intel-gfx.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: Jason Gunthorpe <jgg@nvidia.com>
To: Alex Williamson <alex.williamson@redhat.com>
Cc: David Airlie <airlied@linux.ie>,
	Tony Krowiak <akrowiak@linux.ibm.com>,
	Christian Borntraeger <borntraeger@de.ibm.com>,
	Cornelia Huck <cohuck@redhat.com>,
	Jonathan Corbet <corbet@lwn.net>, Daniel Vetter <daniel@ffwll.ch>,
	Diana Craciun <diana.craciun@oss.nxp.com>,
	dri-devel@lists.freedesktop.org,
	Eric Auger <eric.auger@redhat.com>,
	Eric Farman <farman@linux.ibm.com>,
	Harald Freudenberger <freude@linux.ibm.com>,
	Vasily Gorbik <gor@linux.ibm.com>,
	Heiko Carstens <hca@linux.ibm.com>,
	intel-gfx@lists.freedesktop.org,
	intel-gvt-dev@lists.freedesktop.org,
	Jani Nikula <jani.nikula@linux.intel.com>,
	Jason Herne <jjherne@linux.ibm.com>,
	Joonas Lahtinen <joonas.lahtinen@linux.intel.com>,
	kvm@vger.kernel.org, Kirti Wankhede <kwankhede@nvidia.com>,
	linux-doc@vger.kernel.org, linux-s390@vger.kernel.org,
	Matthew Rosato <mjrosato@linux.ibm.com>,
	Peter Oberparleiter <oberpar@linux.ibm.com>,
	Halil Pasic <pasic@linux.ibm.com>,
	Rodrigo Vivi <rodrigo.vivi@intel.com>,
	Vineeth Vijayan <vneethv@linux.ibm.com>,
	Zhi Wang <zhi.a.wang@intel.com>,
	"Raj, Ashok" <ashok.raj@intel.com>,
	Christoph Hellwig <hch@lst.de>,
	Leon Romanovsky <leonro@nvidia.com>,
	Max Gurtovoy <mgurtovoy@nvidia.com>,
	Yishai Hadas <yishaih@nvidia.com>,
	Zhenyu Wang <zhenyuw@linux.intel.com>
Subject: Re: [Intel-gfx] [PATCH v3 09/14] vfio/pci: Change vfio_pci_try_bus_reset() to use the dev_set
Date: Tue, 3 Aug 2021 13:41:52 -0300	[thread overview]
Message-ID: <20210803164152.GC1721383@nvidia.com> (raw)
In-Reply-To: <20210803103406.5e1be269.alex.williamson@redhat.com>

