Message ID | 5BD19B73.7050607@huawei.com (mailing list archive) |
---|---|
Headers | show |
Series | ocfs2: support IPv6 communication in o2cb cluster | expand |
On Thu, 25 Oct 2018 18:31:15 +0800 piaojun <piaojun@huawei.com> wrote: > o2cb cluster only support IPv4 node configuration nowadays. Along with IPv6 > used more widely, ocfs2 should also keep up with the pace. This patch > series add configfs attribute for IPv6 configure and build connnection > between nodes with IPv6 socket. So the nodes in cluster could send dlm > messages through IPv6 network protocol. Big patchset. I'll await some reviewer input before looking at merging this, I think. Please update Documentation/filesystems/ocfs2.txt to document all the new userspace interfaces and the overall feature, etc. Should the new code be hidden behind `#ifdef CONFIG_IPV6'? That would avoid some code bloat. (That being said, if ocfs2 was concerned about code bloat, it would do something about the fact that mlog() and mlog_errno() more than double the file system's object code size).
Hi Andrew, On 2018/11/8 6:56, Andrew Morton wrote: > On Thu, 25 Oct 2018 18:31:15 +0800 piaojun <piaojun@huawei.com> wrote: > >> o2cb cluster only support IPv4 node configuration nowadays. Along with IPv6 >> used more widely, ocfs2 should also keep up with the pace. This patch >> series add configfs attribute for IPv6 configure and build connnection >> between nodes with IPv6 socket. So the nodes in cluster could send dlm >> messages through IPv6 network protocol. > > Big patchset. I'll await some reviewer input before looking at merging > this, I think. > > Please update Documentation/filesystems/ocfs2.txt to document all the > new userspace interfaces and the overall feature, etc. OK, there is some work in userspace, such as o2cb_ctl, and I will add another patch to update the document. > > Should the new code be hidden behind `#ifdef CONFIG_IPV6'? That would > avoid some code bloat. > > (That being said, if ocfs2 was concerned about code bloat, it would do > something about the fact that mlog() and mlog_errno() more than double > the file system's object code size). Good suggestion, and I will consider using CONFIG_IPV6 to make code thiner. Thanks, Jun > > . >