public inbox for linux-rdma@vger.kernel.org
 help / color / mirror / Atom feed
From: Eli Cohen <eli-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
To: Pradeep Satyanarayana
	<pradeeps-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: Re: Infinihost HCA ->MTT -are they per HCA?
Date: Wed, 9 Feb 2011 08:17:19 +0200	[thread overview]
Message-ID: <20110209061719.GA30553@mtldesk30> (raw)
In-Reply-To: <4D4355ED.50301-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>

On Fri, Jan 28, 2011 at 03:49:01PM -0800, Pradeep Satyanarayana wrote:
> We have the following HCAs in one of the machines:
> 
> InfiniBand: Mellanox Technologies MT25204 [InfiniHost III Lx HCA] (rev 20)
> 
> There are two adapters in the machine and from various experiments
> we find that we are limited to being able to register
> (ibv_reg_mr()) 28 GB. My calculations show one should be able to
> register up to 56 GB. I would like to understand the difference.
> 
> In /sys/module/ib_mthca/parameters I see the following values:
> 
> fmr_reserved_mtts = 262144
> log_mtts_per_seg = 3
> num_mpt = 131072
> num_mtt = 2097152
> 
> Therefore max memory that one can register should be (2097152
> -262144) * 4K * 2^(log_mtts_per_seg)
> i.e. 1.75 MB * 4K * 8 = 56 GB.
> 
> Since ibv_reg_mr() ties one to a pd and thus a particular device, is
> the 56 GB split between the 2 HCAs in the machine?
> How does one explain the difference?
> 

Your calaculations are correct; each device should be able,
theoretically, to register 56 GB of memory. I assume you have enough
memory in your machine for the memory to be registered and also ICM
memory that is required to support all the registered memory. Also it
depends on how many registered memory regions you use and and their
sizes. For example, since the MTT segments allocator is a buddy
allocator, it allocates 2 ^ n mtt segments each time. So if your
memory region requires only [2 ^ (n -1)] + 1 entries, it will still
consume 2 ^ n entries from the allocator.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

      parent reply	other threads:[~2011-02-09  6:17 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-01-28 23:49 Infinihost HCA ->MTT -are they per HCA? Pradeep Satyanarayana
     [not found] ` <4D4355ED.50301-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
2011-02-09  6:17   ` Eli Cohen [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20110209061719.GA30553@mtldesk30 \
    --to=eli-ldsdmyg8hgv8yrgs2mwiifqbs+8scbdb@public.gmane.org \
    --cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=pradeeps-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox