linux-pci.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Alex Williamson <alex.williamson@redhat.com>
To: ethan zhao <ethan.zhao@oracle.com>
Cc: bhelgaas@google.com, konrad.wilk@oracle.com,
	boris.ostrovsky@oracle.com, david.vrabel@citrix.com,
	gleb@kernel.org, pbonzini@redhat.com,
	jeffrey.t.kirsher@intel.com, jesse.brandeburg@intel.com,
	bruce.w.allan@intel.com, carolyn.wyborny@intel.com,
	donald.c.skidmore@intel.com, gregory.v.rose@intel.com,
	alexander.h.duyck@intel.com, john.ronciak@intel.com,
	mitch.a.williams@intel.com, linux-pci@vger.kernel.org,
	kvm@vger.kernel.org, linux.nics@intel.com,
	e1000-devel@lists.sourceforge.net, netdev@vger.kernel.org,
	linux-kernel@vger.kernel.org, ethan.kernel@gmail.com,
	vaughan.cao@oracle.com
Subject: Re: [PATCH 1/4] PCI: introduce VFs reference counter and simplify pci_vfs_assigned() implementation
Date: Thu, 10 Jul 2014 20:22:04 -0600	[thread overview]
Message-ID: <1405045324.4098.50.camel@ul30vt.home> (raw)
In-Reply-To: <53BF479A.5030006@oracle.com>

On Fri, 2014-07-11 at 10:10 +0800, ethan zhao wrote:
> Alex,
>     Thanks for your reviewing, when I create the patch order, I thought 
> about the question you concerned for
> quit a while, make every patch be independent to each other as possible 
> as I could, so we can do bisect when hit
> problem.
> 
>     I manage to take more time to figure out better patch order.
> 
>     Thanks,
>     Ethan
> 
> On 2014/7/11 9:48, Alex Williamson wrote:
> > Since there's no 0th patch, I guess I'll comment here.  This series is
> > not bisectable, patch 1 breaks the existing implementation.  I'd
> > suggest:
> >
> > patch 1 - fix i40e
>              i40e only could be fixed with new interface, so it couldn't 
> be the first one.

It looks like i40e just has it's own copy of pci_vfs_assigned(), why
can't your current patch 4/4 be applied now?

