From: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
To: Sagi Grimberg <sagi@grimberg.me>
Cc: Logan Gunthorpe <logang@deltatee.com>,
Christoph Hellwig <hch@lst.de>,
"James E.J. Bottomley" <jejb@linux.vnet.ibm.com>,
"Martin K. Petersen" <martin.petersen@oracle.com>,
Jens Axboe <axboe@kernel.dk>,
Steve Wise <swise@opengridcomputing.com>,
Stephen Bates <sbates@raithlin.com>,
Max Gurtovoy <maxg@mellanox.com>,
Dan Williams <dan.j.williams@intel.com>,
Keith Busch <keith.busch@intel.com>,
linux-pci@vger.kernel.org, linux-scsi@vger.kernel.org,
linux-nvme@lists.infradead.org, linux-rdma@vger.kernel.org,
linux-nvdimm@ml01.01.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC 6/8] nvmet: Be careful about using iomem accesses when dealing with p2pmem
Date: Thu, 6 Apr 2017 10:35:21 -0600 [thread overview]
Message-ID: <20170406163520.GA7657@obsidianresearch.com> (raw)
In-Reply-To: <4df229d8-8124-664a-9bc4-6401bc034be1@grimberg.me>
On Thu, Apr 06, 2017 at 08:33:38AM +0300, Sagi Grimberg wrote:
>
> >>Note that the nvme completion queues are still on the host memory, so
> >>this means we have lost the ordering between data and completions as
> >>they go to different pcie targets.
> >
> >Hmm, in this simple up/down case with a switch, I think it might
> >actually be OK.
> >
> >Transactions might not complete at the NVMe device before the CPU
> >processes the RDMA completion, however due to the PCI-E ordering rules
> >new TLPs directed to the NVMe will complete after the RMDA TLPs and
> >thus observe the new data. (eg order preserving)
> >
> >It would be very hard to use P2P if fabric ordering is not preserved..
>
> I think it still can race if the p2p device is connected with more than
> a single port to the switch.
>
> Say it's connected via 2 legs, the bar is accessed from leg A and the
> data from the disk comes via leg B. In this case, the data is heading
> towards the p2p device via leg B (might be congested), the completion
> goes directly to the RC, and then the host issues a read from the
> bar via leg A. I don't understand what can guarantee ordering here.
Right, this is why I qualified my statement with 'simple up/down case'
Make it any more complex and it clearly stops working sanely, but I
wouldn't worry about unusual PCI-E fabrics at this point..
> Stephen told me that this still guarantees ordering, but I honestly
> can't understand how, perhaps someone can explain to me in a simple
> way that I can understand.
AFAIK PCI-E ordering is explicitly per link, so things that need order
must always traverse the same link.
Jason
next prev parent reply other threads:[~2017-04-06 16:36 UTC|newest]
Thread overview: 81+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-30 22:12 [RFC 0/8] Copy Offload with Peer-to-Peer PCI Memory Logan Gunthorpe
2017-03-30 22:12 ` [RFC 1/8] Introduce Peer-to-Peer memory (p2pmem) device Logan Gunthorpe
2017-03-31 18:49 ` Sinan Kaya
2017-03-31 21:23 ` Logan Gunthorpe
2017-03-31 21:38 ` Sinan Kaya
2017-03-31 22:42 ` Logan Gunthorpe
2017-03-31 23:51 ` Sinan Kaya
2017-04-01 1:57 ` Logan Gunthorpe
2017-04-01 2:17 ` okaya
2017-04-01 22:16 ` Logan Gunthorpe
2017-04-02 2:26 ` Sinan Kaya
2017-04-02 17:21 ` Logan Gunthorpe
2017-04-02 21:03 ` Sinan Kaya
2017-04-03 4:26 ` Logan Gunthorpe
2017-04-25 11:58 ` Marta Rybczynska
2017-04-25 16:58 ` Logan Gunthorpe
2017-03-30 22:12 ` [RFC 2/8] cxgb4: setup pcie memory window 4 and create p2pmem region Logan Gunthorpe
