Re: Affinity managed interrupts vs non-managed interrupts

[Ming Lei]

On Fri, Aug 31, 2018 at 01:50:31AM -0600, Kashyap Desai wrote:
> > -----Original Message-----
> > From: Ming Lei [mailto:tom.leiming@xxxxxxxxx]
> > Sent: Friday, August 31, 2018 12:54 AM
> > To: sumit.saxena@xxxxxxxxxxxx
> > Cc: Ming Lei; Thomas Gleixner; Christoph Hellwig; Linux Kernel Mailing
> > List;
> > Kashyap Desai; shivasharan.srikanteshwara@xxxxxxxxxxxx; linux-block
> > Subject: Re: Affinity managed interrupts vs non-managed interrupts
> >
> > On Wed, Aug 29, 2018 at 6:47 PM Sumit Saxena
> > <sumit.saxena@xxxxxxxxxxxx> wrote:
> > >
> > > > -----Original Message-----
> > > > From: Ming Lei [mailto:ming.lei@xxxxxxxxxx]
> > > > Sent: Wednesday, August 29, 2018 2:16 PM
> > > > To: Sumit Saxena <sumit.saxena@xxxxxxxxxxxx>
> > > > Cc: tglx@xxxxxxxxxxxxx; hch@xxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> > > > Subject: Re: Affinity managed interrupts vs non-managed interrupts
> > > >
> > > > Hello Sumit,
> > > Hi Ming,
> > > Thanks for response.
> > > >
> > > > On Tue, Aug 28, 2018 at 12:04:52PM +0530, Sumit Saxena wrote:
> > > > >  Affinity managed interrupts vs non-managed interrupts
> > > > >
> > > > > Hi Thomas,
> > > > >
> > > > > We are working on next generation MegaRAID product where
> > requirement
> > > > > is- to allocate additional 16 MSI-x vectors in addition to number of
> > > > > MSI-x vectors megaraid_sas driver usually allocates.  MegaRAID
> > > > > adapter
> > > > > supports 128 MSI-x vectors.
> > > > >
> > > > > To explain the requirement and solution, consider that we have 2
> > > > > socket system (each socket having 36 logical CPUs). Current driver
> > > > > will allocate total 72 MSI-x vectors by calling API-
> > > > > pci_alloc_irq_vectors(with flag- PCI_IRQ_AFFINITY).  All 72 MSI-x
> > > > > vectors will have affinity across NUMA node s and interrupts are
> > > affinity
> > > > managed.
> > > > >
> > > > > If driver calls- pci_alloc_irq_vectors_affinity() with pre_vectors =
> > > > > 16 and, driver can allocate 16 + 72 MSI-x vectors.
> > > >
> > > > Could you explain a bit what the specific use case the extra 16
> > > > vectors
> > > is?
> > > We are trying to avoid the penalty due to one interrupt per IO
> > > completion
> > > and decided to coalesce interrupts on these extra 16 reply queues.
> > > For regular 72 reply queues, we will not coalesce interrupts as for low
> > > IO
> > > workload, interrupt coalescing may take more time due to less IO
> > > completions.
> > > In IO submission path, driver will decide which set of reply queues
> > > (either extra 16 reply queues or regular 72 reply queues) to be picked
> > > based on IO workload.
> >
> > I am just wondering how you can make the decision about using extra
> > 16 or regular 72 queues in submission path, could you share us a bit
> > your idea? How are you going to recognize the IO workload inside your
> > driver? Even the current block layer doesn't recognize IO workload, such
> > as random IO or sequential IO.
> 
> It is not yet finalized, but it can be based on per sdev outstanding,
> shost_busy etc.
> We want to use special 16 reply queue for IO acceleration (these queues are
> working interrupt coalescing mode. This is a h/w feature)

This part is very key to your approach, so I'd suggest to finalize it
first. That said this way doesn't make sense if you can't figure out
one doable approach to decide when to use the coalescing mode, and when to
use the regular 72 reply queues.

