[RFC 0/7] Introduce swiotlb throttling

[Petr Tesařík]

On Fri, 23 Aug 2024 20:40:16 +0000
Michael Kelley <mhklinux at outlook.com> wrote:

> From: Petr Tesařík <petr at tesarici.cz> Sent: Thursday, August 22, 2024 11:45 PM
>[...]
> > > Discussion
> > > ==========
> > > * Since swiotlb isn't visible to device drivers, I've specifically
> > > named the DMA attribute as MAY_BLOCK instead of MAY_THROTTLE or
> > > something swiotlb specific. While this patch set consumes MAY_BLOCK
> > > only on the DMA direct path to do throttling in the swiotlb code,
> > > there might be other uses in the future outside of CoCo VMs, or
> > > perhaps on the IOMMU path.  
> > 
> > I once introduced a similar flag and called it MAY_SLEEP. I chose
> > MAY_SLEEP, because there is already a might_sleep() annotation, but I
> > don't have a strong opinion unless your semantics is supposed to be
> > different from might_sleep(). If it is, then I strongly prefer
> > MAY_BLOCK to prevent confusing the two.  
> 
> My intent is that the semantics are the same as might_sleep(). I
> vacillated between MAY_SLEEP and MAY_BLOCK. The kernel seems
> to treat "sleep" and "block" as equivalent, because blk-mq has
> the BLK_MQ_F_BLOCKING flag, and SCSI has the 
> queuecommand_may_block flag that is translated to
> BLK_MQ_F_BLOCKING. So I settled on MAY_BLOCK, but as you
> point out, that's inconsistent with might_sleep(). Either way will
> be inconsistent somewhere, and I don't have a preference.

Fair enough. Let's stay with MAY_BLOCK then, so you don't have to
change it everywhere.

>[...]
> > > Open Topics
> > > ===========
> > > 1. swiotlb allocations from Xen and the IOMMU code don't make use
> > > of throttling. This could be added if beneficial.
> > >
> > > 2. The throttling values are currently exposed and adjustable in
> > > /sys/kernel/debug/swiotlb. Should any of this be moved so it is
> > > visible even without CONFIG_DEBUG_FS?  
> > 
> > Yes. It should be possible to control the thresholds through
> > sysctl.  
> 
> Good point.  I was thinking about creating /sys/kernel/swiotlb, but
> sysctl is better.

That still leaves the question where it should go.

Under /proc/sys/kernel? Or should we make a /proc/sys/kernel/dma
subdirectory to make room for more dma-related controls?

Petr T