Re: [PATCH v11 10/22] vfio iommu type1: Add support for mediated devices

[Alex Williamson <alex.williamson@redhat.com>]

On Tue, 08 Nov 2016 10:20:14 +0800
Jike Song <jike.song@intel.com> wrote:

> On 11/08/2016 07:16 AM, Alex Williamson wrote:
> > On Sat, 5 Nov 2016 02:40:44 +0530
> > Kirti Wankhede <kwankhede@nvidia.com> wrote:
> >   
> >> VFIO IOMMU drivers are designed for the devices which are IOMMU capable.
> >> Mediated device only uses IOMMU APIs, the underlying hardware can be
> >> managed by an IOMMU domain.
> >>
> >> Aim of this change is:
> >> - To use most of the code of TYPE1 IOMMU driver for mediated devices
> >> - To support direct assigned device and mediated device in single module
> >>
> >> This change adds pin and unpin support for mediated device to TYPE1 IOMMU
> >> backend module. More details:
> >> - vfio_pin_pages() callback here uses task and address space of vfio_dma,
> >>   that is, of the process who mapped that iova range.
> >> - Added pfn_list tracking logic to address space structure. All pages
> >>   pinned through this interface are trached in its address space.  
> >                                           ^ k
> > ------------------------------------------|
> >   
> >> - Pinned pages list is used to verify unpinning request and to unpin
> >>   remaining pages while detaching the group for that device.
> >> - Page accounting is updated to account in its address space where the
> >>   pages are pinned/unpinned.
> >> -  Accouting for mdev device is only done if there is no iommu capable
> >>   domain in the container. When there is a direct device assigned to the
> >>   container and that domain is iommu capable, all pages are already pinned
> >>   during DMA_MAP.
> >> - Page accouting is updated on hot plug and unplug mdev device and pass
> >>   through device.
> >>
> >> Tested by assigning below combinations of devices to a single VM:
> >> - GPU pass through only
> >> - vGPU device only
> >> - One GPU pass through and one vGPU device
> >> - Linux VM hot plug and unplug vGPU device while GPU pass through device
> >>   exist
> >> - Linux VM hot plug and unplug GPU pass through device while vGPU device
> >>   exist
> >>
> >> Signed-off-by: Kirti Wankhede <kwankhede@nvidia.com>
> >> Signed-off-by: Neo Jia <cjia@nvidia.com>
> >> Change-Id: I295d6f0f2e0579b8d9882bfd8fd5a4194b97bd9a
> >> ---
> >>  drivers/vfio/vfio_iommu_type1.c | 538 +++++++++++++++++++++++++++++++++++++---
> >>  1 file changed, 500 insertions(+), 38 deletions(-)
> >>
> >> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> >> index 8d64528dcc22..e511073446a0 100644
> >> --- a/drivers/vfio/vfio_iommu_type1.c
> >> +++ b/drivers/vfio/vfio_iommu_type1.c
> >> @@ -36,6 +36,7 @@
> >>  #include <linux/uaccess.h>
> >>  #include <linux/vfio.h>
> >>  #include <linux/workqueue.h>
> >> +#include <linux/mdev.h>
> >>  
> >>  #define DRIVER_VERSION  "0.2"
> >>  #define DRIVER_AUTHOR   "Alex Williamson <alex.williamson@redhat.com>"
> >> @@ -56,6 +57,7 @@ MODULE_PARM_DESC(disable_hugepages,
> >>  struct vfio_iommu {
> >>  	struct list_head	domain_list;
> >>  	struct list_head	addr_space_list;
> >> +	struct vfio_domain	*external_domain; /* domain for external user */
> >>  	struct mutex		lock;
> >>  	struct rb_root		dma_list;
> >>  	bool			v2;
> >> @@ -67,6 +69,9 @@ struct vfio_addr_space {
> >>  	struct mm_struct	*mm;
> >>  	struct list_head	next;
> >>  	atomic_t		ref_count;
> >> +	/* external user pinned pfns */
> >> +	struct rb_root		pfn_list;	/* pinned Host pfn list */
> >> +	struct mutex		pfn_list_lock;	/* mutex for pfn_list */
> >>  };
> >>  
> >>  struct vfio_domain {
> >> @@ -83,6 +88,7 @@ struct vfio_dma {
> >>  	unsigned long		vaddr;		/* Process virtual addr */
