> -----邮件原件----- > 发件人: Chuck Lever III <chuck.lever@xxxxxxxxxx> > 发送时间: 2024年4月6日 23:36 > 收件人: Chen, Hanxiao/ <chenhx.fnst@xxxxxxxxxxx> > 抄送: Jeff Layton <jlayton@xxxxxxxxxx>; Neil Brown <neilb@xxxxxxx>; Olga > Kornievskaia <kolga@xxxxxxxxxx>; Dai Ngo <dai.ngo@xxxxxxxxxx>; Tom > Talpey <tom@xxxxxxxxxx>; Linux NFS Mailing List <linux-nfs@xxxxxxxxxxxxxxx> > >> 主题: Re: [PATCH v2] NFSD: trace export root filehandle event > >> > >> On Fri, Mar 29, 2024 at 12:16:54AM +0800, Chen Hanxiao wrote: > >>> Add a tracepoint for obtaining root filehandle event > >> > >> I've had a chance to look at this more closely. It turns out we've > >> already got trace_nfsd_ctl_filehandle(). > >> > >> So let's only remove the dprintk call site in exp_rootfh(). > >> > >> > > Hi, > > > > struct svc_export should be a useful info to trace, such as ex_uuid. > > Then the patch description needs to explain why the existing > tracepoint is not adequate. Why does a trouble-shooter need > the UUID and export flags for, for example? > > > > Can we move trace_nfsd_ctl_filehandle into exp_rootfh ,then add an entry > to it to trace ex_uuid? > > No, trace_nfsd_ctl_* all report information from user space, so > they belong where they are now. > > If we need more information, then yes, another tracepoint can > be added. But I haven't seen a detailed rationale for this yet. > > <more-info-needed> ;-) Thanks for your detailed explanation. I'm trying to write patches to obtain root fh in user land by nfs-utils. The ex_uuid is useful info for userland tools to identify whether we got the right one. If the user land root fh patches is merged, maybe we could give the new tracepoint a chance. I'll send a patch to just remove the dprintk part in exp_rootfh. Regards, - Chen
