Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Leon Romanovsky <leon@kernel.org>
Cc: Wenpeng Liang <liangwenpeng@huawei.com>,
	linux-rdma@vger.kernel.org, linuxarm@huawei.com
Subject: Re: [PATCH v4 for-next 1/1] RDMA/hns: Support direct wqe of userspace
Date: Thu, 25 Nov 2021 13:50:44 -0400	[thread overview]
Message-ID: <20211125175044.GA504288@nvidia.com> (raw)
In-Reply-To: <YZtboTThVCL7xs5s@unreal>

On Mon, Nov 22, 2021 at 10:58:09AM +0200, Leon Romanovsky wrote:
> On Mon, Nov 22, 2021 at 11:38:01AM +0800, Wenpeng Liang wrote:
> > From: Yixing Liu <liuyixing1@huawei.com>
> > 
> > Add direct wqe enable switch and address mapping.
> > 
> > Signed-off-by: Yixing Liu <liuyixing1@huawei.com>
> > Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
> >  drivers/infiniband/hw/hns/hns_roce_device.h |  8 +--
> >  drivers/infiniband/hw/hns/hns_roce_main.c   | 38 ++++++++++++---
> >  drivers/infiniband/hw/hns/hns_roce_pd.c     |  3 ++
> >  drivers/infiniband/hw/hns/hns_roce_qp.c     | 54 ++++++++++++++++++++-
> >  include/uapi/rdma/hns-abi.h                 |  2 +
> >  5 files changed, 94 insertions(+), 11 deletions(-)
> 
> <...>
> 
> >  	entry = to_hns_mmap(rdma_entry);
> >  	pfn = entry->address >> PAGE_SHIFT;
> > -	prot = vma->vm_page_prot;
> >  
> > -	if (entry->mmap_type != HNS_ROCE_MMAP_TYPE_TPTR)
> > -		prot = pgprot_noncached(prot);
> > +	switch (entry->mmap_type) {
> > +	case HNS_ROCE_MMAP_TYPE_DB:
> > +		prot = pgprot_noncached(vma->vm_page_prot);
> > +		break;
> > +	case HNS_ROCE_MMAP_TYPE_TPTR:
> > +		prot = vma->vm_page_prot;
> > +		break;
> > +	case HNS_ROCE_MMAP_TYPE_DWQE:
> > +		prot = pgprot_device(vma->vm_page_prot);
> 
> Everything fine, except this pgprot_device(). You probably need to check
> WC internally in your driver and use or pgprot_writecombine() or
> pgprot_noncached() explicitly.

pgprot_device is only used in two places in the kernel
pci_mmap_resource_range() for setting up the sysfs resourceXX mmap

And in pci_remap_iospace() as part of emulationg PIO on mmio
architectures

So, a PCI device should always be using pgprot_device() in its mmap
function

The question is why is pgprot_noncached() being used at all? The only
difference on ARM is that noncached is non-Early Write Acknowledgement
and devices is not.

At the very least this should be explained in a comment why nE vs E is
required in all these cases.

Jason

  parent reply	other threads:[~2021-11-25 17:52 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-22  3:38 [PATCH v4 for-next 0/1] RDMA/hns: Support direct WQE of userspace Wenpeng Liang
2021-11-22  3:38 ` [PATCH v4 for-next 1/1] RDMA/hns: Support direct wqe " Wenpeng Liang
2021-11-22  8:58   ` Leon Romanovsky
2021-11-22  9:28     ` Wenpeng Liang
2021-11-22 11:10       ` Leon Romanovsky
2021-11-22 12:36         ` Wenpeng Liang
2021-11-25 17:50     ` Jason Gunthorpe [this message]
2021-11-26  8:25       ` Wenpeng Liang
2021-11-26 12:16         ` Jason Gunthorpe
2021-11-27  9:04           ` Wenpeng Liang

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=20211125175044.GA504288@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=leon@kernel.org \
    --cc=liangwenpeng@huawei.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=linuxarm@huawei.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