Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 28 Aug 2007 16:50:20 -0400
From:      John Baldwin <jhb@freebsd.org>
To:        src-committers@freebsd.org
Cc:        cvs-src@freebsd.org, cvs-all@freebsd.org
Subject:   Re: cvs commit: src/sys/compat/freebsd32 freebsd32_misc.c src/sys/kern vfs_syscalls.c src/sys/sys mount.h
Message-ID:  <200708281650.20825.jhb@freebsd.org>
In-Reply-To: <200708282028.l7SKSD4u028065@repoman.freebsd.org>
References:  <200708282028.l7SKSD4u028065@repoman.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday 28 August 2007 04:28:12 pm John Baldwin wrote:
> jhb         2007-08-28 20:28:12 UTC
> 
>   FreeBSD src repository
> 
>   Modified files:
>     sys/compat/freebsd32 freebsd32_misc.c 
>     sys/kern             vfs_syscalls.c 
>     sys/sys              mount.h 
>   Log:
>   Rework the routines to convert a 5.x+ statfs structure (with fixed-size
>   64-bit counters) to a 4.x statfs structure (with long-sized counters).
>   - For block counters, we scale up the block size sufficiently large so
>     that the resulting block counts fit into a the long-sized (long for the
>     ABI, so 32-bit in freebsd32) counters.  In 4.x the NFS client's statfs
>     VOP did this already.  This can lie about the block size to 4.x
>     binaries, but it presents a more accurate picture of the ratios of free
>     and available space.
>   - For non-block counters, fix the freebsd32 stats converter to cap the
>     values at INT32_MAX rather than losing the upper 32-bits to match the
>     behavior of the 4.x statfs conversion routine in vfs_syscalls.c

I'm open to suggestion on this point.  I can see that for the true counters 
(like read/write operation counters) it might be more correct for them to 
wrap and just lose the upper 32-bits than to cap them.

-- 
John Baldwin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200708281650.20825.jhb>