Shouldn't be a problem. The socket used for lookup is
AF_UNIX (uses unp_connectat) and the NFS socket
will always be UDP or TCP.
Different sockets imply different socket locks.
At least that's my interpretation, rick
On Wed, Apr 3, 2024 at 11:33 AM Bjoern A. Zeeb
wrote:
>
>
> NFS root boot of a Lab machine; calling wpa_cli:
>
> Thilock order reversal:
> 1st 0x0001d4e1c800 so_snd_sx (so_snd_sx, sx) @
> /usr/src/sys/kern/uipc_socket.c:4020
> 2nd 0xa020cb20e930 nfs (nfs, lockmgr) @
> /usr/src/sys/kern/vfs_lookup.c:1083
> lock order nfs -> so_snd_sx established at:
> #0 0x00529588 at witness_checkorder+0x328
> #1 0x004bdf48 at _sx_xlock+0x70
> #2 0x005687e0 at soiolock+0x5c
> #3 0x00567ff0 at sosend_generic+0x104
> #4 0x005688b8 at sosend+0x48
> #5 0x0076b6a0 at clnt_vc_call+0x570
> #6 0x00769914 at clnt_reconnect_call+0x1c4
> #7 0x003552ec at newnfs_request+0x7e4
> #8 0x0037abf0 at nfsrpc_getattrnovp+0xfc
> #9 0x0039823c at mountnfs+0x6ec
> #10 0x00395c64 at nfs_mount+0xe78
> #11 0x0059b59c at vfs_mount_sigdefer+0x30
> #12 0x005a44c8 at vfs_domount_first+0x254
> #13 0x005a0884 at vfs_domount+0x2d4
> #14 0x0059f1ec at vfs_donmount+0x824
> #15 0x005a3438 at kernel_mount+0x64
> #16 0x005a72b0 at parse_mount+0x494
> #17 0x005a59ac at vfs_mountroot+0x5b8
> lock order so_snd_sx -> nfs attempted at:
> #0 0x00529cd8 at witness_checkorder+0xa78
> #1 0x0047edb8 at lockmgr_lock_flags+0x78
> #2 0x00390044 at nfs_lock+0x34
> #3 0x0059793c at vop_sigdefer+0x38
> #4 0x005c3734 at _vn_lock+0x58
> #5 0x0059cb78 at vfs_lookup+0x12c
> #6 0x0059c01c at namei+0x280
> #7 0x005751f8 at unp_connectat+0x244
> #8 0x00576adc at uipc_sosend_dgram+0x3c0
> #9 0x005689c8 at sousrsend+0x80
> #10 0x0056ec90 at kern_sendit+0x1e4
> #11 0x0056ef78 at sendit+0x1b0
> #12 0x0056edb4 at sys_sendto+0x4c
> #13 0x0086ad40 at do_el0_sync+0x59c
> #14 0x0084391c at handle_el0_sync+0x48
>
>
> --
> Bjoern A. Zeeb r15:7
>
