[ofa-general] Re: [PATCH] rdma_cm: Add debugfs entries to monitor rdma_cm connections

Jason Gunthorpe jgunthorpe at obsidianresearch.com
Wed Jul 8 11:08:39 PDT 2009


On Wed, Jul 08, 2009 at 12:56:40PM +0300, Or Gerlitz wrote:
> Jason Gunthorpe wrote:
> > Well, thats the rub isn't it? debugfs is not 'production ready' (by
> > definition) so why spend time on it?
> 
> to allow debugging, diagnosing problems
> 
> > Why the resistance to doing a proper job and solving the actual problem?
> 
> I didn't see resistance to do a better job,

I don't see any patches..
 
> > debufs is entirely the wrong answer - this is not kernel debugging,
> > this is required user space diagnostic information.
> 
> said who?

Read the kernel devs postings and papers on debugfs and its intended
role in the system.

> how are you suggesting me to know what kernel iser/lustre/rds/rnfs RDMA connections 
> are open now? 
 
A userspace program analogous to ss or netstat is the proper way to do
this - exactly the same as how NFS/iSCSI/etc kernel TCP connections
show up today.

How do you propose to finish the job? RDMACM information is such a
minor export - what about the rest of it:
 * Completion Channel status and information
 * CQ status and information, and mappings to CCs
 * QP status and information and mappings to CQs
  (Including GRH and LRHs used for the QP, RDMACM and IBCM meta
  information used to establish the connection, APM data and status
  from the CM, all the QP attributes, etc)
 * MR and PD information
 * Process and FD bindings for all of the above
 * lsof should be able to show every RDMA object related to a process

Jason



More information about the general mailing list