DPDK patches and discussions
 help / color / mirror / Atom feed
From: Stephen Hemminger <stephen@networkplumber.org>
To: Ferruh Yigit <ferruh.yigit@intel.com>
Cc: Thomas Monjalon <thomas@monjalon.net>,
	David Marchand <david.marchand@redhat.com>,
	Elad Nachman <eladv6@gmail.com>, <dev@dpdk.org>,
	<stable@dpdk.org>, "Igor Ryzhov" <iryzhov@nfware.com>,
	Eric Christian <erclists@gmail.com>
Subject: Re: [PATCH] kni: restrict bifurcated device support
Date: Tue, 23 Nov 2021 08:22:18 -0800
Message-ID: <20211123082218.11b894b0@hermes.local> (raw)
In-Reply-To: <f7269e9c-8238-0e43-96ba-0b9f14bf4f58@intel.com>

On Tue, 23 Nov 2021 09:54:01 +0000
Ferruh Yigit <ferruh.yigit@intel.com> wrote:

> On 10/9/2021 3:03 AM, Stephen Hemminger wrote:
> > On Sat,  9 Oct 2021 00:58:30 +0100
> > Ferruh Yigit <ferruh.yigit@intel.com> wrote:
> >   
> >> To enable bifurcated device support, rtnl_lock is released before calling
> >> userspace callbacks and asynchronous requests are enabled.
> >>
> >> But these changes caused more issues, like bug #809, #816. To reduce the
> >> scope of the problems, the bifurcated device support related changes are
> >> only enabled when it is requested explicitly with new 'enable_bifurcated'
> >> module parameter.
> >> And bifurcated device support is disabled by default.
> >>
> >> So the bifurcated device related problems are isolated and they can be
> >> fixed without impacting all use cases.
> >>
> >> Bugzilla ID: 816
> >> Fixes: 631217c76135 ("kni: fix kernel deadlock with bifurcated device")
> >> Cc: stable@dpdk.org
> >>
> >> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>  
> > 
> > Calling userspace with semaphore held is still risky and buggy.
> > There is no guarantee that the userspace DPDK application will be well behaved.
> > And if it is not, the spinning holding RTNL would break any other network management
> > functions in the kernel.
> >   
> 
> Hi Stephen,
> 
> Because of what you described above, we tried the option that releases the RTNL
> lock before userspace callback,
> That caused a deadlock in the ifdown, and we add a workaround for it.
> 
> But now we are facing with two more issues because of the rtnl lock release,
> - Bugzilla ID: 816, MAC set cause kernel deadlock
> - Some requests are overwritten (because of the workaround we put for ifdown)
> 
> This patch just converts the default behavior to what it was previously.
> Releasing rtnl lock still supported with the module param, but I think it
> is not good idea to make it default while it is know that it is buggy.
> 
> @Thomas, @David,
> Can it be possible to get this patch for -rc4? Since it has potential
> to cause a deadlock in kernel as it is.
> 
> I can send a new version with documentation update.

Is it possible for userspace to choose the correct behavior instead
of module option. Users will do it wrong.


> 
> > These are the kind of problems that make me think it there should be a
> > big "DO NOT USE THIS" onto KNI. Maybe make it print a big nasty message
> > (see kernel VFIO without IOMMU description) or mark kernel as tainted??
> > 
> > See: https://fedoraproject.org/wiki/KernelStagingPolicy
> > 
> > Something like:
> > 
> > diff --git a/kernel/linux/kni/kni_net.c b/kernel/linux/kni/kni_net.c
> > index 611719b5ee27..d47fc6133cbe 100644
> > --- a/kernel/linux/kni/kni_net.c
> > +++ b/kernel/linux/kni/kni_net.c
> > @@ -838,6 +838,14 @@ kni_net_init(struct net_device *dev)
> >   	dev->header_ops      = &kni_net_header_ops;
> >   	dev->ethtool_ops     = &kni_net_ethtool_ops;
> >   	dev->watchdog_timeo = WD_TIMEOUT;
> > +
> > +	/*
> > +	 * KNI is unsafe since it requires calling userspace to do
> > +	 * control operations. And the overall quality according to
> > +	 * kernel standards is the same as devices in staging.
> > +	 */
> > +	add_taint(TAINT_CRAP, LOCKDEP_STILL_OK);
> > +	netdev_warn(dev, "Adding kernel taint for KNI because it is not safe\n");  
> 
> I am for discussing above separate from this patch, since this patch
> restores the behavior that exist since KNI module exists.

Any user of KNI will already get tainted because KNI is out of tree driver.

  reply	other threads:[~2021-11-23 16:22 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-10-08 23:58 [dpdk-dev] " Ferruh Yigit
2021-10-09  2:03 ` Stephen Hemminger
2021-11-23  9:54   ` Ferruh Yigit
2021-11-23 16:22     ` Stephen Hemminger [this message]
2021-11-23 16:51       ` Ferruh Yigit
2021-11-23 19:10         ` Stephen Hemminger
2021-11-17 16:42 ` Igor Ryzhov
2021-11-23 16:46 ` [PATCH v2] " Ferruh Yigit
2021-11-24 13:51   ` Thomas Monjalon

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=20211123082218.11b894b0@hermes.local \
    --to=stephen@networkplumber.org \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    --cc=eladv6@gmail.com \
    --cc=erclists@gmail.com \
    --cc=ferruh.yigit@intel.com \
    --cc=iryzhov@nfware.com \
    --cc=stable@dpdk.org \
    --cc=thomas@monjalon.net \
    /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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror http://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ http://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git