> > patch 2 - create assign/deassign that uses dev_flags
>              This will be the first ?
> > patch 3 - convert users to new interface
>              Have to be the later step.
> > patch 4 - convert interface to use atomic_t
>              Could it be standalone step ?
> 
>   Let me think about it.
> 
> > IMHO, the assigned flag is a horrible hack and I don't know why drivers
> > like xenback need to use it.  KVM needs to use it because it doesn't
> > actually have a driver to bind to when a device is assigned, it's happy
> > to assign devices without any driver.  Thanks,
> >
> > Alex
> >
> >
> > On Fri, 2014-07-11 at 08:47 +0800, Ethan Zhao wrote:
> >> Current implementation of helper function pci_vfs_assigned() is a little complex, to
> >> get sum of VFs that assigned to VM, access low level configuration space register and
> >> then loop in traversing device tree.
> >>
> >> This patch introduce an atomic reference counter for VFs that assigned to VM in struct
> >> pci_sriov, and compose two more helper functions
> >>
> >> pci_sriov_assign_device(),
> >> pci_sriov_deassign_device()
> >>
> >> to replace manipulation to device flag and the meanwhile increase and decease the counter.
> >>
> >> Passed building on 3.15.5
> >>
> >> Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
> >> ---
> >>   drivers/pci/iov.c   |   65 ++++++++++++++++++++++++++++----------------------
> >>   drivers/pci/pci.h   |    1 +
> >>   include/linux/pci.h |    4 +++
> >>   3 files changed, 41 insertions(+), 29 deletions(-)
> >>
> >> diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
> >> index de7a747..72e267f 100644
> >> --- a/drivers/pci/iov.c
> >> +++ b/drivers/pci/iov.c
> >> @@ -382,6 +382,7 @@ found:
> >>   	iov->nres = nres;
> >>   	iov->ctrl = ctrl;
> >>   	iov->total_VFs = total;
> >> +	atomic_set(&iov->VFs_assigned_cnt, 0);
> >>   	iov->offset = offset;
> >>   	iov->stride = stride;
> >>   	iov->pgsz = pgsz;
> >> @@ -603,45 +604,51 @@ int pci_num_vf(struct pci_dev *dev)
> >>   EXPORT_SYMBOL_GPL(pci_num_vf);
> >>   
> >>   /**
> >> - * pci_vfs_assigned - returns number of VFs are assigned to a guest
> >> - * @dev: the PCI device
> >> + * pci_vfs_assigned - returns number of VFs are assigned to VM
> >> + * @dev: the physical PCI device that contains the VFs.
> >>    *
> >> - * Returns number of VFs belonging to this device that are assigned to a guest.
> >> + * Returns number of VFs belonging to this device that are assigned to VM.
> >>    * If device is not a physical function returns 0.
> >>    */
> >>   int pci_vfs_assigned(struct pci_dev *dev)
> >>   {
> >> -	struct pci_dev *vfdev;
> >> -	unsigned int vfs_assigned = 0;
> >> -	unsigned short dev_id;
> >> -
> >> -	/* only search if we are a PF */
> >>   	if (!dev->is_physfn)
> >>   		return 0;
> >> +	if (dev->sriov)
> >> +		return atomic_read(&dev->sriov->VFs_assigned_cnt);
> >> +	return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(pci_vfs_assigned);
> >>   
> >> -	/*
> >> -	 * determine the device ID for the VFs, the vendor ID will be the
> >> -	 * same as the PF so there is no need to check for that one
> >> -	 */
> >> -	pci_read_config_word(dev, dev->sriov->pos + PCI_SRIOV_VF_DID, &dev_id);
> >> -
> >> -	/* loop through all the VFs to see if we own any that are assigned */
> >> -	vfdev = pci_get_device(dev->vendor, dev_id, NULL);
> >> -	while (vfdev) {
> >> -		/*
> >> -		 * It is considered assigned if it is a virtual function with
> >> -		 * our dev as the physical function and the assigned bit is set
> >> -		 */
> >> -		if (vfdev->is_virtfn && (vfdev->physfn == dev) &&
> >> -		    (vfdev->dev_flags & PCI_DEV_FLAGS_ASSIGNED))
> >> -			vfs_assigned++;
> >> -
> >> -		vfdev = pci_get_device(dev->vendor, dev_id, vfdev);
> >> -	}
> >> +/**
> >> + * pci_sriov_assign_device - assign device to VM
> >> + * @pdev: the device to be assigned.
> >> + */
> >> +void pci_sriov_assign_device(struct pci_dev *pdev)
> >> +{
> >> +	pdev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
> >> +	if (pdev->is_virtfn && !pdev->is_physfn)
> >> +		if (pdev->physfn)
> >> +			if (pdev->physfn->sriov)
> >> +				atomic_inc(&pdev->physfn->sriov->
> >> +						VFs_assigned_cnt);
> >> +}
> >> +EXPORT_SYMBOL_GPL(pci_sriov_assign_device);
> >>   
> >> -	return vfs_assigned;
> >> +/**
> >> + * pci_sriov_deassign_device - deasign device from VM
> >> + * @pdev: the device to be deassigned.
> >> + */
> >> +void pci_sriov_deassign_device(struct pci_dev *pdev)
> >> +{
> >> +	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
> >> +	if (pdev->is_virtfn && !pdev->is_physfn)
> >> +		if (pdev->physfn)
> >> +			if (pdev->physfn->sriov)
> >> +				atomic_dec(&pdev->physfn->sriov->
> >> +						VFs_assigned_cnt);
> >>   }
> >> -EXPORT_SYMBOL_GPL(pci_vfs_assigned);
> >> +EXPORT_SYMBOL_GPL(pci_sriov_deassign_device);
> >>   
> >>   /**
> >>    * pci_sriov_set_totalvfs -- reduce the TotalVFs available
> >> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> >> index 6bd0822..d17bda2 100644
> >> --- a/drivers/pci/pci.h
> >> +++ b/drivers/pci/pci.h
> >> @@ -235,6 +235,7 @@ struct pci_sriov {
> >>   	u32 pgsz;		/* page size for BAR alignment */
> >>   	u8 link;		/* Function Dependency Link */
> >>   	u16 driver_max_VFs;	/* max num VFs driver supports */
> >> +	atomic_t VFs_assigned_cnt; /* counter of VFs assigned to VM */
> >>   	struct pci_dev *dev;	/* lowest numbered PF */
> >>   	struct pci_dev *self;	/* this PF */
> >>   	struct mutex lock;	/* lock for VF bus */
> >> diff --git a/include/linux/pci.h b/include/linux/pci.h
> >> index aab57b4..5cf6833 100644
> >> --- a/include/linux/pci.h
> >> +++ b/include/linux/pci.h
> >> @@ -1601,6 +1601,8 @@ int pci_enable_sriov(struct pci_dev *dev, int nr_virtfn);
> >>   void pci_disable_sriov(struct pci_dev *dev);
> >>   int pci_num_vf(struct pci_dev *dev);
> >>   int pci_vfs_assigned(struct pci_dev *dev);
> >> +void pci_sriov_assign_device(struct pci_dev *dev);
> >> +void pci_sriov_deassign_device(struct pci_dev *dev);
> >>   int pci_sriov_set_totalvfs(struct pci_dev *dev, u16 numvfs);
> >>   int pci_sriov_get_totalvfs(struct pci_dev *dev);
> >>   #else
> >> @@ -1610,6 +1612,8 @@ static inline void pci_disable_sriov(struct pci_dev *dev) { }
> >>   static inline int pci_num_vf(struct pci_dev *dev) { return 0; }
> >>   static inline int pci_vfs_assigned(struct pci_dev *dev)
> >>   { return 0; }
> >> +static inline void pci_sriov_assign_device(struct pci_dev *dev) { }
> >> +static inline void pci_sriov_deassign_device(struct pci_dev *dev) { }
> >>   static inline int pci_sriov_set_totalvfs(struct pci_dev *dev, u16 numvfs)
> >>   { return 0; }
> >>   static inline int pci_sriov_get_totalvfs(struct pci_dev *dev)
> >
> >
> 




