Re: Kernel WARNING in Xenomai3 with RTnet

On 08.04.26 22:53, Hannes Diethelm wrote:
> Configuration:
> 
> Debian Trixie
> xenomai3:master
> linux-dovetail:v6.12.67-dovetail1
> 
> I have the following warning as soon as I enable the network interface:
> Apr 08 01:52:12 trixie-box kernel: ------------[ cut here ]------------
> Apr 08 01:52:12 trixie-box kernel: memcpy: detected field-spanning write
> (size 80) of single field "call->priv_data" at drivers/xenomai/net/
> stack/rtnet_rtpc.c:93 (size 0)
> Apr 08 01:52:12 trixie-box kernel: WARNING: CPU: 13 PID: 2480 at
> drivers/xenomai/net/stack/rtnet_rtpc.c:93
> rtnet_rtpc_dispatch_call+0x2df/0x3d0 [rtnet]
> Apr 08 01:52:12 trixie-box kernel: Modules linked in: rtudp rtipv4
> rt_loopback joydev uinput qrtr rfkill binfmt_misc intel_rapl_msr
> intel_rapl_common ccp kvm snd_hda_codec_generic snd_hda_intel ir>
> Apr 08 01:52:12 trixie-box kernel: CPU: 13 UID: 0 PID: 2480 Comm: rtping
> Not tainted 6.12.67-xenomai3-73ad3b #3
> Apr 08 01:52:12 trixie-box kernel: Hardware name: QEMU Standard PC (Q35
> + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> Apr 08 01:52:12 trixie-box kernel: IRQ stage: Linux
> Apr 08 01:52:12 trixie-box kernel: RIP:
> 0010:rtnet_rtpc_dispatch_call+0x2df/0x3d0 [rtnet]
> Apr 08 01:52:12 trixie-box kernel: Code: 26 50 00 00 00 0f 85 d1 fd ff
> ff 31 c9 48 c7 c2 00 13 94 c0 4c 89 ee 48 c7 c7 48 13 94 c0 c6 05 06 50
> 00 00 01 e8 51 d6 97 eb <0f> 0b e9 ab fd ff ff 84 d2 >
> Apr 08 01:52:12 trixie-box kernel: RSP: 0018:ffffcdff82ebbd68 EFLAGS:
> 00010286
> Apr 08 01:52:12 trixie-box kernel: RAX: 0000000000000000 RBX:
> ffff8988585012c0 RCX: 0000000000000027
> Apr 08 01:52:12 trixie-box kernel: RDX: ffff898b6fca0948 RSI:
> 0000000000000001 RDI: ffff898b6fca0940
> Apr 08 01:52:12 trixie-box kernel: RBP: 00000000000001f4 R08:
> 0000000000000000 R09: ffffcdff82ebbb28
> Apr 08 01:52:12 trixie-box kernel: R10: ffffffffadeb9ee8 R11:
> 0000000000000003 R12: 0000000000000000
> Apr 08 01:52:12 trixie-box kernel: R13: 0000000000000050 R14:
> ffffffffc0cff5b0 R15: ffffffffc0cff6e0
> Apr 08 01:52:12 trixie-box kernel: FS:  00007f1499d11740(0000)
> GS:ffff898b6fc80000(0000) knlGS:0000000000000000
> Apr 08 01:52:12 trixie-box kernel: CS:  0010 DS: 0000 ES: 0000 CR0:
> 0000000080050033
> Apr 08 01:52:12 trixie-box kernel: CR2: 00007ffe3de7dff8 CR3:
> 000000014d74e000 CR4: 0000000000750ef0
> Apr 08 01:52:12 trixie-box kernel: PKRU: 55555554
> Apr 08 01:52:12 trixie-box kernel: Call Trace:
> Apr 08 01:52:12 trixie-box kernel:  <TASK>
> Apr 08 01:52:12 trixie-box kernel:  ipv4_ioctl+0x16b/0x230 [rtipv4]
> Apr 08 01:52:12 trixie-box kernel:  rtnet_ioctl+0x164/0x190 [rtnet]
> Apr 08 01:52:12 trixie-box kernel:  ? __count_memcg_events+0x67/0x100
> Apr 08 01:52:12 trixie-box kernel:  __x64_sys_ioctl+0x91/0xd0
> Apr 08 01:52:12 trixie-box kernel:  do_syscall_64+0xc0/0x230
> Apr 08 01:52:12 trixie-box kernel: entry_SYSCALL_64_after_hwframe+0x76/0x7e
> Apr 08 01:52:12 trixie-box kernel: RIP: 0033:0x7f1499e2091b
> Apr 08 01:52:12 trixie-box kernel: Code: 00 48 89 44 24 18 31 c0 48 8d
> 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24
> 10 b8 10 00 00 00 0f 05 <89> c2 3d 00 f0 ff ff 77 1c >
> Apr 08 01:52:12 trixie-box kernel: RSP: 002b:00007ffe3de7e9f0 EFLAGS:
> 00000246 ORIG_RAX: 0000000000000010
> Apr 08 01:52:12 trixie-box kernel: RAX: ffffffffffffffda RBX:
> 00007f1499d11740 RCX: 00007f1499e2091b
> Apr 08 01:52:12 trixie-box kernel: RDX: 000055c8e00d60e0 RSI:
> 00000000c0500285 RDI: 0000000000000003
> Apr 08 01:52:12 trixie-box kernel: RBP: 0000000000000002 R08:
> 0000000000000000 R09: 0000000000000000
> Apr 08 01:52:12 trixie-box kernel: R10: 0000000000000003 R11:
> 0000000000000246 R12: 00007ffe3de7f818
> Apr 08 01:52:12 trixie-box kernel: R13: 0000000000000002 R14:
> 00007f1499f60000 R15: 000055c8e00d5dd8
> Apr 08 01:52:12 trixie-box kernel:  </TASK>
> Apr 08 01:52:12 trixie-box kernel: ---[ end trace 0000000000000000 ]---
> 
> This can be fixed with the following patch. However, I am not sure if
> it's the right way to do it.
> diff --git a/kernel/drivers/net/stack/rtnet_rtpc.c b/kernel/drivers/net/
> stack/rtnet_rtpc.c
> index cd5f0546d..caed27269 100644
> --- a/kernel/drivers/net/stack/rtnet_rtpc.c
> +++ b/kernel/drivers/net/stack/rtnet_rtpc.c
> @@ -90,7 +90,7 @@ int rtnet_rtpc_dispatch_call(rtpc_proc proc, unsigned
> int timeout,
>      if (call == NULL)
>          return -ENOMEM;
> 
> -    memcpy(call->priv_data, priv_data, priv_data_size);
> +    unsafe_memcpy(call->priv_data, priv_data, priv_data_size, /
> *embedded variable length data*/);
> 

Yeah, good catch - the kernel and the toolchains improved their
diagnosis features a lot since this was written more than 20 years ago.
I bet we have a few more of those variable length fields in RTnet data
structures. Those should be converted to flexible array members, see
also
https://www.kernel.org/doc/html/latest/process/deprecated.html#zero-length-and-one-element-arrays.
Happy to review patches!

Jan

>      call->processed = 0;
>      call->proc = proc;
> 
> 

-- 
Siemens AG, Foundational Technologies
Linux Expert Center