> >>  	size_t			size;		/* Map size (bytes) */
> >>  	int			prot;		/* IOMMU_READ/WRITE */
> >> +	bool			iommu_mapped;
> >>  	struct vfio_addr_space	*addr_space;
> >>  	struct task_struct	*task;
> >>  	bool			mlock_cap;
> >> @@ -94,6 +100,19 @@ struct vfio_group {
> >>  };
> >>  
> >>  /*
> >> + * Guest RAM pinning working set or DMA target
> >> + */
> >> +struct vfio_pfn {
> >> +	struct rb_node		node;
> >> +	unsigned long		pfn;		/* Host pfn */
> >> +	int			prot;
> >> +	atomic_t		ref_count;
> >> +};
> >> +
> >> +#define IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu)	\
> >> +					(!list_empty(&iommu->domain_list))
> >> +
> >> +/*
> >>   * This code handles mapping and unmapping of user data buffers
> >>   * into DMA'ble space using the IOMMU
> >>   */
> >> @@ -153,6 +172,93 @@ static struct vfio_addr_space *vfio_find_addr_space(struct vfio_iommu *iommu,
> >>  	return NULL;
> >>  }
> >>  
> >> +/*
> >> + * Helper Functions for host pfn list
> >> + */
> >> +static struct vfio_pfn *vfio_find_pfn(struct vfio_addr_space *addr_space,
> >> +				      unsigned long pfn)
> >> +{
> >> +	struct vfio_pfn *vpfn;
> >> +	struct rb_node *node = addr_space->pfn_list.rb_node;
> >> +
> >> +	while (node) {
> >> +		vpfn = rb_entry(node, struct vfio_pfn, node);
> >> +
> >> +		if (pfn < vpfn->pfn)
> >> +			node = node->rb_left;
> >> +		else if (pfn > vpfn->pfn)
> >> +			node = node->rb_right;
> >> +		else
> >> +			return vpfn;
> >> +	}
> >> +
> >> +	return NULL;
> >> +}
> >> +
> >> +static void vfio_link_pfn(struct vfio_addr_space *addr_space,
> >> +			  struct vfio_pfn *new)
> >> +{
> >> +	struct rb_node **link, *parent = NULL;
> >> +	struct vfio_pfn *vpfn;
> >> +
> >> +	link = &addr_space->pfn_list.rb_node;
> >> +	while (*link) {
> >> +		parent = *link;
> >> +		vpfn = rb_entry(parent, struct vfio_pfn, node);
> >> +
> >> +		if (new->pfn < vpfn->pfn)
> >> +			link = &(*link)->rb_left;
> >> +		else
> >> +			link = &(*link)->rb_right;
> >> +	}
> >> +
> >> +	rb_link_node(&new->node, parent, link);
> >> +	rb_insert_color(&new->node, &addr_space->pfn_list);
> >> +}
> >> +
> >> +static void vfio_unlink_pfn(struct vfio_addr_space *addr_space,
> >> +			    struct vfio_pfn *old)
> >> +{
> >> +	rb_erase(&old->node, &addr_space->pfn_list);
> >> +}
> >> +
> >> +static int vfio_add_to_pfn_list(struct vfio_addr_space *addr_space,
> >> +				unsigned long pfn, int prot)
> >> +{
> >> +	struct vfio_pfn *vpfn;
> >> +
> >> +	vpfn = kzalloc(sizeof(*vpfn), GFP_KERNEL);
> >> +	if (!vpfn)
> >> +		return -ENOMEM;
> >> +
> >> +	vpfn->pfn = pfn;
> >> +	vpfn->prot = prot;
> >> +	atomic_set(&vpfn->ref_count, 1);
> >> +	vfio_link_pfn(addr_space, vpfn);
> >> +	return 0;
> >> +}
> >> +
> >> +static void vfio_remove_from_pfn_list(struct vfio_addr_space *addr_space,
> >> +				      struct vfio_pfn *vpfn)
> >> +{
> >> +	vfio_unlink_pfn(addr_space, vpfn);
> >> +	kfree(vpfn);
> >> +}
> >> +
> >> +static int vfio_pfn_account(struct vfio_addr_space *addr_space,
> >> +			    unsigned long pfn)
> >> +{
> >> +	struct vfio_pfn *p;
> >> +	int ret = 1;
> >> +
> >> +	mutex_lock(&addr_space->pfn_list_lock);
> >> +	p = vfio_find_pfn(addr_space, pfn);
> >> +	if (p)
> >> +		ret = 0;
> >> +	mutex_unlock(&addr_space->pfn_list_lock);
> >> +	return ret;
> >> +}
> >> +
> >>  struct vwork {
> >>  	struct mm_struct	*mm;
> >>  	long			npage;
> >> @@ -304,16 +410,18 @@ static long __vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
> >>  	unsigned long limit = task_rlimit(task, RLIMIT_MEMLOCK) >> PAGE_SHIFT;
> >>  	bool lock_cap = dma->mlock_cap;
> >>  	struct mm_struct *mm = dma->addr_space->mm;
> >> -	long ret, i;
> >> +	long ret, i, lock_acct;
> >>  	bool rsvd;
> >>  
> >>  	ret = vaddr_get_pfn(mm, vaddr, prot, pfn_base);
