Subject: Re: stat(1)
To: None <>
From: Geoff Adams <>
List: current-users
Date: 12/28/2001 14:58:24
On Friday, December 28, 2001, at 01:06  PM, Geoff Adams wrote:

> On Friday, December 28, 2001, at 10:03  AM, Martin Weber wrote:
>> b) wth is st_gen about ? I think I only see files with st_gen of 
>> 0 ... ?
> I believe this is the inode generation number. If yours are all zero, 
> then that just indicates that you haven't run fsirand(8):

Actually, I think I need to take that back.

After reading the tech-security message that mouss referred to, I 
decided to have a look at the NetBSD vn_stat() function in 
src/sys/kern/vfs_vnops.c. It appears that we *always* return zero for 
the st_gen field, not just in the case of non-superuser calls, as the 
advisory included in the tech-security message suggests be done. Here 
are the last few lines of vn_stat():

           sb->st_blksize = va.va_blocksize;
           sb->st_flags = va.va_flags;
           sb->st_gen = 0;
           sb->st_blocks = va.va_bytes / S_BLKSIZE;
           return (0);

(This has been the case since version 1.22 of that file, committed in 
March of 1997 -- a couple weeks after that mail message.)

So, I submit that in any stat(1) that various people have mentioned 
writing, there is no use in printing out the st_gen field at all, at 
least on NetBSD. It'll always be returned as zero by stat(2). (This 
makes one wonder why it's in struct stat. Compatibility, I presume.)

- Geoff