If it is just for IO acceleration, why not always use the coalescing mode?

> 
> >
> > Frankly speaking, you may reuse the 72 reply queues to do interrupt
> > coalescing by configuring one extra register to enable the coalescing
> > mode,
> > and you may just use small part of the 72 reply queues under the
> > interrupt coalescing mode.
> Our h/w can set interrupt coalescing per 8 reply queues. So smallest is 8.
> If we choose to take 8 reply queue from existing 72 reply queue (without
> asking for extra reply queue), we still have  an issue on more numa node
> systems.  Example - in 8 numa node system each node will have only *one*
> reply queue for effective interrupt coalescing. (since irq subsystem will
> spread msix per numa).
> 
> To keep things scalable we cherry picked few reply queues and wanted them to
> be out of cpu-msix mapping.

I mean you can group the reply queues according to the queue's numa node
info, given the mapping has been figured out there by genirq affinity
code.

> 
> >
> > Or you can learn from SPDK to use one or small number of dedicated cores
> > or kernel threads to poll the interrupts from all reply queues, then I
> > guess you may benefit much compared with the extra 16 queue approach.
> Problem with polling -  It requires some steady completion, otherwise
> prediction in driver gives different results on different profiles.
> We attempted irq-poll and thread ISR based polling, but it has pros and
> cons. One of the key usage of method what we are trying is not to impact
> latency for lower QD workloads.

Interrupt coalescing should effect latency too[1], or could you share your
idea how to use interrupt coalescing to address the latency issue?

	"Interrupt coalescing, also known as interrupt moderation,[1] is a
	technique in which events which would normally trigger a hardware interrupt
	are held back, either until a certain amount of work is pending, or a
	timeout timer triggers."[1] 

[1] https://en.wikipedia.org/wiki/Interrupt_coalescing

> I posted RFC at
> https://www.spinics.net/lists/linux-scsi/msg122874.html
> 
> We have done extensive study and concluded to use interrupt coalescing is
> better if h/w can manage two different modes (coalescing on/off).

Could you explain a bit why coalescing is better?

In theory, interrupt coalescing is just to move the implementation into
hardware. And the IO submitted from the same coalescing group is usually
irrelevant. The same problem you found in polling should have been in
coalescing too.

> 
> >
> > Introducing extra 16 queues just for interrupt coalescing and making it
> > coexisting with the regular 72 reply queues seems one very unusual use
> > case, not sure the current genirq affinity can support it well.
> 
> Yes. This is unusual case. I think it is not used by any other drivers.
> 
> >
> > > >
> > > > >
> > > > > All pre_vectors (16) will be mapped to all available online CPUs but
> > > > > e
> > > > > ffective affinity of each vector is to CPU 0. Our requirement is to
> > > > > have pre _vectors 16 reply queues to be mapped to local NUMA node
> > with
> > > > > effective CPU should be spread within local node cpu mask. Without
> > > > > changing kernel code, we can
> > > >
> > > > If all CPUs in one NUMA node is offline, can this use case work as
> > > expected?
> > > > Seems we have to understand what the use case is and how it works.
> > >
> > > Yes, if all CPUs of the NUMA node is offlined, IRQ-CPU affinity will be
> > > broken and irqbalancer takes care of migrating affected IRQs to online
> > > CPUs of different NUMA node.
> > > When offline CPUs are onlined again, irqbalancer restores affinity.
> >
> >  irqbalance daemon can't cover managed interrupts, or you mean
> > you don't use pci_alloc_irq_vectors_affinity(PCI_IRQ_AFFINITY)?
> 
> Yes. We did not used " pci_alloc_irq_vectors_affinity".
> We used " pci_enable_msix_range" and manually set affinity in driver using
> irq_set_affinity_hint.

Then you have to cover all kind of CPU hotplug issues in your driver
because you switch to driver to maintain the queue mapping.

Thanks,
Ming