Date: Sat, 15 Dec 2001 10:32:32 -0500 From: Jake Burkholder <jake@locore.ca> To: Kirk McKusick <mckusick@FreeBSD.org> Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/ufs/ffs ffs_alloc.c ffs_snapshot.c ffs_vfsops.c fs.h Message-ID: <20011215103232.C12719@locore.ca> In-Reply-To: <200112140015.fBE0F6Q99440@freefall.freebsd.org>; from mckusick@FreeBSD.org on Thu, Dec 13, 2001 at 04:15:06PM -0800 References: <200112140015.fBE0F6Q99440@freefall.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Apparently, On Thu, Dec 13, 2001 at 04:15:06PM -0800,
Kirk McKusick said words to the effect of;
> mckusick 2001/12/13 16:15:06 PST
>
> Modified files:
> sys/ufs/ffs ffs_alloc.c ffs_snapshot.c ffs_vfsops.c
> fs.h
> Log:
> Minimize the time necessary to suspend operations on a filesystem
> when taking a snapshot. The two time consuming operations are
Atomic_clear_char and atomic_set_char shouldn't be used in MI code
because they can't be implemented efficiently on all architectures. [1]
Sparc64 only has compare and set instructions for 4 bytes or 8 bytes,
which makes the char and short operations very difficult to implement
properly.
The whole thing is also still under Giant, which has the correct
memory barriers to ensure that stores to locations other than the
lock itself are correctly drained to memory as its released.
Is there a way you can do this without the atomic ops?
Cheers,
Jake
[1] I removed all use of the char and short atomic ops in MI code
when the sparc64 port was first committed, but neglected to update
the man page to mention that their use is discouraged. I will do so.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011215103232.C12719>