2017-04-04 10:42 ` Sagi Grimberg
2017-04-04 15:56 ` Logan Gunthorpe
2017-04-05 15:41 ` Steve Wise
2017-03-30 22:12 ` [RFC 3/8] nvmet: Use p2pmem in nvme target Logan Gunthorpe
2017-04-04 10:40 ` Sagi Grimberg
2017-04-04 16:16 ` Logan Gunthorpe
2017-04-06 5:47 ` Sagi Grimberg
2017-04-06 15:52 ` Logan Gunthorpe
2017-03-30 22:12 ` [RFC 4/8] p2pmem: Add debugfs "stats" file Logan Gunthorpe
2017-04-04 10:46 ` Sagi Grimberg
2017-04-04 17:25 ` Logan Gunthorpe
2017-04-05 15:43 ` Steve Wise
2017-03-30 22:12 ` [RFC 5/8] scatterlist: Modify SG copy functions to support io memory Logan Gunthorpe
2017-03-31 7:09 ` Christoph Hellwig
2017-03-31 15:41 ` Logan Gunthorpe
2017-04-03 21:20 ` Logan Gunthorpe
2017-04-03 21:44 ` Dan Williams
2017-04-03 22:10 ` Logan Gunthorpe
2017-04-03 22:47 ` Dan Williams
2017-04-03 23:12 ` Logan Gunthorpe
2017-04-04 0:07 ` Dan Williams
2017-04-07 17:59 ` Logan Gunthorpe
2017-03-30 22:12 ` [RFC 6/8] nvmet: Be careful about using iomem accesses when dealing with p2pmem Logan Gunthorpe
2017-04-04 10:59 ` Sagi Grimberg
2017-04-04 15:46 ` Jason Gunthorpe
2017-04-04 17:21 ` Logan Gunthorpe
2017-04-06 5:33 ` Sagi Grimberg
2017-04-06 16:02 ` Logan Gunthorpe
2017-04-06 16:35 ` Jason Gunthorpe [this message]
2017-04-07 11:19 ` Stephen Bates
2017-04-10 8:29 ` Sagi Grimberg
2017-04-10 16:03 ` Logan Gunthorpe
2017-03-30 22:12 ` [RFC 7/8] p2pmem: Support device removal Logan Gunthorpe
2017-03-30 22:12 ` [RFC 8/8] p2pmem: Added char device user interface Logan Gunthorpe
2017-04-12 5:22 ` [RFC 0/8] Copy Offload with Peer-to-Peer PCI Memory Benjamin Herrenschmidt
2017-04-12 17:09 ` Logan Gunthorpe
2017-04-12 21:55 ` Benjamin Herrenschmidt
2017-04-13 21:22 ` Logan Gunthorpe
2017-04-13 22:37 ` Benjamin Herrenschmidt
2017-04-13 23:26 ` Bjorn Helgaas
2017-04-14 4:16 ` Jason Gunthorpe
2017-04-14 4:40 ` Logan Gunthorpe
2017-04-14 11:37 ` Benjamin Herrenschmidt
2017-04-14 11:39 ` Benjamin Herrenschmidt
2017-04-14 11:37 ` Benjamin Herrenschmidt
2017-04-14 17:30 ` Logan Gunthorpe
2017-04-14 19:04 ` Bjorn Helgaas
2017-04-14 22:07 ` Benjamin Herrenschmidt
2017-04-15 17:41 ` Logan Gunthorpe
2017-04-15 22:09 ` Dan Williams
2017-04-16 3:01 ` Benjamin Herrenschmidt
2017-04-16 4:46 ` Logan Gunthorpe
2017-04-16 15:53 ` Dan Williams
2017-04-16 16:34 ` Logan Gunthorpe
2017-04-16 22:31 ` Benjamin Herrenschmidt
2017-04-24 7:36 ` Knut Omang
2017-04-24 16:14 ` Logan Gunthorpe
2017-04-25 6:30 ` Knut Omang
2017-04-25 17:03 ` Logan Gunthorpe
2017-04-25 21:23 ` Stephen Bates
2017-04-25 21:23 ` Stephen Bates
2017-04-16 22:26 ` Benjamin Herrenschmidt
2017-04-15 22:17 ` Benjamin Herrenschmidt
2017-04-16 5:36 ` Logan Gunthorpe
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=20170406163520.GA7657@obsidianresearch.com \
--to=jgunthorpe@obsidianresearch.com \
--cc=axboe@kernel.dk \
--cc=dan.j.williams@intel.com \
--cc=hch@lst.de \
--cc=jejb@linux.vnet.ibm.com \
--cc=keith.busch@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nvdimm@ml01.01.org \
--cc=linux-nvme@lists.infradead.org \
--cc=linux-pci@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=logang@deltatee.com \
--cc=martin.petersen@oracle.com \
--cc=maxg@mellanox.com \
--cc=sagi@grimberg.me \
--cc=sbates@raithlin.com \
--cc=swise@opengridcomputing.com \
/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;
as well as URLs for NNTP newsgroup(s).