On Tue, Sep 01, 2026 at 05:34:52PM +0000, Rick Macklem wrote: R> The branch main has been updated by rmacklem: R> R> URL: https://cgit.FreeBSD.org/src/commit/?id=cc352a850f9860eadf9558cf0c920a3655fae576 R> R> commit cc352a850f9860eadf9558cf0c920a3655fae576 R> Author: Rick Macklem <[email protected]> R> AuthorDate: 2026-09-01 17:33:08 +0000 R> Commit: Rick Macklem <[email protected]> R> CommitDate: 2026-09-01 17:33:08 +0000 R> R> nfs_clvfsops.c: Fix the non-VIMAGE build R> R> The OFED code checks for a vnet argument, but it is R> is not defined. R> R> Reported by: glebius R> MFC after: 3 months R> Fixes: 884ee8d6c9b4 ("nfscl: Add some glue for client side NFS over RDMA") R> --- R> sys/fs/nfsclient/nfs_clvfsops.c | 4 ++++ R> 1 file changed, 4 insertions(+) R> R> diff --git a/sys/fs/nfsclient/nfs_clvfsops.c b/sys/fs/nfsclient/nfs_clvfsops.c R> index afe954c2a69d..f336e320b04c 100644 R> --- a/sys/fs/nfsclient/nfs_clvfsops.c R> +++ b/sys/fs/nfsclient/nfs_clvfsops.c R> @@ -1544,7 +1544,11 @@ nfs_mount(struct mount *mp) R> error = EINVAL; R> goto out; R> } R> +#ifdef VIMAGE R> error = rdma_check_route(vnet0, nam, NFSV4_CBSLOTS); R> +#else R> + error = rdma_check_route(NULL, nam, NFSV4_CBSLOTS); R> +#endif
We might want a define for vnet0, so that this commit could have been a one liner. In net/vnet.h under !VIMAGE case: #define vnet0 NULL On the other hand, any module that wants to use vnet0, or check if curvnet is vnet0 or not, is very likely doing something wrong :) I haven't looked at rdma, but passing vnet as function parameter seems wrong. The above was rhetorical grumbling, not a request for any action :) -- Gleb Smirnoff
