From: Vitaliy Makkoveev Subject: Re: Skip fuse FS by /usr/libexec/security To: Claudio Jeker Cc: Ingo Schwarze , tech@openbsd.org Date: Sun, 19 May 2024 00:59:45 +0300 > On 18 May 2024, at 12:46, Claudio Jeker wrote: > > On Sat, May 18, 2024 at 10:08:09AM +0100, Kirill A. Korinsky wrote: >> On Fri, 17 May 2024 23:41:41 +0100, >> Vitaliy Makkoveev wrote: >>> >>> I was confused with explicitly setting MNT_LOCAL on `mnt_flag' which >>> seems to be not required. However, it is enough to drop MNT_LOCAL in >>> corresponding `vfsconflist' entry. >>> >> >> Just tested, and it indeed ok. >> >> If I recall right, I've started my original patch as drop MNT_LOCAL in >> vfsconflist but after that I've tried to keep it as option, that makes >> things much more complicated and confusing, indeed. >> >>> Index: sys/kern/vfs_init.c >>> =================================================================== >>> RCS file: /cvs/src/sys/kern/vfs_init.c,v >>> diff -u -p -r1.43 vfs_init.c >>> --- sys/kern/vfs_init.c 26 Dec 2019 13:30:54 -0000 1.43 >>> +++ sys/kern/vfs_init.c 17 May 2024 22:25:20 -0000 >>> @@ -92,7 +92,7 @@ static struct vfsconf vfsconflist[] = { >>> #endif >>> >>> #ifdef FUSE >>> - { &fusefs_vfsops, MOUNT_FUSEFS, 18, 0, MNT_LOCAL, >>> + { &fusefs_vfsops, MOUNT_FUSEFS, 18, 0, 0, >>> sizeof(struct fusefs_args) }, >>> #endif >>> >> >> Shall we preserve comment with explanation why it's no MNT_LOCAL here? >> It was suitable in fusefs_mount() to explain why we clean this flag. But in the `vfsconflist' it is declared as declared. >> BTW a bit above I see: >> >> { &nfs_vfsops, MOUNT_NFS, 2, 0, MNT_SWAPPABLE, >> sizeof(struct nfs_args) }, >> >> which seems quite wired because keep SWAP on NFS probably extreamly bad idea >> due to network latency. Shall it be dropped as well? Have I missed something? > > This is for pure diskless setups and "works". While it is not great to > swap on NFS it is sometimes the only way. So no, this should remain. So, no objections to commit my diff?