIP: 0010:skb_panic+0x13e/0x230<br /> Call Trace:<br /> skb_push+0xbd/0x100<br /> ppp_receive_nonmp_frame+0x48a/0x1d10<br /> ppp_input+0x4e9/0x2f80<br /> ppp_async_process+0x2a/0xe0<br /> tasklet_action_common+0x20f/0x8a0<br /> handle_softirqs+0x18e/0x590<br /> Kernel panic - not syncing: Fatal exception in interrupt<br /> <br /> Zeroing the headroom violates the NET_SKB_PAD guarantee that dev_alloc_skb()<br /> gives the rest of the receive path. Besides the filter panic above, when CCP<br /> compression is enabled ppp_decompress_frame() hands skb-&gt;data - 2 to<br /> -&gt;decompress()/-&gt;incomp(), which then reads out of bounds before skb-&gt;head<br /> for the same reason.<br /> <br /> Rather than restore the headroom, drop the errored frame - as ppp_synctty<br /> already does on its error path - and clear ap-&gt;rpkt so the next frame is<br /> reassembled into a fresh skb with proper headroom. This is simpler and fixes<br /> both the filter under-panic and the CCP out-of-bounds read.<br /> <br /> The original V1 of this patch made room in ppp_receive_nonmp_frame() with<br /> skb_cow_head(); Eric pointed out that fixing the root cause in the transport<br /> is the right approach.<br /> <br /> Found by fuzzing the PPP receive path with a mutating peer on a pty; it is an<br /> interesting (remote) DoS: root configures PPP, the peer supplies two crashing<br /> frames. The reproducer (repro-ppp-skb.c, unchanged from v1) panics in about a<br /> second, and returns cleanly with this applied.