Date: Sat, 11 Jun 2005 18:20:16 +0900 From: Pyun YongHyeon <yongari@rndsoft.co.kr> To: John-Mark Gurney <gurney_j@resnet.uoregon.edu> Cc: Hiroki Sato <hrs@freebsd.org>, sparc64@freebsd.org Subject: Re: E4500 with 24GB RAM Message-ID: <20050611092015.GE19976@rndsoft.co.kr> In-Reply-To: <20050611084748.GF742@funkthat.com> References: <20050606132756.X16994@carver.gumbysoft.com> <20050611.004435.59726356.hrs@allbsd.org> <20050610211239.GA59402@xor.obsecurity.org> <20050611.154028.102195481.hrs@allbsd.org> <20050611072632.GB19976@rndsoft.co.kr> <20050611084748.GF742@funkthat.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jun 11, 2005 at 01:47:48AM -0700, John-Mark Gurney wrote: > Pyun YongHyeon wrote this message on Sat, Jun 11, 2005 at 16:26 +0900: > > I wonder how you can use NFS reliably on sparc64. Due to failure of > > alignment(both server and client) it's really easy to get panic on sparc64. > > I was recently trying to upgrade my sparc64 using a cross-built world, > and I couldn't get through an installworld with my sparc64 mounted.. > It kept on panicing.. I plan to take a look at them shortly, but I > don't think I was using tcp... I do have a few back traces if someone > wants to look at them... > Long ago I had a local hack that addresses the alignment issue. Basically it re-inspected whether the mbuf needs realignment in nfsm_disct(). It was ugly hack and gave performance penalty for architectures that allow non-aligned memory access. Personally I believe we should remove nfs_realign() that fails its mission and rewrite nfsm_disct to handle alignments as NetBSD did. In this way we can remove duplicated code, nfs_realign(client and server side), and get clean/central implementation of nfsm_disct since there is no need to call nfs_realign. -- Regards, Pyun YongHyeon http://www.kr.freebsd.org/~yongari | yongari@freebsd.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050611092015.GE19976>