Unaligned 32-bit reads following casts on dns_socket_dispatcher::local_addr (Cortex-M0)

[IWillDonut]

Describe the bug

dns_socket_dispatcher::local_addr is not aligned for the 32-bit read of local_addr.sin_addr.s_addr in net_context_bind() leading to a hard-fault on a Cortex-M0. I've noticed this for the IPv4 path, but I suspect this impacts other paths in net_context_bind() as well.

I've been able to get around with either of the two workarounds:

(a) Forcing the alignment of dns_socket_dispatcher::local_addr to the maximum alignment of the various address types resolves the issue.

(b) Updating net_context_bind() to use the unaligned helpers from net_ip.h.

To Reproduce

Apologies, I'm fairly new and not sure if there's a way to reproduce this in QEMU, so the following assumes access to the physical hardware.

Add the board files from Additional context below to samples/net/dhcpv4_client/boards.

west build -b rpi_pico samples/net/dhcpv4_client --pristine

west debug

Set a breakpoint for the faulting instruction:

b zephyr/subsys/net/ip/net_context.c:868

The following ldr tries to perform an unaligned read from address 0x200015E2 leading to the hard-fault shown below:

0x1000c1a0 <net_context_bind+92>        ldr     r2, [r5, #4]

# r5             0x200015de

Expected behavior

No hard faults in DNS path.

Impact

Unable to use DNS (and DHCP) on my board.

Logs and console output

[00:00:03.348,000] <err> os: ***** HARD FAULT *****
[00:00:03.348,000] <err> os: r0/a1:  0x00000000  r1/a2:  0x20001eee  r2/a3:  0x00000000
[00:00:03.348,000] <err> os: r3/a4:  0xe0000000 r12/ip:  0x2b4be420 r14/lr:  0x100169e3
[00:00:03.348,000] <err> os:  xpsr:  0x61000000
[00:00:03.348,000] <err> os: Faulting instruction address (r15/pc): 0x10016a1c
[00:00:03.348,000] <err> os: >>> ZEPHYR FATAL ERROR 0: CPU exception on CPU 0
[00:00:03.348,000] <err> os: Current thread: 0x20001ca8 (rx_q[0])
[00:00:03.396,000] <err> os: Halting system

