[RFC PATCH 0/4] use MemTxAttrs to signal cpu index

[Alex Bennée]

Hi,

This ostensibly fixes a bug with gdbstub when accessing the GIC state
(via system registers). It also proposes a pattern for removing
current_cpu/qtest_enabled() hacks in hw/ by passing the cpu index via
MemTxAttrs. For the ARM timer code we also assert that those accesses
do come from a CPU (as opposed to HW DMA attempting to overwrite
stuff).

We do need to audit helpers in target/ which use the address_space API
to read and write data and ensure they properly form their MxTxAttrs
so this works cleanly.

What do people think?

Alex Bennée (4):
  hw: encode accessing CPU index in MemTxAttrs
  qtest: make read/write operation appear to be from CPU
  hw/intc/gic: use MxTxAttrs to divine accessing CPU
  hw/timer: convert mptimer access to attrs to derive cpu index

 include/exec/memattrs.h |  4 +++-
 accel/tcg/cputlb.c      | 22 ++++++++++++++++------
 hw/core/cpu-sysemu.c    | 17 +++++++++++++----
 hw/intc/arm_gic.c       | 39 ++++++++++++++++++++++-----------------
 hw/timer/arm_mptimer.c  | 25 ++++++++++++++-----------
 softmmu/qtest.c         | 31 +++++++++++++++++++------------
 6 files changed, 87 insertions(+), 51 deletions(-)

-- 
2.34.1