On Tue, Aug 03, 2021 at 10:34:06AM -0600, Alex Williamson wrote:
> On Wed, 28 Jul 2021 21:49:18 -0300
> Jason Gunthorpe <jgg@nvidia.com> wrote:
> 
> > Keep track of all the vfio_devices that have been added to the device set
> > and use this list in vfio_pci_try_bus_reset() instead of trying to work
> > backwards from the pci_device.
> > 
> > The dev_set->lock directly prevents devices from joining/leaving the set,
> > which further implies the pci_device cannot change drivers or that the
> > vfio_device be freed, eliminating the need for get/put's.
> > 
> > Completeness of the device set can be directly measured by checking if
> > every PCI device in the reset group is also in the device set - which
> > proves that VFIO drivers are attached to everything.
> > 
> > This restructuring corrects a call to pci_dev_driver() without holding the
> > device_lock() and removes a hard wiring to &vfio_pci_driver.
> > 
> > Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> >  drivers/vfio/pci/vfio_pci.c | 148 +++++++++++++++---------------------
> >  1 file changed, 62 insertions(+), 86 deletions(-)
> > 
> > diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
> > index 5d6db93d6c680f..a1ae9a83a38621 100644
> > +++ b/drivers/vfio/pci/vfio_pci.c
> > @@ -404,6 +404,9 @@ static void vfio_pci_disable(struct vfio_pci_device *vdev)
> >  	struct vfio_pci_ioeventfd *ioeventfd, *ioeventfd_tmp;
> >  	int i, bar;
> >  
> > +	/* For needs_reset */
> > +	lockdep_assert_held(&vdev->vdev.dev_set->lock);
> > +
> >  	/* Stop the device from further DMA */
> >  	pci_clear_master(pdev);
> >  
> > @@ -2145,7 +2148,7 @@ static struct pci_driver vfio_pci_driver = {
> >  	.err_handler		= &vfio_err_handlers,
> >  };
> >  
> > -static int vfio_pci_get_unused_devs(struct pci_dev *pdev, void *data)
> > +static int vfio_pci_try_zap_and_vma_lock_cb(struct pci_dev *pdev, void *data)
> >  {
> >  	struct vfio_devices *devs = data;
> >  	struct vfio_device *device;
> > @@ -2165,8 +2168,11 @@ static int vfio_pci_get_unused_devs(struct pci_dev *pdev, void *data)
> >  
> >  	vdev = container_of(device, struct vfio_pci_device, vdev);
> >  
> > -	/* Fault if the device is not unused */
> > -	if (device->open_count) {
> > +	/*
> > +	 * Locking multiple devices is prone to deadlock, runaway and
> > +	 * unwind if we hit contention.
> > +	 */
> > +	if (!vfio_pci_zap_and_vma_lock(vdev, true)) {
> >  		vfio_device_put(device);
> >  		return -EBUSY;
> >  	}
> > @@ -2175,112 +2181,82 @@ static int vfio_pci_get_unused_devs(struct pci_dev *pdev, void *data)
> >  	return 0;
> >  }
> >  
> > -static int vfio_pci_try_zap_and_vma_lock_cb(struct pci_dev *pdev, void *data)
> > +static int vfio_pci_is_device_in_set(struct pci_dev *pdev, void *data)
> >  {
> > -	struct vfio_devices *devs = data;
> > -	struct vfio_device *device;
> > -	struct vfio_pci_device *vdev;
> > +	struct vfio_device_set *dev_set = data;
> > +	struct vfio_device *cur;
> >  
> > -	if (devs->cur_index == devs->max_index)
> > -		return -ENOSPC;
> > +	lockdep_assert_held(&dev_set->lock);
> >  
> > -	device = vfio_device_get_from_dev(&pdev->dev);
> > -	if (!device)
> > -		return -EINVAL;
> > -
> > -	if (pci_dev_driver(pdev) != &vfio_pci_driver) {
> > -		vfio_device_put(device);
> > -		return -EBUSY;
> > -	}
> > -
> > -	vdev = container_of(device, struct vfio_pci_device, vdev);
> > +	list_for_each_entry(cur, &dev_set->device_list, dev_set_list)
> > +		if (cur->dev == &pdev->dev)
> > +			return 0;
> > +	return -EBUSY;
> > +}
> >  
> > -	/*
> > -	 * Locking multiple devices is prone to deadlock, runaway and
> > -	 * unwind if we hit contention.
> > -	 */
> > -	if (!vfio_pci_zap_and_vma_lock(vdev, true)) {
> > -		vfio_device_put(device);
> > -		return -EBUSY;
> > +/*
> > + * vfio-core considers a group to be viable and will create a vfio_device even
> > + * if some devices are bound to drivers like pci-stub or pcieport.  Here we
> > + * require all PCI devices to be inside our dev_set since that ensures they stay
> > + * put and that every driver controlling the device can co-ordinate with the
> > + * device reset.
> > + */
> > +static struct pci_dev *vfio_pci_find_reset_target(struct vfio_pci_device *vdev)
> > +{
> > +	struct vfio_device_set *dev_set = vdev->vdev.dev_set;
> > +	struct vfio_pci_device *cur;
> > +	bool needs_reset = false;
> > +
> > +	/* No VFIO device has an open device FD */
> > +	list_for_each_entry(cur, &dev_set->device_list, vdev.dev_set_list) {
> > +		if (cur->vdev.open_count)
> > +			return NULL;
> > +		needs_reset |= cur->needs_reset;
> >  	}
> > +	if (!needs_reset)
> > +		return NULL;
> >  
> > -	devs->devices[devs->cur_index++] = vdev;
> > -	return 0;
> > +	/* All PCI devices in the group to be reset need to be in our dev_set */
> > +	if (vfio_pci_for_each_slot_or_bus(
> > +		    vdev->pdev, vfio_pci_is_device_in_set, dev_set,
> > +		    !pci_probe_reset_slot(vdev->pdev->slot)))
> > +		return NULL;
> > +	return cur->pdev;
> 
> 
> I don't understand the "reset target" aspect of this, cur->pdev is
> simply the last entry in the dev_set->devices_list...

