Re: [PATCH 1/2] iommu/vt-d: Use rbtree to track iommu probed devices

[Baolu Lu]

On 2024/2/16 1:47, Jason Gunthorpe wrote:
> On Thu, Feb 15, 2024 at 03:22:48PM +0800, Lu Baolu wrote:
> > Use a red-black tree(rbtree) to track devices probed by the driver's
> > probe_device callback. These devices need to be looked up quickly by
> > a source ID when the hardware reports a fault, either recoverable or
> > unrecoverable.
> >
> > Fault reporting paths are critical. Searching a list in this scenario
> > is inefficient, with an algorithm complexity of O(n). An rbtree is a
> > self-balancing binary search tree, offering an average search time
> > complexity of O(log(n)). This significant performance improvement
> > makes rbtrees a better choice.
> >
> > Furthermore, rbtrees are implemented on a per-iommu basis, eliminating
> > the need for global searches and further enhancing efficiency in
> > critical fault paths. The rbtree is protected by a spin lock with
> > interrupts disabled to ensure thread-safe access even within interrupt
> > contexts.
> >
> > Co-developed-by: Huang Jiaqing<jiaqing.huang@xxxxxxxxx>
> > Signed-off-by: Huang Jiaqing<jiaqing.huang@xxxxxxxxx>
> > Signed-off-by: Lu Baolu<baolu.lu@xxxxxxxxxxxxxxx>
> > ---
> >   drivers/iommu/intel/iommu.h |  7 +++++
> >   drivers/iommu/intel/dmar.c  |  3 +-
> >   drivers/iommu/intel/iommu.c | 62 +++++++++++++++++++++++++++++++++++--
> >   3 files changed, 69 insertions(+), 3 deletions(-)
> Reviewed-by: Jason Gunthorpe<jgg@xxxxxxxxxx>
>
> > +static int device_rbtree_insert(struct intel_iommu *iommu,
> > +				struct device_domain_info *info)
> > +{
> > +	struct rb_node *curr;
> > +	unsigned long flags;
> > +
> > +	spin_lock_irqsave(&iommu->device_rbtree_lock, flags);
> > +	curr = rb_find_add(&info->node, &iommu->device_rbtree, device_rid_cmp);
> > +	spin_unlock_irqrestore(&iommu->device_rbtree_lock, flags);
> > +	if (curr)
> > +		dev_warn(info->dev, "device already in rbtree\n");
> I would suggest
>
> WARN_ON(curr);
>
> Something has gone really wonky at this point, right?

Yes. This is not expected, and it is worth a WARN_ON().

Best regards,
baolu