[PATCH 00/18] dbtr: Fixes and heterogeneous trigger types

[Nicholas Piggin]

In the course of implementing and testing heterogeneous trigger
types as found in the Tenstorrent Ascalon CPU, I found some issues
to fix as well.

Patches 1-11 contain fixes and improvements for existing function,
and make most of the new kvm-unit-tests tests pass. 12-17 implement
"heterogenous" triggers where different HW triggers support
different functions so we have to relax the 1:1 mapping from DBTR trig
to HW trigger index and allocate HW triggers independently from what
can satisfy the requested type. Patch 18 is another possible fix but
it relies on update to SBI spec so I put it last. I can move it ahead
into the fixes block if the issue is confirmed.

Thanks,
Nick

Nicholas Piggin (18):
  dbtr: Add consistent range checks to trigger ecalls
  dbtr: Trigger update should set sbiret.value on failure
  dbtr: Fix endian conversion in trigger install handler
  dbtr: Return correct error on install not supported
  dbtr: Do not support chain bit
  dbtr: Improve trigger update error checking
  dbtr: Check for invalid and unsupported triggers in update
  dbtr: Improve error handling for trigger enable, disable, uninstall
  dbtr: Read triggers should not read HW trigger if not mapped
  dbtr: Avoid crash in sbi_debug_read_triggers
  dbtr: Succeed operations with no triggers in mask
  dbtr: Move hardware trigger probing to a function
  dbtr: Rework install and update error handling
  dbtr: Decouple dbtr trigger index from hardware trigger number
  dbtr: Move trigger feature support test into a function
  dbtr: Heterogeneous trigger type support
  dbtr: Heterogeneous access type matching for mcontrol triggers
  dbtr: Work around specification bug in range checks

 include/sbi/sbi_dbtr.h   |  17 +-
 lib/sbi/sbi_dbtr.c       | 499 ++++++++++++++++++++++++++++++---------
 lib/sbi/sbi_ecall_dbtr.c |   2 +-
 3 files changed, 407 insertions(+), 111 deletions(-)

-- 
2.51.0