On 8/26/19 4:21 PM, Tony Nguyen wrote:
This series is an attempt to re-declare devices with DEVICE_NATIVE_ENDIAN as
DEVICE_LITTLE_ENDIAN or DEVICE_BIG_ENDIAN.
This clean up was split out from an earlier series which collapsed byte swaps
along the I/O path.
On Wed, 7 Aug 2019 at 12:42, Paolo Bonzini wrote:
On 07/08/19 10:31, tony.nguyen@bt wrote:
Device realizing code with MemorRegionOps endianness as
DEVICE_NATIVE_ENDIAN is not common code.
Corrected devices were identified by making the declaration of
DEVICE_NATIVE_ENDIAN conditional upon NEED_CPU_H and then listing
what failed to compile.
The general approach makes sense. However, most of these should not be
DEVICE_NATIVE_ENDIAN. I can help with some of them.
On Fri, 16 Aug 2019 at 13:02, Peter Maydell wrote:
OTOH it's worth noting that it's quite likely that most of
the implementations of these DEVICE_NATIVE_ENDIAN devices
picked it in an equally naive way, by just copying some other
device's code...
Approach this in two steps.
1. Naively. For each device declared with DEVICE_NATIVE_ENDIAN, find the set of
targets from the set of target/hw/*/device.o.
If the set of targets are all little or all big endian, re-declare
as DEVICE_LITTLE_ENDIAN or DEVICE_BIG_ENDIAN respectively.
2. Manually. Inspect with heuristics (thanks Paolo):
- if not used, re-declare as DEVICE_HOST_ENDIAN.
- if max/min size=1, re-declare as DEVICE_HOST_ENDIAN.
- if just a bit bucket, re-declare as DEVICE_HOST_ENDIAN
- if PCI, re-declare as DEVICE_LITTLE_ENDIAN.
- if for {ARM|unicore32} only, re-declare as DEVICE_LITTLE_ENDIAN.
- if for SPARC only, re-declare as DEVICE_BIG_ENDIAN.
Maybe this list is better to split your patches.
Which are the 'not used' ones?
TYPE_TPCI200 (from IndustryPack) can go with PCI.
You can put all the PrimeCell devices together (eventually with the
other ARM devices):
hw/audio/pl041.c
hw/char/pl011.c
hw/display/pl110.c
hw/dma/pl080.c
hw/dma/pl330.c
hw/gpio/pl061.c
hw/input/pl050.c
hw/intc/pl190.c
hw/sd/pl181.c
hw/ssi/pl022.c
hw/timer/pl031.c
