In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
net: stmmac: restore NET_IP_ALIGN in the RX DMA offset<\/p>\n
Since the RX path was converted to zero-copy, the page pool page is handed
\nto the stack directly as the skb head, and the offset the DMA engine writes
\nat is what determines the alignment of the packet headers.<\/p>\n
Before the conversion the payload was copied into an skb obtained from
\nnapi_alloc_skb(), which reserves NET_SKB_PAD + NET_IP_ALIGN. The
\nconversion moved the headroom into stmmac_rx_offset() but did not carry
\nover NET_IP_ALIGN, so on architectures where NET_IP_ALIGN is 2 the IP
\nheader now lands misaligned:<\/p>\n
64 (NET_SKB_PAD) + 14 (ethernet) + 20 (IP) = 98<\/p>\n
Same for the XDP branch:<\/p>\n
256 (XDP_PACKET_HEADROOM) + 14 (ethernet) + 20 (IP) = 290<\/p>\n
On ARM32 this is fatal, because ldm and ldrd trap on unaligned addresses
\neven when CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS is set.<\/p>\n
Any received echo request panics the machine, e.g:<\/p>\n
Unhandled fault: alignment exception (0x001) at 0x81873062
\n Internal error: : 1 [#1] SMP ARM
\n Hardware name: Altera SOCFPGA Arria10
\n PC is at icmp_echo+0x38\/0xa8
\n LR is at icmp_rcv+0x22c\/0x370
\n Call trace:
\n icmp_echo from icmp_rcv+0x22c\/0x370
\n icmp_rcv from ip_protocol_deliver_rcu+0x2c\/0x224
\n ip_protocol_deliver_rcu from ip_local_deliver+0xc8\/0x1a0
\n ip_local_deliver from ip_sublist_rcv_finish+0x3c\/0x50
\n ip_sublist_rcv_finish from ip_list_rcv_finish+0x110\/0x118
\n ip_list_rcv_finish from ip_list_rcv+0xc8\/0xdc
\n ip_list_rcv from __netif_receive_skb_list_core+0x170\/0x1c0
\n …
\n napi_complete_done from stmmac_napi_poll_rx+0xcb0\/0x1030
\n Code: e24dd068 e59020a0 e28dc010 e0822001 (e8920003)
\n Kernel panic – not syncing: Fatal exception in interrupt<\/p>\n
The faulting instruction is the ldm of *icmp_hdr(skb) in icmp_echo().<\/p>\n
Fix by adding NET_IP_ALIGN back to the RX offset, which restores the
\nalignment the stack used to get.<\/p>\n
Note that commit a955318fe67e (“stmmac: align RX buffers”) made a similar
\nchange in 2021 and was reverted by commit 12d125b4574b (“stmmac: Revert
\n“stmmac: align RX buffers””) because it caused packet corruption. That
\npatch raised the offset from 0 without adjusting the buffer size
\naccounting, so the DMA engine could arguably write past the end of the RX
\nbuffers, though this was never root caused.
\nCommit df542f669307 (“net: stmmac: Switch to zero-copy in non-XDP RX
\npath”) since derives the page pool allocation from stmmac_rx_offset(), so
\nthe extra bytes are accounted for.<\/p>\n","protected":false},"excerpt":{"rendered":"
In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
net: stmmac: restore NET_IP_ALIGN in the RX DMA offset<\/p>\n
Since the RX path was converted to zero-copy, the page pool page is handed<\/p>\n","protected":false},"template":"","meta":[],"categories":[],"tags":[],"class_list":["post-25468","vuln","type-vuln","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln\/25468","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln"}],"about":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/types\/vuln"}],"wp:attachment":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=25468"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=25468"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=25468"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}