  reply	other threads:[~2014-07-11  2:22 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-07-11  0:47 [PATCH 1/4] PCI: introduce VFs reference counter and simplify pci_vfs_assigned() implementation Ethan Zhao
2014-07-11  0:47 ` [PATCH 2/4] xen-pciback: use PCI VFs assignment helper functions Ethan Zhao
2014-07-11  0:47 ` [PATCH 3/4] KVM: " Ethan Zhao
2014-07-11  0:47 ` [PATCH 4/4] i40e: use PCI VFs assignment helper function simplify i40e_vfs_are_assigned() Ethan Zhao
2014-07-11  1:48 ` [PATCH 1/4] PCI: introduce VFs reference counter and simplify pci_vfs_assigned() implementation Alex Williamson
2014-07-11  2:10   ` ethan zhao
2014-07-11  2:22     ` Alex Williamson [this message]
2014-07-11  2:29       ` ethan zhao
2014-07-11  2:33         ` Alex Williamson
2014-07-11  3:11           ` ethan zhao

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=1405045324.4098.50.camel@ul30vt.home \
    --to=alex.williamson@redhat.com \
    --cc=alexander.h.duyck@intel.com \
    --cc=bhelgaas@google.com \
    --cc=boris.ostrovsky@oracle.com \
    --cc=bruce.w.allan@intel.com \
    --cc=carolyn.wyborny@intel.com \
    --cc=david.vrabel@citrix.com \
    --cc=donald.c.skidmore@intel.com \
    --cc=e1000-devel@lists.sourceforge.net \
    --cc=ethan.kernel@gmail.com \
    --cc=ethan.zhao@oracle.com \
    --cc=gleb@kernel.org \
    --cc=gregory.v.rose@intel.com \
    --cc=jeffrey.t.kirsher@intel.com \
    --cc=jesse.brandeburg@intel.com \
    --cc=john.ronciak@intel.com \
    --cc=konrad.wilk@oracle.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=linux.nics@intel.com \
    --cc=mitch.a.williams@intel.com \
    --cc=netdev@vger.kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=vaughan.cao@oracle.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).