[Intel-gfx] [RFC 7/8] drm/i915: Skip CSB processing on invalid CSB tail

[Chris Wilson]

Quoting Jeff McGee (2018-03-21 17:31:45)
> On Wed, Mar 21, 2018 at 10:26:24AM -0700, jeff.mcgee at intel.com wrote:
> > From: Jeff McGee <jeff.mcgee at intel.com>
> > 
> > Engine reset is fast. A context switch interrupt may be generated just
> > prior to the reset such that the top half handler is racing with reset
> > post-processing. The handler may set the irq_posted bit again after
> > the reset code has cleared it to start fresh. Then the re-enabled
> > tasklet will read the CSB head and tail from MMIO, which will be at
> > the hardware reset values of 0 and 7 respectively, given that no
> > actual CSB event has occurred since the reset. Mayhem then ensues as
> > the tasklet starts processing invalid CSB entries.
> > 
> > We can handle this corner case without adding any new synchronization
> > between the irq top half and the reset work item. The tasklet can
> > just skip CSB processing if the tail is not sane.
> > 
> > Signed-off-by: Jeff McGee <jeff.mcgee at intel.com>
> > ---
> If I drop this patch and substitute https://patchwork.freedesktop.org/patch/211831/
> I will see irq_posted get set after reset which causes the first tasklet
> run to re-process a previous CSB event and hit GEM_BUG_ON that nothing
> was active.

However, for reset+sync to be followed by an interrupt is surprising.
What more do we need to do after the reset to flush the last interrupt?

What I've been trying to get right is disabling the RING_IMR across the
reset so that we do not get any new interrupts generated for this engine.
(So couple that also with a sync_irq.) We've been kicking around that
plan for yonks, since the beginning of doing per-engine reset.
-Chris