#0  net_context_bind (context=context@entry=0x20001ff8 <contexts>, addr=0x200015de <dns_default_ctx+30>, addrlen=<optimized out>) at zephyr/subsys/net/ip/net_context.c:868
#1  0x10005cfc in zsock_bind_ctx (ctx=0x20001ff8 <contexts>, addr=<optimized out>, addrlen=<optimized out>) at zephyr/subsys/net/lib/sockets/sockets.c:478
#2  0x100174ce in sock_bind_vmeth (obj=<optimized out>, addr=<optimized out>, addrlen=<optimized out>) at zephyr/subsys/net/lib/sockets/sockets.c:3852
#3  0x10005d72 in z_impl_zsock_bind (sock=<optimized out>, addr=addr@entry=0x200015de <dns_default_ctx+30>, addrlen=8) at zephyr/subsys/net/lib/sockets/sockets.c:497
#4  0x1000ef6e in zsock_bind (addrlen=<optimized out>, addr=0x200015de <dns_default_ctx+30>, sock=<optimized out>) at zephyr/build/zephyr/include/generated/zephyr/syscalls/socket.h:158
#5  dns_dispatcher_register (ctx=ctx@entry=0x200015d0 <dns_default_ctx+16>) at zephyr/subsys/net/lib/dns/dispatcher.c:247
#6  0x1000f1dc in register_dispatcher (svc=0x1001bc10 <resolve_svc>, addr6=0x0, addr4=<optimized out>, local=0x20003d68 <rx_stack+1064>, server=0x200015c0 <dns_default_ctx>, ctx=0x200015c0 <dns_default_ctx>) at zephyr/subsys/net/lib/dns/resolve.c:317
#7  dns_resolve_init_locked (ctx=ctx@entry=0x200015c0 <dns_default_ctx>, servers=servers@entry=0x0, servers_sa=servers_sa@entry=0x20003de8 <rx_stack+1192>) at zephyr/subsys/net/lib/dns/resolve.c:475
#8  0x1001a1de in dns_resolve_reconfigure (ctx=0x200015c0 <dns_default_ctx>, servers=servers@entry=0x0, servers_sa=0x20003de8 <rx_stack+1192>, servers_sa@entry=0x20003e28 <rx_stack+1256>) at zephyr/subsys/net/lib/dns/resolve.c:1617
#9  0x1000e778 in dhcpv4_parse_options (msg_type=<synthetic pointer>, iface=0x200008d8 <__net_if_dts_ord_74_0>, pkt=0x200081ac <_k_mem_slab_buf_rx_pkts+60>) at zephyr/subsys/net/lib/dhcpv4/dhcpv4.c:1110
#10 net_dhcpv4_input (conn=<optimized out>, pkt=0x200081ac <_k_mem_slab_buf_rx_pkts+60>, ip_hdr=0x200008d8 <__net_if_dts_ord_74_0>, proto_hdr=<optimized out>, user_data=0x0) at zephyr/subsys/net/lib/dhcpv4/dhcpv4.c:1534
#11 0x1000d60a in net_conn_input (pkt=pkt@entry=0x200081ac <_k_mem_slab_buf_rx_pkts+60>, ip_hdr=ip_hdr@entry=0x20003e94 <rx_stack+1364>, proto=<optimized out>, proto_hdr=proto_hdr@entry=0x20003e90 <rx_stack+1360>) at zephyr/subsys/net/ip/connection.c:919
#12 0x10019aba in net_ipv4_input (pkt=pkt@entry=0x200081ac <_k_mem_slab_buf_rx_pkts+60>, is_loopback=is_loopback@entry=false) at zephyr/subsys/net/ip/ipv4.c:437
#13 0x10017c6a in process_data (is_loopback=<optimized out>, pkt=0x200081ac <_k_mem_slab_buf_rx_pkts+60>) at zephyr/subsys/net/ip/net_core.c:138
#14 processing_data (pkt=0x200081ac <_k_mem_slab_buf_rx_pkts+60>, is_loopback=<optimized out>) at zephyr/subsys/net/ip/net_core.c:156
#15 0x10017e6c in net_rx (iface=<optimized out>, pkt=<optimized out>) at zephyr/subsys/net/ip/net_core.c:452
#16 net_process_rx_packet (pkt=<optimized out>) at zephyr/subsys/net/ip/net_core.c:464
#17 0x10019226 in tc_rx_handler (p1=0x200014b8 <rx_classes>, p2=<optimized out>, p3=<optimized out>) at zephyr/subsys/net/ip/net_tc.c:257
#18 0x10001930 in z_thread_entry (entry=0x1001920b <tc_rx_handler>, p1=0x200014b8 <rx_classes>, p2=0x0, p3=0x0) at zephyr/lib/os/thread_entry.c:48

Environment:

• OS: Windows
• Toolchain: Zephyr SDK
• Commit SHA: 15aab68
• Board: rpi-pico
• Ethernet Controller: enc28j60

Additional context

boards/rpi_pico.conf

CONFIG_SPI=y
CONFIG_ETH_DRIVER=y
CONFIG_ETH_ENC28J60=y
CONFIG_NET_L2_ETHERNET=y

boards/rpi_pico.overlay

&spi0 {
	status = "okay";
	pinctrl-names = "default";
	cs-gpios = <&gpio0 5 GPIO_ACTIVE_LOW>;
	clock-frequency = <4000000>;

	eth_enc28j60: eth_enc28j60@0 {
		compatible = "microchip,enc28j60";
		reg = <0x0>;
		spi-max-frequency = <2000000>;
		label = "ETH_0";
		local-mac-address = [b8 27 eb d2 ec 5c];
		int-gpios = <&gpio0 20 GPIO_ACTIVE_LOW>;
	};
};

&pinctrl {
	spi0_default: spi0_default {
		group1 {
			pinmux = <SPI0_CSN_P5>, <SPI0_SCK_P2>, <SPI0_TX_P3>;
		};
		group2 {
			pinmux = <SPI0_RX_P4>;
			input-enable;
		};
	};
};

[github-actions]

Hi @IWillDonut! We appreciate you submitting your first issue for our open-source project. 🌟

Even though I'm a bot, I can assure you that the whole community is genuinely grateful for your time and effort. 🤖💙

[jukkar]

(b) Updating net_context_bind() to use the unaligned helpers from net_ip.h.

This sounds like a proper way, could you send a PR for it?

[jukkar]

There was a similar report discussed in #77184. Would either @IWillDonut or @hulryung be able to send a PR to fix this?

[IWillDonut]

Sure thing. I can take a look at getting a PR up for this later this week.