> >>  	if (ret)
> >>  		return ret;
> >>  
> >> +	lock_acct = vfio_pfn_account(dma->addr_space, *pfn_base);
> >> +
> >>  	rsvd = is_invalid_reserved_pfn(*pfn_base);
> >>  
> >> -	if (!rsvd && !lock_cap && mm->locked_vm + 1 > limit) {
> >> +	if (!rsvd && !lock_cap && mm->locked_vm + lock_acct > limit) {
> >>  		put_pfn(*pfn_base, prot);
> >>  		pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n", __func__,
> >>  			limit << PAGE_SHIFT);
> >> @@ -340,8 +448,10 @@ static long __vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
> >>  			break;
> >>  		}
> >>  
> >> +		lock_acct += vfio_pfn_account(dma->addr_space, pfn);
> >> +
> >>  		if (!rsvd && !lock_cap &&
> >> -		    mm->locked_vm + i + 1 > limit) {
> >> +		    mm->locked_vm + lock_acct + 1 > limit) {
> >>  			put_pfn(pfn, prot);
> >>  			pr_warn("%s: RLIMIT_MEMLOCK (%ld) exceeded\n",
> >>  				__func__, limit << PAGE_SHIFT);
> >> @@ -350,7 +460,7 @@ static long __vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
> >>  	}
> >>  
> >>  	if (!rsvd)
> >> -		vfio_lock_acct(mm, i);
> >> +		vfio_lock_acct(mm, lock_acct);
> >>  
> >>  	return i;
> >>  }
> >> @@ -370,14 +480,214 @@ static long __vfio_unpin_pages_remote(struct vfio_dma *dma, unsigned long pfn,
> >>  	return unlocked;
> >>  }
> >>  
> >> -static void vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma)
> >> +static int __vfio_pin_page_external(struct vfio_dma *dma, unsigned long vaddr,
> >> +				    int prot, unsigned long *pfn_base,
> >> +				    bool do_accounting)
> >> +{
> >> +	struct task_struct *task = dma->task;
> >> +	unsigned long limit = task_rlimit(task, RLIMIT_MEMLOCK) >> PAGE_SHIFT;
> >> +	bool lock_cap = dma->mlock_cap;
> >> +	struct mm_struct *mm = dma->addr_space->mm;
> >> +	int ret;
> >> +	bool rsvd;
> >> +
> >> +	ret = vaddr_get_pfn(mm, vaddr, prot, pfn_base);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	rsvd = is_invalid_reserved_pfn(*pfn_base);
> >> +
> >> +	if (!rsvd && !lock_cap && mm->locked_vm + 1 > limit) {
> >> +		put_pfn(*pfn_base, prot);
> >> +		pr_warn("%s: Task %s (%d) RLIMIT_MEMLOCK (%ld) exceeded\n",
> >> +			__func__, task->comm, task_pid_nr(task),
> >> +			limit << PAGE_SHIFT);
> >> +		return -ENOMEM;
> >> +	}
> >> +
> >> +	if (!rsvd && do_accounting)
> >> +		vfio_lock_acct(mm, 1);
> >> +
> >> +	return 1;
> >> +}
> >> +
> >> +static void __vfio_unpin_page_external(struct vfio_addr_space *addr_space,
> >> +				       unsigned long pfn, int prot,
> >> +				       bool do_accounting)
> >> +{
> >> +	put_pfn(pfn, prot);
> >> +
> >> +	if (do_accounting)
> >> +		vfio_lock_acct(addr_space->mm, -1);  
> > 
> > Can't we batch this like we do elsewhere?  Intel folks, AIUI you intend
> > to pin all VM memory through this side channel, have you tested the
> > scalability and performance of this with larger VMs?  Our vfio_pfn
> > data structure alone is 40 bytes per pinned page, which means for
> > each 1GB of VM memory, we have 10MBs worth of struct vfio_pfn!
> > Additionally, unmapping each 1GB of VM memory will result in 256k
> > separate vfio_lock_acct() callbacks.  I'm concerned that we're not
> > being efficient enough in either space or time.  
> 
> Hi Alex,
> 
> Sorry for being confusing, Intel vGPU actually doesn't necessarily need
> to pin all guest memory. A vGPU has its page table (GTT), whose access
> is trapped. Whenever guest driver wants to specify a page for DMA, it
> writes the GTT entry - thereby we could know the event and pin that
> page only.
> 
> Performance data will be shared once available. Thanks :)

Ok, so maybe if none of the initial users are mapping full VM memory
then scalability can be saved for future optimization.  Thanks,

Alex