Oh, hum, this got messed up someplace along the way, the original code
was just:

        /* Does at least one need a reset? */
        for (i = 0; i < devs.cur_index; i++) {
                tmp = devs.devices[i];
                if (tmp->needs_reset) {
                        ret = pci_reset_bus(vdev->pdev);
                        break;

So should this, I'll fix it up, thanks

> I think the vfio_pci_find_reset_target() function needs to be re-worked
> to just tell us true/false that it's ok to reset the provided device,
> not to anoint an arbitrary target device.  Thanks,

Yes, though this logic is confusing, why do we need to check if any
device needs a reset at this point? If we are being asked to reset
vdev shouldn't vdev needs_reset?

Or is the function more of a 'synchronize pending reset' kind of
thing?

Jason

  reply	other threads:[~2021-08-03 16:41 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-29  0:49 [Intel-gfx] [PATCH v3 00/14] Provide core infrastructure for managing open/release Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 01/14] vfio/samples: Remove module get/put Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 02/14] vfio/mbochs: Fix missing error unwind of mbochs_used_mbytes Jason Gunthorpe
2021-07-29  7:21   ` Christoph Hellwig
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 03/14] vfio: Introduce a vfio_uninit_group_dev() API call Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 04/14] vfio: Provide better generic support for open/release vfio_device_ops Jason Gunthorpe
2021-07-29  7:26   ` Christoph Hellwig
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 05/14] vfio/samples: Delete useless open/close Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 06/14] vfio/fsl: Move to the device set infrastructure Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 07/14] vfio/platform: Use open_device() instead of open coding a refcnt scheme Jason Gunthorpe
2021-08-05 12:37   ` Eric Auger
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 08/14] vfio/pci: Move to the device set infrastructure Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 09/14] vfio/pci: Change vfio_pci_try_bus_reset() to use the dev_set Jason Gunthorpe
2021-07-29  7:30   ` Christoph Hellwig
2021-08-03 16:34   ` Alex Williamson
2021-08-03 16:41     ` Jason Gunthorpe [this message]
2021-08-03 16:52       ` Alex Williamson
2021-08-05 11:47         ` Jason Gunthorpe
2021-08-05 17:33           ` Alex Williamson
2021-08-05 23:05             ` Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 10/14] vfio/pci: Reorganize VFIO_DEVICE_PCI_HOT_RESET to use the device set Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 11/14] vfio/mbochs: Fix close when multiple device FDs are open Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 12/14] vfio/ap, ccw: Fix open/close " Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 13/14] vfio/gvt: " Jason Gunthorpe
2021-07-29  0:49 ` [Intel-gfx] [PATCH v3 14/14] vfio: Remove struct vfio_device_ops open/release Jason Gunthorpe
2021-07-29  2:13 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for Provide core infrastructure for managing open/release (rev7) Patchwork
2021-07-29  2:37 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-07-29  8:16 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork
2021-08-05 14:26 ` [Intel-gfx] ✗ Fi.CI.BUILD: failure for Provide core infrastructure for managing open/release (rev8) Patchwork

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=20210803164152.GC1721383@nvidia.com \
    --to=jgg@nvidia.com \
    --cc=airlied@linux.ie \
    --cc=akrowiak@linux.ibm.com \
    --cc=alex.williamson@redhat.com \
    --cc=ashok.raj@intel.com \
    --cc=borntraeger@de.ibm.com \
    --cc=cohuck@redhat.com \
    --cc=corbet@lwn.net \
    --cc=daniel@ffwll.ch \
    --cc=diana.craciun@oss.nxp.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=eric.auger@redhat.com \
    --cc=farman@linux.ibm.com \
    --cc=freude@linux.ibm.com \
    --cc=gor@linux.ibm.com \
    --cc=hca@linux.ibm.com \
    --cc=hch@lst.de \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=intel-gvt-dev@lists.freedesktop.org \
    --cc=jani.nikula@linux.intel.com \
    --cc=jjherne@linux.ibm.com \
    --cc=joonas.lahtinen@linux.intel.com \
    --cc=kvm@vger.kernel.org \
    --cc=kwankhede@nvidia.com \
    --cc=leonro@nvidia.com \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-s390@vger.kernel.org \
    --cc=mgurtovoy@nvidia.com \
    --cc=mjrosato@linux.ibm.com \
    --cc=oberpar@linux.ibm.com \
    --cc=pasic@linux.ibm.com \
    --cc=rodrigo.vivi@intel.com \
    --cc=vneethv@linux.ibm.com \
    --cc=yishaih@nvidia.com \
    --cc=zhenyuw@linux.intel.com \
    --cc=zhi.a.wang@intel.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).